mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-01-18 04:53:01 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/sql/table.cc

4926 lines
146 KiB
C++
Raw Normal View History

WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
/* Copyright 2000-2008 MySQL AB, 2008-2009 Sun Microsystems, Inc.
Update copyright Fixed memory leak on shutdown (Affects the embedded version & MyODBC)
2001-12-06 13:10:51 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
Many files: Changed header to GPL version 2 only
2006-12-23 20:17:15 +01:00
the Free Software Foundation; version 2 of the License.
Update copyright Fixed memory leak on shutdown (Affects the embedded version & MyODBC)
2001-12-06 13:10:51 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
Update copyright Fixed memory leak on shutdown (Affects the embedded version & MyODBC)
2001-12-06 13:10:51 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
/* Some general useful functions */
WL#5030: Splitting mysql_priv.h Adding my_global.h first in all files using NO_EMBEDDED_ACCESS_CHECKS. Correcting a merge problem resulting from a changed definition of check_some_access compared to the original patches.
2010-04-07 13:58:40 +02:00
#include "my_global.h" /* NO_EMBEDDED_ACCESS_CHECKS */
WL#5030: Split and remove mysql_priv.h This patch: - Moves all definitions from the mysql_priv.h file into header files for the component where the variable is defined - Creates header files if the component lacks one - Eliminates all include directives from mysql_priv.h - Eliminates all circular include cycles - Rename time.cc to sql_time.cc - Rename mysql_priv.h to sql_priv.h
2010-03-31 16:05:33 +02:00
#include "sql_priv.h"
#include "unireg.h" // REQUIRED: for other includes
#include "table.h"
#include "frm_crypt.h" // get_crypt_for_frm
#include "key.h" // find_ref_key
#include "sql_table.h" // build_table_filename,
// primary_key_name
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
#include "sql_trigger.h"
WL#5030: Split and remove mysql_priv.h This patch: - Moves all definitions from the mysql_priv.h file into header files for the component where the variable is defined - Creates header files if the component lacks one - Eliminates all include directives from mysql_priv.h - Eliminates all circular include cycles - Rename time.cc to sql_time.cc - Rename mysql_priv.h to sql_priv.h
2010-03-31 16:05:33 +02:00
#include "sql_parse.h" // free_items
#include "strfunc.h" // unhex_type2
#include "sql_partition.h" // mysql_unpack_partition,
// fix_partition_func, partition_info
#include "sql_acl.h" // *_ACL, acl_getroot_no_password
#include "sql_base.h" // release_table_share
Import changeset
2000-07-31 21:29:14 +02:00
#include <m_ctype.h>
Avoid the name conflict between the system-provided "md5.h" and the MySQL one by renaming "include/md5.h" to "include/my_md5.h". Fixes bug#14151.
2007-07-19 14:14:03 +02:00
#include "my_md5.h"
Import changeset
2000-07-31 21:29:14 +02:00
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
/* INFORMATION_SCHEMA name */
LEX_STRING INFORMATION_SCHEMA_NAME= {C_STRING_WITH_LEN("information_schema")};
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
/* PERFORMANCE_SCHEMA name */
LEX_STRING PERFORMANCE_SCHEMA_DB_NAME= {C_STRING_WITH_LEN("performance_schema")};
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
/* MYSQL_SCHEMA name */
LEX_STRING MYSQL_SCHEMA_NAME= {C_STRING_WITH_LEN("mysql")};
/* GENERAL_LOG name */
LEX_STRING GENERAL_LOG_NAME= {C_STRING_WITH_LEN("general_log")};
/* SLOW_LOG name */
LEX_STRING SLOW_LOG_NAME= {C_STRING_WITH_LEN("slow_log")};
Import changeset
2000-07-31 21:29:14 +02:00
/* Functions defined in this file */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
void open_table_error(TABLE_SHARE *share, int error, int db_errno,
myf errortype, int errarg);
fixed reading of free'd memory when opening a table with non-existing plugin
2006-04-06 15:49:00 +02:00
static int open_binary_frm(THD *thd, TABLE_SHARE *share,
uchar *head, File file);
Import changeset
2000-07-31 21:29:14 +02:00
static void fix_type_pointers(const char ***array, TYPELIB *point_to_type,
uint types, char **names);
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
static uint find_field(Field **fields, uchar *record, uint start, uint length);
Import changeset
2000-07-31 21:29:14 +02:00
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
inline bool is_system_table_name(const char *name, uint length);
Import changeset
2000-07-31 21:29:14 +02:00
Patch for the following bugs: - BUG#11986: Stored routines and triggers can fail if the code has a non-ascii symbol - BUG#16291: mysqldump corrupts string-constants with non-ascii-chars - BUG#19443: INFORMATION_SCHEMA does not support charsets properly - BUG#21249: Character set of SP-var can be ignored - BUG#25212: Character set of string constant is ignored (stored routines) - BUG#25221: Character set of string constant is ignored (triggers) There were a few general problems that caused these bugs: 1. Character set information of the original (definition) query for views, triggers, stored routines and events was lost. 2. mysqldump output query in client character set, which can be inappropriate to encode definition-query. 3. INFORMATION_SCHEMA used strings with mixed encodings to display object definition; 1. No query-definition-character set. In order to compile query into execution code, some extra data (such as environment variables or the database character set) is used. The problem here was that this context was not preserved. So, on the next load it can differ from the original one, thus the result will be different. The context contains the following data: - client character set; - connection collation (character set and collation); - collation of the owner database; The fix is to store this context and use it each time we parse (compile) and execute the object (stored routine, trigger, ...). 2. Wrong mysqldump-output. The original query can contain several encodings (by means of character set introducers). The problem here was that we tried to convert original query to the mysqldump-client character set. Moreover, we stored queries in different character sets for different objects (views, for one, used UTF8, triggers used original character set). The solution is - to store definition queries in the original character set; - to change SHOW CREATE statement to output definition query in the binary character set (i.e. without any conversion); - introduce SHOW CREATE TRIGGER statement; - to dump special statements to switch the context to the original one before dumping and restore it afterwards. Note, in order to preserve the database collation at the creation time, additional ALTER DATABASE might be used (to temporary switch the database collation back to the original value). In this case, ALTER DATABASE privilege will be required. This is a backward-incompatible change. 3. INFORMATION_SCHEMA showed non-UTF8 strings The fix is to generate UTF8-query during the parsing, store it in the object and show it in the INFORMATION_SCHEMA. Basically, the idea is to create a copy of the original query convert it to UTF8. Character set introducers are removed and all text literals are converted to UTF8. This UTF8 query is intended to provide user-readable output. It must not be used to recreate the object. Specialized SHOW CREATE statements should be used for this. The reason for this limitation is the following: the original query can contain symbols from several character sets (by means of character set introducers). Example: - original query: CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1; - UTF8 query (for INFORMATION_SCHEMA): CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
/**************************************************************************
Object_creation_ctx implementation.
**************************************************************************/
Object_creation_ctx *Object_creation_ctx::set_n_backup(THD *thd)
{
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
Object_creation_ctx *backup_ctx;
DBUG_ENTER("Object_creation_ctx::set_n_backup");
Patch for the following bugs: - BUG#11986: Stored routines and triggers can fail if the code has a non-ascii symbol - BUG#16291: mysqldump corrupts string-constants with non-ascii-chars - BUG#19443: INFORMATION_SCHEMA does not support charsets properly - BUG#21249: Character set of SP-var can be ignored - BUG#25212: Character set of string constant is ignored (stored routines) - BUG#25221: Character set of string constant is ignored (triggers) There were a few general problems that caused these bugs: 1. Character set information of the original (definition) query for views, triggers, stored routines and events was lost. 2. mysqldump output query in client character set, which can be inappropriate to encode definition-query. 3. INFORMATION_SCHEMA used strings with mixed encodings to display object definition; 1. No query-definition-character set. In order to compile query into execution code, some extra data (such as environment variables or the database character set) is used. The problem here was that this context was not preserved. So, on the next load it can differ from the original one, thus the result will be different. The context contains the following data: - client character set; - connection collation (character set and collation); - collation of the owner database; The fix is to store this context and use it each time we parse (compile) and execute the object (stored routine, trigger, ...). 2. Wrong mysqldump-output. The original query can contain several encodings (by means of character set introducers). The problem here was that we tried to convert original query to the mysqldump-client character set. Moreover, we stored queries in different character sets for different objects (views, for one, used UTF8, triggers used original character set). The solution is - to store definition queries in the original character set; - to change SHOW CREATE statement to output definition query in the binary character set (i.e. without any conversion); - introduce SHOW CREATE TRIGGER statement; - to dump special statements to switch the context to the original one before dumping and restore it afterwards. Note, in order to preserve the database collation at the creation time, additional ALTER DATABASE might be used (to temporary switch the database collation back to the original value). In this case, ALTER DATABASE privilege will be required. This is a backward-incompatible change. 3. INFORMATION_SCHEMA showed non-UTF8 strings The fix is to generate UTF8-query during the parsing, store it in the object and show it in the INFORMATION_SCHEMA. Basically, the idea is to create a copy of the original query convert it to UTF8. Character set introducers are removed and all text literals are converted to UTF8. This UTF8 query is intended to provide user-readable output. It must not be used to recreate the object. Specialized SHOW CREATE statements should be used for this. The reason for this limitation is the following: the original query can contain symbols from several character sets (by means of character set introducers). Example: - original query: CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1; - UTF8 query (for INFORMATION_SCHEMA): CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
backup_ctx= create_backup_ctx(thd);
Patch for the following bugs: - BUG#11986: Stored routines and triggers can fail if the code has a non-ascii symbol - BUG#16291: mysqldump corrupts string-constants with non-ascii-chars - BUG#19443: INFORMATION_SCHEMA does not support charsets properly - BUG#21249: Character set of SP-var can be ignored - BUG#25212: Character set of string constant is ignored (stored routines) - BUG#25221: Character set of string constant is ignored (triggers) There were a few general problems that caused these bugs: 1. Character set information of the original (definition) query for views, triggers, stored routines and events was lost. 2. mysqldump output query in client character set, which can be inappropriate to encode definition-query. 3. INFORMATION_SCHEMA used strings with mixed encodings to display object definition; 1. No query-definition-character set. In order to compile query into execution code, some extra data (such as environment variables or the database character set) is used. The problem here was that this context was not preserved. So, on the next load it can differ from the original one, thus the result will be different. The context contains the following data: - client character set; - connection collation (character set and collation); - collation of the owner database; The fix is to store this context and use it each time we parse (compile) and execute the object (stored routine, trigger, ...). 2. Wrong mysqldump-output. The original query can contain several encodings (by means of character set introducers). The problem here was that we tried to convert original query to the mysqldump-client character set. Moreover, we stored queries in different character sets for different objects (views, for one, used UTF8, triggers used original character set). The solution is - to store definition queries in the original character set; - to change SHOW CREATE statement to output definition query in the binary character set (i.e. without any conversion); - introduce SHOW CREATE TRIGGER statement; - to dump special statements to switch the context to the original one before dumping and restore it afterwards. Note, in order to preserve the database collation at the creation time, additional ALTER DATABASE might be used (to temporary switch the database collation back to the original value). In this case, ALTER DATABASE privilege will be required. This is a backward-incompatible change. 3. INFORMATION_SCHEMA showed non-UTF8 strings The fix is to generate UTF8-query during the parsing, store it in the object and show it in the INFORMATION_SCHEMA. Basically, the idea is to create a copy of the original query convert it to UTF8. Character set introducers are removed and all text literals are converted to UTF8. This UTF8 query is intended to provide user-readable output. It must not be used to recreate the object. Specialized SHOW CREATE statements should be used for this. The reason for this limitation is the following: the original query can contain symbols from several character sets (by means of character set introducers). Example: - original query: CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1; - UTF8 query (for INFORMATION_SCHEMA): CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
change_env(thd);
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
DBUG_RETURN(backup_ctx);
Patch for the following bugs: - BUG#11986: Stored routines and triggers can fail if the code has a non-ascii symbol - BUG#16291: mysqldump corrupts string-constants with non-ascii-chars - BUG#19443: INFORMATION_SCHEMA does not support charsets properly - BUG#21249: Character set of SP-var can be ignored - BUG#25212: Character set of string constant is ignored (stored routines) - BUG#25221: Character set of string constant is ignored (triggers) There were a few general problems that caused these bugs: 1. Character set information of the original (definition) query for views, triggers, stored routines and events was lost. 2. mysqldump output query in client character set, which can be inappropriate to encode definition-query. 3. INFORMATION_SCHEMA used strings with mixed encodings to display object definition; 1. No query-definition-character set. In order to compile query into execution code, some extra data (such as environment variables or the database character set) is used. The problem here was that this context was not preserved. So, on the next load it can differ from the original one, thus the result will be different. The context contains the following data: - client character set; - connection collation (character set and collation); - collation of the owner database; The fix is to store this context and use it each time we parse (compile) and execute the object (stored routine, trigger, ...). 2. Wrong mysqldump-output. The original query can contain several encodings (by means of character set introducers). The problem here was that we tried to convert original query to the mysqldump-client character set. Moreover, we stored queries in different character sets for different objects (views, for one, used UTF8, triggers used original character set). The solution is - to store definition queries in the original character set; - to change SHOW CREATE statement to output definition query in the binary character set (i.e. without any conversion); - introduce SHOW CREATE TRIGGER statement; - to dump special statements to switch the context to the original one before dumping and restore it afterwards. Note, in order to preserve the database collation at the creation time, additional ALTER DATABASE might be used (to temporary switch the database collation back to the original value). In this case, ALTER DATABASE privilege will be required. This is a backward-incompatible change. 3. INFORMATION_SCHEMA showed non-UTF8 strings The fix is to generate UTF8-query during the parsing, store it in the object and show it in the INFORMATION_SCHEMA. Basically, the idea is to create a copy of the original query convert it to UTF8. Character set introducers are removed and all text literals are converted to UTF8. This UTF8 query is intended to provide user-readable output. It must not be used to recreate the object. Specialized SHOW CREATE statements should be used for this. The reason for this limitation is the following: the original query can contain symbols from several character sets (by means of character set introducers). Example: - original query: CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1; - UTF8 query (for INFORMATION_SCHEMA): CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
}
void Object_creation_ctx::restore_env(THD *thd, Object_creation_ctx *backup_ctx)
{
if (!backup_ctx)
return;
backup_ctx->change_env(thd);
delete backup_ctx;
}
/**************************************************************************
Default_object_creation_ctx implementation.
**************************************************************************/
Default_object_creation_ctx::Default_object_creation_ctx(THD *thd)
: m_client_cs(thd->variables.character_set_client),
m_connection_cl(thd->variables.collation_connection)
{ }
Default_object_creation_ctx::Default_object_creation_ctx(
CHARSET_INFO *client_cs, CHARSET_INFO *connection_cl)
: m_client_cs(client_cs),
m_connection_cl(connection_cl)
{ }
Object_creation_ctx *
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
Default_object_creation_ctx::create_backup_ctx(THD *thd) const
Patch for the following bugs: - BUG#11986: Stored routines and triggers can fail if the code has a non-ascii symbol - BUG#16291: mysqldump corrupts string-constants with non-ascii-chars - BUG#19443: INFORMATION_SCHEMA does not support charsets properly - BUG#21249: Character set of SP-var can be ignored - BUG#25212: Character set of string constant is ignored (stored routines) - BUG#25221: Character set of string constant is ignored (triggers) There were a few general problems that caused these bugs: 1. Character set information of the original (definition) query for views, triggers, stored routines and events was lost. 2. mysqldump output query in client character set, which can be inappropriate to encode definition-query. 3. INFORMATION_SCHEMA used strings with mixed encodings to display object definition; 1. No query-definition-character set. In order to compile query into execution code, some extra data (such as environment variables or the database character set) is used. The problem here was that this context was not preserved. So, on the next load it can differ from the original one, thus the result will be different. The context contains the following data: - client character set; - connection collation (character set and collation); - collation of the owner database; The fix is to store this context and use it each time we parse (compile) and execute the object (stored routine, trigger, ...). 2. Wrong mysqldump-output. The original query can contain several encodings (by means of character set introducers). The problem here was that we tried to convert original query to the mysqldump-client character set. Moreover, we stored queries in different character sets for different objects (views, for one, used UTF8, triggers used original character set). The solution is - to store definition queries in the original character set; - to change SHOW CREATE statement to output definition query in the binary character set (i.e. without any conversion); - introduce SHOW CREATE TRIGGER statement; - to dump special statements to switch the context to the original one before dumping and restore it afterwards. Note, in order to preserve the database collation at the creation time, additional ALTER DATABASE might be used (to temporary switch the database collation back to the original value). In this case, ALTER DATABASE privilege will be required. This is a backward-incompatible change. 3. INFORMATION_SCHEMA showed non-UTF8 strings The fix is to generate UTF8-query during the parsing, store it in the object and show it in the INFORMATION_SCHEMA. Basically, the idea is to create a copy of the original query convert it to UTF8. Character set introducers are removed and all text literals are converted to UTF8. This UTF8 query is intended to provide user-readable output. It must not be used to recreate the object. Specialized SHOW CREATE statements should be used for this. The reason for this limitation is the following: the original query can contain symbols from several character sets (by means of character set introducers). Example: - original query: CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1; - UTF8 query (for INFORMATION_SCHEMA): CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
{
return new Default_object_creation_ctx(thd);
}
void Default_object_creation_ctx::change_env(THD *thd) const
{
thd->variables.character_set_client= m_client_cs;
thd->variables.collation_connection= m_connection_cl;
thd->update_charset();
}
/**************************************************************************
View_creation_ctx implementation.
**************************************************************************/
View_creation_ctx *View_creation_ctx::create(THD *thd)
{
View_creation_ctx *ctx= new (thd->mem_root) View_creation_ctx(thd);
return ctx;
}
/*************************************************************************/
View_creation_ctx * View_creation_ctx::create(THD *thd,
Post-merge fixes (merge from the main).
2007-07-16 22:59:21 +02:00
TABLE_LIST *view)
Patch for the following bugs: - BUG#11986: Stored routines and triggers can fail if the code has a non-ascii symbol - BUG#16291: mysqldump corrupts string-constants with non-ascii-chars - BUG#19443: INFORMATION_SCHEMA does not support charsets properly - BUG#21249: Character set of SP-var can be ignored - BUG#25212: Character set of string constant is ignored (stored routines) - BUG#25221: Character set of string constant is ignored (triggers) There were a few general problems that caused these bugs: 1. Character set information of the original (definition) query for views, triggers, stored routines and events was lost. 2. mysqldump output query in client character set, which can be inappropriate to encode definition-query. 3. INFORMATION_SCHEMA used strings with mixed encodings to display object definition; 1. No query-definition-character set. In order to compile query into execution code, some extra data (such as environment variables or the database character set) is used. The problem here was that this context was not preserved. So, on the next load it can differ from the original one, thus the result will be different. The context contains the following data: - client character set; - connection collation (character set and collation); - collation of the owner database; The fix is to store this context and use it each time we parse (compile) and execute the object (stored routine, trigger, ...). 2. Wrong mysqldump-output. The original query can contain several encodings (by means of character set introducers). The problem here was that we tried to convert original query to the mysqldump-client character set. Moreover, we stored queries in different character sets for different objects (views, for one, used UTF8, triggers used original character set). The solution is - to store definition queries in the original character set; - to change SHOW CREATE statement to output definition query in the binary character set (i.e. without any conversion); - introduce SHOW CREATE TRIGGER statement; - to dump special statements to switch the context to the original one before dumping and restore it afterwards. Note, in order to preserve the database collation at the creation time, additional ALTER DATABASE might be used (to temporary switch the database collation back to the original value). In this case, ALTER DATABASE privilege will be required. This is a backward-incompatible change. 3. INFORMATION_SCHEMA showed non-UTF8 strings The fix is to generate UTF8-query during the parsing, store it in the object and show it in the INFORMATION_SCHEMA. Basically, the idea is to create a copy of the original query convert it to UTF8. Character set introducers are removed and all text literals are converted to UTF8. This UTF8 query is intended to provide user-readable output. It must not be used to recreate the object. Specialized SHOW CREATE statements should be used for this. The reason for this limitation is the following: the original query can contain symbols from several character sets (by means of character set introducers). Example: - original query: CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1; - UTF8 query (for INFORMATION_SCHEMA): CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
{
View_creation_ctx *ctx= new (thd->mem_root) View_creation_ctx(thd);
/* Throw a warning if there is NULL cs name. */
if (!view->view_client_cs_name.str ||
!view->view_connection_cl_name.str)
{
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_NOTE,
ER_VIEW_NO_CREATION_CTX,
ER(ER_VIEW_NO_CREATION_CTX),
(const char *) view->db,
(const char *) view->table_name);
ctx->m_client_cs= system_charset_info;
ctx->m_connection_cl= system_charset_info;
return ctx;
}
/* Resolve cs names. Throw a warning if there is unknown cs name. */
bool invalid_creation_ctx;
invalid_creation_ctx= resolve_charset(view->view_client_cs_name.str,
system_charset_info,
&ctx->m_client_cs);
invalid_creation_ctx= resolve_collation(view->view_connection_cl_name.str,
system_charset_info,
&ctx->m_connection_cl) ||
invalid_creation_ctx;
if (invalid_creation_ctx)
{
sql_print_warning("View '%s'.'%s': there is unknown charset/collation "
"names (client: '%s'; connection: '%s').",
(const char *) view->db,
(const char *) view->table_name,
(const char *) view->view_client_cs_name.str,
(const char *) view->view_connection_cl_name.str);
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_NOTE,
ER_VIEW_INVALID_CREATION_CTX,
ER(ER_VIEW_INVALID_CREATION_CTX),
(const char *) view->db,
(const char *) view->table_name);
}
return ctx;
}
/*************************************************************************/
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/* Get column name from column hash */
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
static uchar *get_field_name(Field **buff, size_t *length,
my_bool not_used __attribute__((unused)))
Import changeset
2000-07-31 21:29:14 +02:00
{
"Fix" for BUG #2050 "10 to 1 performance drop with server 4.1.1". Actually it is not a bug but right behavior observed as pefomance degradation after we have forced Item_field::fix_fields() to re-execute each time when we are executing prep stmt. This patch implements small optimization which heals this bad behavior. We are caching field position in TABLE::field array in Item's member and are using this position for speeding up field lookups in fix_fields() in case of its re-execution.
2004-03-28 01:11:54 +01:00
*length= (uint) strlen((*buff)->field_name);
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
return (uchar*) (*buff)->field_name;
Import changeset
2000-07-31 21:29:14 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
WL#1324 table name to file name encoding - Encoding itself, implemented as a charset "filename". Originally planned to use '.' as an escape character, but now changed to '@' for two reasons: "ls" does not return file names starting with '.' considering them as a kind of hidden files; some platforms do not allow several dots in a file name. - replacing many calls of my_snprintf() and strnxmov() to the new build_table_filename(). - Adding MY_APPEND_EXT mysys flag, to append an extention rather that replace it. - Replacing all numeric constants in fn_format flag arguments to their mysys definitions, e.g. MY_UNPACK_FILENAME, - Predictability in several function/methods: when a table name can appear with or withot .frm extension. Some functions/methods were changed so accept names strictly with .frm, other - strictly without .frm extensions. Several DBUG_ASSERTs were added to check whether an extension is passed. Many files: table name to file name encoding mysql_priv.h: Prototypes for new table name encoding tools. ctype-utf8.c: Implementing "filename" charset for table name to file name encoding. row0mysql.c: Fixing table name prefix. mf_format.c: Adding MY_APPEND_EXT processing. Many files: Fixing tests. my_sys.h: Adding new flag to append rather than replace an extension. m_ctype.h: Adding "filename" charset definition.
2005-12-31 06:01:26 +01:00
/*
Returns pointer to '.frm' extension of the file name.
SYNOPSIS
fn_rext()
name file name
DESCRIPTION
Checks file name part starting with the rightmost '.' character,
and returns it if it is equal to '.frm'.
TODO
It is a good idea to get rid of this function modifying the code
to garantee that the functions presently calling fn_rext() always
get arguments in the same format: either with '.frm' or without '.frm'.
RETURN VALUES
Pointer to the '.frm' extension. If there is no extension,
or extension is not '.frm', pointer at the end of file name.
*/
char *fn_rext(char *name)
{
char *res= strrchr(name, '.');
Bug#18775 - Temporary table from alter table visible to other threads Continued implementation of WL#1324 (table name to filename encoding) The intermediate (not temporary) files of the new table during ALTER TABLE was visible for SHOW TABLES. These intermediate files are copies of the original table with the changes done by ALTER TABLE. After all the data is copied over from the original table, these files are renamed to the original tables file names. So they are not temporary files. They persist after ALTER TABLE, but just with another name. In 5.0 the intermediate files are invisible for SHOW TABLES because all file names beginning with "#sql" were suppressed. This failed since 5.1.6 because even temporary table names were converted when making file names from them. The prefix became converted to "@0023sql". Converting the prefix during SHOW TABLES would suppress the listing of user tables that start with "#sql". The solution of the problem is to continue the implementation of the table name to file name conversion feature. One requirement is to suppress the conversion for temporary table names. This change is straightforward for real temporary tables as there is a function that creates temporary file names. But the generated path names are located in TMPDIR and have no relation to the internal table name. This cannot be used for ALTER TABLE. Its intermediate files need to be in the same directory as the old table files. And it is necessary to be able to deduce the same path from the same table name repeatedly. Consequently the intermediate table files must be handled like normal tables. Their internal names shall start with tmp_file_prefix (#sql) and they shall not be converted like normal table names. I added a flags parameter to all relevant functions that are called from ALTER TABLE. It is used to suppress the conversion for the intermediate table files. The outcome is that the suppression of #sql in SHOW TABLES works again. It does not suppress user tables as these are converted to @0023sql on file level. This patch does also fix ALTER TABLE ... RENAME, which could not rename a table with non-ASCII characters in its name. It does also fix the problem that a user could create a table like `#sql-xxxx-yyyy`, where xxxx is mysqld's pid and yyyy is the thread ID of some other thread, which prevented this thread from running ALTER TABLE. Some of the above problems are mentioned in Bug 1405, which can be closed with this patch. This patch does also contain some minor fixes for other forgotten conversions. Still known problems are reported as bugs 21370, 21373, and 21387.
2006-08-02 17:57:06 +02:00
if (res && !strcmp(res, reg_ext))
WL#1324 table name to file name encoding - Encoding itself, implemented as a charset "filename". Originally planned to use '.' as an escape character, but now changed to '@' for two reasons: "ls" does not return file names starting with '.' considering them as a kind of hidden files; some platforms do not allow several dots in a file name. - replacing many calls of my_snprintf() and strnxmov() to the new build_table_filename(). - Adding MY_APPEND_EXT mysys flag, to append an extention rather that replace it. - Replacing all numeric constants in fn_format flag arguments to their mysys definitions, e.g. MY_UNPACK_FILENAME, - Predictability in several function/methods: when a table name can appear with or withot .frm extension. Some functions/methods were changed so accept names strictly with .frm, other - strictly without .frm extensions. Several DBUG_ASSERTs were added to check whether an extension is passed. Many files: table name to file name encoding mysql_priv.h: Prototypes for new table name encoding tools. ctype-utf8.c: Implementing "filename" charset for table name to file name encoding. row0mysql.c: Fixing table name prefix. mf_format.c: Adding MY_APPEND_EXT processing. Many files: Fixing tests. my_sys.h: Adding new flag to append rather than replace an extension. m_ctype.h: Adding "filename" charset definition.
2005-12-31 06:01:26 +01:00
return res;
return name + strlen(name);
}
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
TABLE_CATEGORY get_table_category(const LEX_STRING *db, const LEX_STRING *name)
{
DBUG_ASSERT(db != NULL);
DBUG_ASSERT(name != NULL);
Manual merge from mysql-trunk-merge. Conflicts: - sql/event_data_objects.cc - sql/events.cc - sql/mysql_priv.h - sql/repl_failsafe.cc - sql/sql_parse.cc - sql/sql_show.cc - sql/sql_view.cc - sql/table.cc
2010-01-27 22:10:37 +01:00
if (is_infoschema_db(db->str, db->length))
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
return TABLE_CATEGORY_INFORMATION;
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if ((db->length == PERFORMANCE_SCHEMA_DB_NAME.length) &&
(my_strcasecmp(system_charset_info,
PERFORMANCE_SCHEMA_DB_NAME.str,
db->str) == 0))
return TABLE_CATEGORY_PERFORMANCE;
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
if ((db->length == MYSQL_SCHEMA_NAME.length) &&
(my_strcasecmp(system_charset_info,
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
MYSQL_SCHEMA_NAME.str,
db->str) == 0))
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
{
if (is_system_table_name(name->str, name->length))
return TABLE_CATEGORY_SYSTEM;
if ((name->length == GENERAL_LOG_NAME.length) &&
(my_strcasecmp(system_charset_info,
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
GENERAL_LOG_NAME.str,
name->str) == 0))
return TABLE_CATEGORY_LOG;
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
if ((name->length == SLOW_LOG_NAME.length) &&
(my_strcasecmp(system_charset_info,
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
SLOW_LOG_NAME.str,
name->str) == 0))
return TABLE_CATEGORY_LOG;
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
}
return TABLE_CATEGORY_USER;
}
WL#1324 table name to file name encoding - Encoding itself, implemented as a charset "filename". Originally planned to use '.' as an escape character, but now changed to '@' for two reasons: "ls" does not return file names starting with '.' considering them as a kind of hidden files; some platforms do not allow several dots in a file name. - replacing many calls of my_snprintf() and strnxmov() to the new build_table_filename(). - Adding MY_APPEND_EXT mysys flag, to append an extention rather that replace it. - Replacing all numeric constants in fn_format flag arguments to their mysys definitions, e.g. MY_UNPACK_FILENAME, - Predictability in several function/methods: when a table name can appear with or withot .frm extension. Some functions/methods were changed so accept names strictly with .frm, other - strictly without .frm extensions. Several DBUG_ASSERTs were added to check whether an extension is passed. Many files: table name to file name encoding mysql_priv.h: Prototypes for new table name encoding tools. ctype-utf8.c: Implementing "filename" charset for table name to file name encoding. row0mysql.c: Fixing table name prefix. mf_format.c: Adding MY_APPEND_EXT processing. Many files: Fixing tests. my_sys.h: Adding new flag to append rather than replace an extension. m_ctype.h: Adding "filename" charset definition.
2005-12-31 06:01:26 +01:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/*
Allocate a setup TABLE_SHARE structure
SYNOPSIS
alloc_table_share()
TABLE_LIST Take database and table name from there
key Table cache key (db \0 table_name \0...)
key_length Length of key
RETURN
0 Error (out of memory)
# Share
*/
Fix byte * vs. char * mismatches that break Win32 compilation.
2005-11-29 09:07:21 +01:00
TABLE_SHARE *alloc_table_share(TABLE_LIST *table_list, char *key,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
uint key_length)
{
MEM_ROOT mem_root;
TABLE_SHARE *share;
Fix for bug#19403/12212 "Crash that happens during removing of database name from cache" and #21216 "Simultaneous DROP TABLE and SHOW OPEN TABLES causes server to crash". Crash happened when one ran DROP DATABASE or SHOW OPEN TABLES statements while concurrently doing DROP TABLE (or RENAME TABLE, CREATE TABLE LIKE or any other command that takes name-lock) in other connection. This problem was caused by the fact that table placeholders which were added to table cache in order to obtain name-lock on table had TABLE_SHARE::db and table_name set to 0. Therefore they broke assumption that these members are non-0 for all tables in table cache on which some of our code relies. The fix sets these members for such placeholders to appropriate value making this assumption true again. As attempt to avoid such problems in future we introduce auxiliary TABLE_SHARE::set_table_cache_key() methods which should be used when one wants to set TABLE_SHARE::table_cache_key and which ensure that TABLE_SHARE::table_name/db are set properly. Test cases for these bugs were added to 5.0 test-suite (with 5.0-specific fix for bug #21216).
2006-08-21 17:02:11 +02:00
char *key_buff, *path_buff;
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
char path[FN_REFLEN];
uint path_length;
Bug#18775 - Temporary table from alter table visible to other threads Continued implementation of WL#1324 (table name to filename encoding) The intermediate (not temporary) files of the new table during ALTER TABLE was visible for SHOW TABLES. These intermediate files are copies of the original table with the changes done by ALTER TABLE. After all the data is copied over from the original table, these files are renamed to the original tables file names. So they are not temporary files. They persist after ALTER TABLE, but just with another name. In 5.0 the intermediate files are invisible for SHOW TABLES because all file names beginning with "#sql" were suppressed. This failed since 5.1.6 because even temporary table names were converted when making file names from them. The prefix became converted to "@0023sql". Converting the prefix during SHOW TABLES would suppress the listing of user tables that start with "#sql". The solution of the problem is to continue the implementation of the table name to file name conversion feature. One requirement is to suppress the conversion for temporary table names. This change is straightforward for real temporary tables as there is a function that creates temporary file names. But the generated path names are located in TMPDIR and have no relation to the internal table name. This cannot be used for ALTER TABLE. Its intermediate files need to be in the same directory as the old table files. And it is necessary to be able to deduce the same path from the same table name repeatedly. Consequently the intermediate table files must be handled like normal tables. Their internal names shall start with tmp_file_prefix (#sql) and they shall not be converted like normal table names. I added a flags parameter to all relevant functions that are called from ALTER TABLE. It is used to suppress the conversion for the intermediate table files. The outcome is that the suppression of #sql in SHOW TABLES works again. It does not suppress user tables as these are converted to @0023sql on file level. This patch does also fix ALTER TABLE ... RENAME, which could not rename a table with non-ASCII characters in its name. It does also fix the problem that a user could create a table like `#sql-xxxx-yyyy`, where xxxx is mysqld's pid and yyyy is the thread ID of some other thread, which prevented this thread from running ALTER TABLE. Some of the above problems are mentioned in Bug 1405, which can be closed with this patch. This patch does also contain some minor fixes for other forgotten conversions. Still known problems are reported as bugs 21370, 21373, and 21387.
2006-08-02 17:57:06 +02:00
DBUG_ENTER("alloc_table_share");
DBUG_PRINT("enter", ("table: '%s'.'%s'",
table_list->db, table_list->table_name));
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
WL#1324 table name to file name encoding - Encoding itself, implemented as a charset "filename". Originally planned to use '.' as an escape character, but now changed to '@' for two reasons: "ls" does not return file names starting with '.' considering them as a kind of hidden files; some platforms do not allow several dots in a file name. - replacing many calls of my_snprintf() and strnxmov() to the new build_table_filename(). - Adding MY_APPEND_EXT mysys flag, to append an extention rather that replace it. - Replacing all numeric constants in fn_format flag arguments to their mysys definitions, e.g. MY_UNPACK_FILENAME, - Predictability in several function/methods: when a table name can appear with or withot .frm extension. Some functions/methods were changed so accept names strictly with .frm, other - strictly without .frm extensions. Several DBUG_ASSERTs were added to check whether an extension is passed. Many files: table name to file name encoding mysql_priv.h: Prototypes for new table name encoding tools. ctype-utf8.c: Implementing "filename" charset for table name to file name encoding. row0mysql.c: Fixing table name prefix. mf_format.c: Adding MY_APPEND_EXT processing. Many files: Fixing tests. my_sys.h: Adding new flag to append rather than replace an extension. m_ctype.h: Adding "filename" charset definition.
2005-12-31 06:01:26 +01:00
path_length= build_table_filename(path, sizeof(path) - 1,
table_list->db,
Bug#18775 - Temporary table from alter table visible to other threads Continued implementation of WL#1324 (table name to filename encoding) The intermediate (not temporary) files of the new table during ALTER TABLE was visible for SHOW TABLES. These intermediate files are copies of the original table with the changes done by ALTER TABLE. After all the data is copied over from the original table, these files are renamed to the original tables file names. So they are not temporary files. They persist after ALTER TABLE, but just with another name. In 5.0 the intermediate files are invisible for SHOW TABLES because all file names beginning with "#sql" were suppressed. This failed since 5.1.6 because even temporary table names were converted when making file names from them. The prefix became converted to "@0023sql". Converting the prefix during SHOW TABLES would suppress the listing of user tables that start with "#sql". The solution of the problem is to continue the implementation of the table name to file name conversion feature. One requirement is to suppress the conversion for temporary table names. This change is straightforward for real temporary tables as there is a function that creates temporary file names. But the generated path names are located in TMPDIR and have no relation to the internal table name. This cannot be used for ALTER TABLE. Its intermediate files need to be in the same directory as the old table files. And it is necessary to be able to deduce the same path from the same table name repeatedly. Consequently the intermediate table files must be handled like normal tables. Their internal names shall start with tmp_file_prefix (#sql) and they shall not be converted like normal table names. I added a flags parameter to all relevant functions that are called from ALTER TABLE. It is used to suppress the conversion for the intermediate table files. The outcome is that the suppression of #sql in SHOW TABLES works again. It does not suppress user tables as these are converted to @0023sql on file level. This patch does also fix ALTER TABLE ... RENAME, which could not rename a table with non-ASCII characters in its name. It does also fix the problem that a user could create a table like `#sql-xxxx-yyyy`, where xxxx is mysqld's pid and yyyy is the thread ID of some other thread, which prevented this thread from running ALTER TABLE. Some of the above problems are mentioned in Bug 1405, which can be closed with this patch. This patch does also contain some minor fixes for other forgotten conversions. Still known problems are reported as bugs 21370, 21373, and 21387.
2006-08-02 17:57:06 +02:00
table_list->table_name, "", 0);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
init_sql_alloc(&mem_root, TABLE_ALLOC_BLOCK_SIZE, 0);
Fix for bug#19403/12212 "Crash that happens during removing of database name from cache" and #21216 "Simultaneous DROP TABLE and SHOW OPEN TABLES causes server to crash". Crash happened when one ran DROP DATABASE or SHOW OPEN TABLES statements while concurrently doing DROP TABLE (or RENAME TABLE, CREATE TABLE LIKE or any other command that takes name-lock) in other connection. This problem was caused by the fact that table placeholders which were added to table cache in order to obtain name-lock on table had TABLE_SHARE::db and table_name set to 0. Therefore they broke assumption that these members are non-0 for all tables in table cache on which some of our code relies. The fix sets these members for such placeholders to appropriate value making this assumption true again. As attempt to avoid such problems in future we introduce auxiliary TABLE_SHARE::set_table_cache_key() methods which should be used when one wants to set TABLE_SHARE::table_cache_key and which ensure that TABLE_SHARE::table_name/db are set properly. Test cases for these bugs were added to 5.0 test-suite (with 5.0-specific fix for bug #21216).
2006-08-21 17:02:11 +02:00
if (multi_alloc_root(&mem_root,
&share, sizeof(*share),
&key_buff, key_length,
&path_buff, path_length + 1,
NULL))
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
bzero((char*) share, sizeof(*share));
Fix for bug#19403/12212 "Crash that happens during removing of database name from cache" and #21216 "Simultaneous DROP TABLE and SHOW OPEN TABLES causes server to crash". Crash happened when one ran DROP DATABASE or SHOW OPEN TABLES statements while concurrently doing DROP TABLE (or RENAME TABLE, CREATE TABLE LIKE or any other command that takes name-lock) in other connection. This problem was caused by the fact that table placeholders which were added to table cache in order to obtain name-lock on table had TABLE_SHARE::db and table_name set to 0. Therefore they broke assumption that these members are non-0 for all tables in table cache on which some of our code relies. The fix sets these members for such placeholders to appropriate value making this assumption true again. As attempt to avoid such problems in future we introduce auxiliary TABLE_SHARE::set_table_cache_key() methods which should be used when one wants to set TABLE_SHARE::table_cache_key and which ensure that TABLE_SHARE::table_name/db are set properly. Test cases for these bugs were added to 5.0 test-suite (with 5.0-specific fix for bug #21216).
2006-08-21 17:02:11 +02:00
share->set_table_cache_key(key_buff, key, key_length);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Fix for bug#19403/12212 "Crash that happens during removing of database name from cache" and #21216 "Simultaneous DROP TABLE and SHOW OPEN TABLES causes server to crash". Crash happened when one ran DROP DATABASE or SHOW OPEN TABLES statements while concurrently doing DROP TABLE (or RENAME TABLE, CREATE TABLE LIKE or any other command that takes name-lock) in other connection. This problem was caused by the fact that table placeholders which were added to table cache in order to obtain name-lock on table had TABLE_SHARE::db and table_name set to 0. Therefore they broke assumption that these members are non-0 for all tables in table cache on which some of our code relies. The fix sets these members for such placeholders to appropriate value making this assumption true again. As attempt to avoid such problems in future we introduce auxiliary TABLE_SHARE::set_table_cache_key() methods which should be used when one wants to set TABLE_SHARE::table_cache_key and which ensure that TABLE_SHARE::table_name/db are set properly. Test cases for these bugs were added to 5.0 test-suite (with 5.0-specific fix for bug #21216).
2006-08-21 17:02:11 +02:00
share->path.str= path_buff;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->path.length= path_length;
strmov(share->path.str, path);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
share->normalized_path.str= share->path.str;
share->normalized_path.length= path_length;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->version= refresh_version;
Bug#17181 (mysqlslap test server crash): Moving assignments to table_map_id for thread-safe handling of table shares.
2006-02-23 13:34:03 +01:00
/*
Since alloc_table_share() can be called without any locking (for
example, ha_create_table... functions), we do not assign a table
map id here. Instead we assign a value that is not used
elsewhere, and then assign a table map id inside open_table()
under the protection of the LOCK_open mutex.
*/
WL#3023 (Use locks in statement-like manner): Post-merge patches.
2006-03-09 03:56:14 +01:00
share->table_map_id= ~0UL;
Bug#18280 (RBR: Extreneous maps when writing to general_log and slow_log): Filter out replication general_log and slow_log entirely from binary log. Caching result of table share-specific tests.
2006-03-17 18:11:07 +01:00
share->cached_row_logging_check= -1;
Initial import of WL#3726 "DDL locking for all metadata objects". Backport of: ------------------------------------------------------------ revno: 2630.4.1 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Fri 2008-05-23 17:54:03 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ------------------------------------------------------------ This is the first patch in series. It transforms the metadata locking subsystem to use a dedicated module (mdl.h,cc). No significant changes in the locking protocol. The import passes the test suite with the exception of deprecated/removed 6.0 features, and MERGE tables. The latter are subject to a fix by WL#4144. Unfortunately, the original changeset comments got lost in a merge, thus this import has its own (largely insufficient) comments. This patch fixes Bug#25144 "replication / binlog with view breaks". Warning: this patch introduces an incompatible change: Under LOCK TABLES, it's no longer possible to FLUSH a table that was not locked for WRITE. Under LOCK TABLES, it's no longer possible to DROP a table or VIEW that was not locked for WRITE. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.2 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:03:45 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.3 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:08:51 +0400 message: WL#3726 "DDL locking for all metadata objects" Fixed failing Windows builds by adding mdl.cc to the lists of files needed to build server/libmysqld on Windows. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.4 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 21:57:58 +0400 message: WL#3726 "DDL locking for all metadata objects". Fix for assert failures in kill.test which occured when one tried to kill ALTER TABLE statement on merge table while it was waiting in wait_while_table_is_used() for other connections to close this table. These assert failures stemmed from the fact that cleanup code in this case assumed that temporary table representing new version of table was open with adding to THD::temporary_tables list while code which were opening this temporary table wasn't always fulfilling this. This patch changes code that opens new version of table to always do this linking in. It also streamlines cleanup process for cases when error occurs while we have new version of table open. ****** WL#3726 "DDL locking for all metadata objects" Add libmysqld/mdl.cc to .bzrignore. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.6 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sun 2008-05-25 00:33:22 +0400 message: WL#3726 "DDL locking for all metadata objects". Addition to the fix of assert failures in kill.test caused by changes for this worklog. Make sure we close the new table only once.
2009-11-30 16:55:03 +01:00
share->used_tables.empty();
share->free_tables.empty();
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
memcpy((char*) &share->mem_root, (char*) &mem_root, sizeof(mem_root));
Merge next-mr -> next-4284
2010-02-03 14:43:03 +01:00
mysql_mutex_init(key_TABLE_SHARE_LOCK_ha_data,
&share->LOCK_ha_data, MY_MUTEX_INIT_FAST);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
}
Bug#18775 - Temporary table from alter table visible to other threads Continued implementation of WL#1324 (table name to filename encoding) The intermediate (not temporary) files of the new table during ALTER TABLE was visible for SHOW TABLES. These intermediate files are copies of the original table with the changes done by ALTER TABLE. After all the data is copied over from the original table, these files are renamed to the original tables file names. So they are not temporary files. They persist after ALTER TABLE, but just with another name. In 5.0 the intermediate files are invisible for SHOW TABLES because all file names beginning with "#sql" were suppressed. This failed since 5.1.6 because even temporary table names were converted when making file names from them. The prefix became converted to "@0023sql". Converting the prefix during SHOW TABLES would suppress the listing of user tables that start with "#sql". The solution of the problem is to continue the implementation of the table name to file name conversion feature. One requirement is to suppress the conversion for temporary table names. This change is straightforward for real temporary tables as there is a function that creates temporary file names. But the generated path names are located in TMPDIR and have no relation to the internal table name. This cannot be used for ALTER TABLE. Its intermediate files need to be in the same directory as the old table files. And it is necessary to be able to deduce the same path from the same table name repeatedly. Consequently the intermediate table files must be handled like normal tables. Their internal names shall start with tmp_file_prefix (#sql) and they shall not be converted like normal table names. I added a flags parameter to all relevant functions that are called from ALTER TABLE. It is used to suppress the conversion for the intermediate table files. The outcome is that the suppression of #sql in SHOW TABLES works again. It does not suppress user tables as these are converted to @0023sql on file level. This patch does also fix ALTER TABLE ... RENAME, which could not rename a table with non-ASCII characters in its name. It does also fix the problem that a user could create a table like `#sql-xxxx-yyyy`, where xxxx is mysqld's pid and yyyy is the thread ID of some other thread, which prevented this thread from running ALTER TABLE. Some of the above problems are mentioned in Bug 1405, which can be closed with this patch. This patch does also contain some minor fixes for other forgotten conversions. Still known problems are reported as bugs 21370, 21373, and 21387.
2006-08-02 17:57:06 +02:00
DBUG_RETURN(share);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
}
/*
Initialize share for temporary tables
SYNOPSIS
init_tmp_table_share()
Bug#26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Bug 26377 - Deadlock with MERGE and FLUSH TABLE Bug 25038 - Waiting TRUNCATE Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Bug 19627 - temporary merge table locking Bug 27660 - Falcon: merge table possible Bug 30273 - merge tables: Can't lock file (errno: 155) The problems were: Bug 26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table 1. A thread trying to lock a MERGE table performs busy waiting while REPAIR TABLE or a similar table administration task is ongoing on one or more of its MyISAM tables. 2. A thread trying to lock a MERGE table performs busy waiting until all threads that did REPAIR TABLE or similar table administration tasks on one or more of its MyISAM tables in LOCK TABLES segments do UNLOCK TABLES. The difference against problem #1 is that the busy waiting takes place *after* the administration task. It is terminated by UNLOCK TABLES only. 3. Two FLUSH TABLES within a LOCK TABLES segment can invalidate the lock. This does *not* require a MERGE table. The first FLUSH TABLES can be replaced by any statement that requires other threads to reopen the table. In 5.0 and 5.1 a single FLUSH TABLES can provoke the problem. Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Trying DML on a MERGE table, which has a child locked and repaired by another thread, made an infinite loop in the server. Bug 26377 - Deadlock with MERGE and FLUSH TABLE Locking a MERGE table and its children in parent-child order and flushing the child deadlocked the server. Bug 25038 - Waiting TRUNCATE Truncating a MERGE child, while the MERGE table was in use, let the truncate fail instead of waiting for the table to become free. Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Repairing a child of an open MERGE table corrupted the child. It was necessary to FLUSH the child first. Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Flushing and optimizing locked MERGE children crashed the server. Bug 19627 - temporary merge table locking Use of a temporary MERGE table with non-temporary children could corrupt the children. Temporary tables are never locked. So we do now prohibit non-temporary chidlren of a temporary MERGE table. Bug 27660 - Falcon: merge table possible It was possible to create a MERGE table with non-MyISAM children. Bug 30273 - merge tables: Can't lock file (errno: 155) This was a Windows-only bug. Table administration statements sometimes failed with "Can't lock file (errno: 155)". These bugs are fixed by a new implementation of MERGE table open. When opening a MERGE table in open_tables() we do now add the child tables to the list of tables to be opened by open_tables() (the "query_list"). The children are not opened in the handler at this stage. After opening the parent, open_tables() opens each child from the now extended query_list. When the last child is opened, we remove the children from the query_list again and attach the children to the parent. This behaves similar to the old open. However it does not open the MyISAM tables directly, but grabs them from the already open children. When closing a MERGE table in close_thread_table() we detach the children only. Closing of the children is done implicitly because they are in thd->open_tables. For more detail see the comment at the top of ha_myisammrg.cc. Changed from open_ltable() to open_and_lock_tables() in all places that can be relevant for MERGE tables. The latter can handle tables added to the list on the fly. When open_ltable() was used in a loop over a list of tables, the list must be temporarily terminated after every table for open_and_lock_tables(). table_list->required_type is set to FRMTYPE_TABLE to avoid open of special tables. Handling of derived tables is suppressed. These details are handled by the new function open_n_lock_single_table(), which has nearly the same signature as open_ltable() and can replace it in most cases. In reopen_tables() some of the tables open by a thread can be closed and reopened. When a MERGE child is affected, the parent must be closed and reopened too. Closing of the parent is forced before the first child is closed. Reopen happens in the order of thd->open_tables. MERGE parents do not attach their children automatically at open. This is done after all tables are reopened. So all children are open when attaching them. Special lock handling like mysql_lock_abort() or mysql_lock_remove() needs to be suppressed for MERGE children or forwarded to the parent. This depends on the situation. In loops over all open tables one suppresses child lock handling. When a single table is touched, forwarding is done. Behavioral changes: =================== This patch changes the behavior of temporary MERGE tables. Temporary MERGE must have temporary children. The old behavior was wrong. A temporary table is not locked. Hence even non-temporary children were not locked. See Bug 19627 - temporary merge table locking. You cannot change the union list of a non-temporary MERGE table when LOCK TABLES is in effect. The following does *not* work: CREATE TABLE m1 ... ENGINE=MRG_MYISAM ...; LOCK TABLES t1 WRITE, t2 WRITE, m1 WRITE; ALTER TABLE m1 ... UNION=(t1,t2) ...; However, you can do this with a temporary MERGE table. You cannot create a MERGE table with CREATE ... SELECT, neither as a temporary MERGE table, nor as a non-temporary MERGE table. CREATE TABLE m1 ... ENGINE=MRG_MYISAM ... SELECT ...; Gives error message: table is not BASE TABLE.
2007-11-15 20:25:43 +01:00
thd thread handle
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share Share to fill
key Table_cache_key, as generated from create_table_def_key.
must start with db name.
key_length Length of key
table_name Table name
path Path to file (possible in lower case) without .frm
NOTES
This is different from alloc_table_share() because temporary tables
don't have to be shared between threads or put into the table def
cache, so we can do some things notable simpler and faster
If table is not put in thd->temporary_tables (happens only when
one uses OPEN TEMPORARY) then one can specify 'db' as key and
use key_length= 0 as neither table_cache_key or key_length will be used).
*/
Bug#26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Bug 26377 - Deadlock with MERGE and FLUSH TABLE Bug 25038 - Waiting TRUNCATE Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Bug 19627 - temporary merge table locking Bug 27660 - Falcon: merge table possible Bug 30273 - merge tables: Can't lock file (errno: 155) The problems were: Bug 26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table 1. A thread trying to lock a MERGE table performs busy waiting while REPAIR TABLE or a similar table administration task is ongoing on one or more of its MyISAM tables. 2. A thread trying to lock a MERGE table performs busy waiting until all threads that did REPAIR TABLE or similar table administration tasks on one or more of its MyISAM tables in LOCK TABLES segments do UNLOCK TABLES. The difference against problem #1 is that the busy waiting takes place *after* the administration task. It is terminated by UNLOCK TABLES only. 3. Two FLUSH TABLES within a LOCK TABLES segment can invalidate the lock. This does *not* require a MERGE table. The first FLUSH TABLES can be replaced by any statement that requires other threads to reopen the table. In 5.0 and 5.1 a single FLUSH TABLES can provoke the problem. Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Trying DML on a MERGE table, which has a child locked and repaired by another thread, made an infinite loop in the server. Bug 26377 - Deadlock with MERGE and FLUSH TABLE Locking a MERGE table and its children in parent-child order and flushing the child deadlocked the server. Bug 25038 - Waiting TRUNCATE Truncating a MERGE child, while the MERGE table was in use, let the truncate fail instead of waiting for the table to become free. Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Repairing a child of an open MERGE table corrupted the child. It was necessary to FLUSH the child first. Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Flushing and optimizing locked MERGE children crashed the server. Bug 19627 - temporary merge table locking Use of a temporary MERGE table with non-temporary children could corrupt the children. Temporary tables are never locked. So we do now prohibit non-temporary chidlren of a temporary MERGE table. Bug 27660 - Falcon: merge table possible It was possible to create a MERGE table with non-MyISAM children. Bug 30273 - merge tables: Can't lock file (errno: 155) This was a Windows-only bug. Table administration statements sometimes failed with "Can't lock file (errno: 155)". These bugs are fixed by a new implementation of MERGE table open. When opening a MERGE table in open_tables() we do now add the child tables to the list of tables to be opened by open_tables() (the "query_list"). The children are not opened in the handler at this stage. After opening the parent, open_tables() opens each child from the now extended query_list. When the last child is opened, we remove the children from the query_list again and attach the children to the parent. This behaves similar to the old open. However it does not open the MyISAM tables directly, but grabs them from the already open children. When closing a MERGE table in close_thread_table() we detach the children only. Closing of the children is done implicitly because they are in thd->open_tables. For more detail see the comment at the top of ha_myisammrg.cc. Changed from open_ltable() to open_and_lock_tables() in all places that can be relevant for MERGE tables. The latter can handle tables added to the list on the fly. When open_ltable() was used in a loop over a list of tables, the list must be temporarily terminated after every table for open_and_lock_tables(). table_list->required_type is set to FRMTYPE_TABLE to avoid open of special tables. Handling of derived tables is suppressed. These details are handled by the new function open_n_lock_single_table(), which has nearly the same signature as open_ltable() and can replace it in most cases. In reopen_tables() some of the tables open by a thread can be closed and reopened. When a MERGE child is affected, the parent must be closed and reopened too. Closing of the parent is forced before the first child is closed. Reopen happens in the order of thd->open_tables. MERGE parents do not attach their children automatically at open. This is done after all tables are reopened. So all children are open when attaching them. Special lock handling like mysql_lock_abort() or mysql_lock_remove() needs to be suppressed for MERGE children or forwarded to the parent. This depends on the situation. In loops over all open tables one suppresses child lock handling. When a single table is touched, forwarding is done. Behavioral changes: =================== This patch changes the behavior of temporary MERGE tables. Temporary MERGE must have temporary children. The old behavior was wrong. A temporary table is not locked. Hence even non-temporary children were not locked. See Bug 19627 - temporary merge table locking. You cannot change the union list of a non-temporary MERGE table when LOCK TABLES is in effect. The following does *not* work: CREATE TABLE m1 ... ENGINE=MRG_MYISAM ...; LOCK TABLES t1 WRITE, t2 WRITE, m1 WRITE; ALTER TABLE m1 ... UNION=(t1,t2) ...; However, you can do this with a temporary MERGE table. You cannot create a MERGE table with CREATE ... SELECT, neither as a temporary MERGE table, nor as a non-temporary MERGE table. CREATE TABLE m1 ... ENGINE=MRG_MYISAM ... SELECT ...; Gives error message: table is not BASE TABLE.
2007-11-15 20:25:43 +01:00
void init_tmp_table_share(THD *thd, TABLE_SHARE *share, const char *key,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
uint key_length, const char *table_name,
const char *path)
{
DBUG_ENTER("init_tmp_table_share");
Bug#18775 - Temporary table from alter table visible to other threads Continued implementation of WL#1324 (table name to filename encoding) The intermediate (not temporary) files of the new table during ALTER TABLE was visible for SHOW TABLES. These intermediate files are copies of the original table with the changes done by ALTER TABLE. After all the data is copied over from the original table, these files are renamed to the original tables file names. So they are not temporary files. They persist after ALTER TABLE, but just with another name. In 5.0 the intermediate files are invisible for SHOW TABLES because all file names beginning with "#sql" were suppressed. This failed since 5.1.6 because even temporary table names were converted when making file names from them. The prefix became converted to "@0023sql". Converting the prefix during SHOW TABLES would suppress the listing of user tables that start with "#sql". The solution of the problem is to continue the implementation of the table name to file name conversion feature. One requirement is to suppress the conversion for temporary table names. This change is straightforward for real temporary tables as there is a function that creates temporary file names. But the generated path names are located in TMPDIR and have no relation to the internal table name. This cannot be used for ALTER TABLE. Its intermediate files need to be in the same directory as the old table files. And it is necessary to be able to deduce the same path from the same table name repeatedly. Consequently the intermediate table files must be handled like normal tables. Their internal names shall start with tmp_file_prefix (#sql) and they shall not be converted like normal table names. I added a flags parameter to all relevant functions that are called from ALTER TABLE. It is used to suppress the conversion for the intermediate table files. The outcome is that the suppression of #sql in SHOW TABLES works again. It does not suppress user tables as these are converted to @0023sql on file level. This patch does also fix ALTER TABLE ... RENAME, which could not rename a table with non-ASCII characters in its name. It does also fix the problem that a user could create a table like `#sql-xxxx-yyyy`, where xxxx is mysqld's pid and yyyy is the thread ID of some other thread, which prevented this thread from running ALTER TABLE. Some of the above problems are mentioned in Bug 1405, which can be closed with this patch. This patch does also contain some minor fixes for other forgotten conversions. Still known problems are reported as bugs 21370, 21373, and 21387.
2006-08-02 17:57:06 +02:00
DBUG_PRINT("enter", ("table: '%s'.'%s'", key, table_name));
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
bzero((char*) share, sizeof(*share));
init_sql_alloc(&share->mem_root, TABLE_ALLOC_BLOCK_SIZE, 0);
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
share->table_category= TABLE_CATEGORY_TEMPORARY;
share->tmp_table= INTERNAL_TMP_TABLE;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->db.str= (char*) key;
share->db.length= strlen(key);
share->table_cache_key.str= (char*) key;
share->table_cache_key.length= key_length;
share->table_name.str= (char*) table_name;
share->table_name.length= strlen(table_name);
share->path.str= (char*) path;
share->normalized_path.str= (char*) path;
share->path.length= share->normalized_path.length= strlen(path);
share->frm_version= FRM_VER_TRUE_VARCHAR;
Bug#18280 (RBR: Extreneous maps when writing to general_log and slow_log): Filter out replication general_log and slow_log entirely from binary log. Caching result of table share-specific tests.
2006-03-17 18:11:07 +01:00
share->cached_row_logging_check= -1;
Bug#17181 (mysqlslap test server crash): Moving assignments to table_map_id for thread-safe handling of table shares.
2006-02-23 13:34:03 +01:00
Bug#26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Bug 26377 - Deadlock with MERGE and FLUSH TABLE Bug 25038 - Waiting TRUNCATE Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Bug 19627 - temporary merge table locking Bug 27660 - Falcon: merge table possible Bug 30273 - merge tables: Can't lock file (errno: 155) The problems were: Bug 26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table 1. A thread trying to lock a MERGE table performs busy waiting while REPAIR TABLE or a similar table administration task is ongoing on one or more of its MyISAM tables. 2. A thread trying to lock a MERGE table performs busy waiting until all threads that did REPAIR TABLE or similar table administration tasks on one or more of its MyISAM tables in LOCK TABLES segments do UNLOCK TABLES. The difference against problem #1 is that the busy waiting takes place *after* the administration task. It is terminated by UNLOCK TABLES only. 3. Two FLUSH TABLES within a LOCK TABLES segment can invalidate the lock. This does *not* require a MERGE table. The first FLUSH TABLES can be replaced by any statement that requires other threads to reopen the table. In 5.0 and 5.1 a single FLUSH TABLES can provoke the problem. Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Trying DML on a MERGE table, which has a child locked and repaired by another thread, made an infinite loop in the server. Bug 26377 - Deadlock with MERGE and FLUSH TABLE Locking a MERGE table and its children in parent-child order and flushing the child deadlocked the server. Bug 25038 - Waiting TRUNCATE Truncating a MERGE child, while the MERGE table was in use, let the truncate fail instead of waiting for the table to become free. Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Repairing a child of an open MERGE table corrupted the child. It was necessary to FLUSH the child first. Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Flushing and optimizing locked MERGE children crashed the server. Bug 19627 - temporary merge table locking Use of a temporary MERGE table with non-temporary children could corrupt the children. Temporary tables are never locked. So we do now prohibit non-temporary chidlren of a temporary MERGE table. Bug 27660 - Falcon: merge table possible It was possible to create a MERGE table with non-MyISAM children. Bug 30273 - merge tables: Can't lock file (errno: 155) This was a Windows-only bug. Table administration statements sometimes failed with "Can't lock file (errno: 155)". These bugs are fixed by a new implementation of MERGE table open. When opening a MERGE table in open_tables() we do now add the child tables to the list of tables to be opened by open_tables() (the "query_list"). The children are not opened in the handler at this stage. After opening the parent, open_tables() opens each child from the now extended query_list. When the last child is opened, we remove the children from the query_list again and attach the children to the parent. This behaves similar to the old open. However it does not open the MyISAM tables directly, but grabs them from the already open children. When closing a MERGE table in close_thread_table() we detach the children only. Closing of the children is done implicitly because they are in thd->open_tables. For more detail see the comment at the top of ha_myisammrg.cc. Changed from open_ltable() to open_and_lock_tables() in all places that can be relevant for MERGE tables. The latter can handle tables added to the list on the fly. When open_ltable() was used in a loop over a list of tables, the list must be temporarily terminated after every table for open_and_lock_tables(). table_list->required_type is set to FRMTYPE_TABLE to avoid open of special tables. Handling of derived tables is suppressed. These details are handled by the new function open_n_lock_single_table(), which has nearly the same signature as open_ltable() and can replace it in most cases. In reopen_tables() some of the tables open by a thread can be closed and reopened. When a MERGE child is affected, the parent must be closed and reopened too. Closing of the parent is forced before the first child is closed. Reopen happens in the order of thd->open_tables. MERGE parents do not attach their children automatically at open. This is done after all tables are reopened. So all children are open when attaching them. Special lock handling like mysql_lock_abort() or mysql_lock_remove() needs to be suppressed for MERGE children or forwarded to the parent. This depends on the situation. In loops over all open tables one suppresses child lock handling. When a single table is touched, forwarding is done. Behavioral changes: =================== This patch changes the behavior of temporary MERGE tables. Temporary MERGE must have temporary children. The old behavior was wrong. A temporary table is not locked. Hence even non-temporary children were not locked. See Bug 19627 - temporary merge table locking. You cannot change the union list of a non-temporary MERGE table when LOCK TABLES is in effect. The following does *not* work: CREATE TABLE m1 ... ENGINE=MRG_MYISAM ...; LOCK TABLES t1 WRITE, t2 WRITE, m1 WRITE; ALTER TABLE m1 ... UNION=(t1,t2) ...; However, you can do this with a temporary MERGE table. You cannot create a MERGE table with CREATE ... SELECT, neither as a temporary MERGE table, nor as a non-temporary MERGE table. CREATE TABLE m1 ... ENGINE=MRG_MYISAM ... SELECT ...; Gives error message: table is not BASE TABLE.
2007-11-15 20:25:43 +01:00
/*
table_map_id is also used for MERGE tables to suppress repeated
compatibility checks.
*/
share->table_map_id= (ulong) thd->query_id;
Initial import of WL#3726 "DDL locking for all metadata objects". Backport of: ------------------------------------------------------------ revno: 2630.4.1 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Fri 2008-05-23 17:54:03 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ------------------------------------------------------------ This is the first patch in series. It transforms the metadata locking subsystem to use a dedicated module (mdl.h,cc). No significant changes in the locking protocol. The import passes the test suite with the exception of deprecated/removed 6.0 features, and MERGE tables. The latter are subject to a fix by WL#4144. Unfortunately, the original changeset comments got lost in a merge, thus this import has its own (largely insufficient) comments. This patch fixes Bug#25144 "replication / binlog with view breaks". Warning: this patch introduces an incompatible change: Under LOCK TABLES, it's no longer possible to FLUSH a table that was not locked for WRITE. Under LOCK TABLES, it's no longer possible to DROP a table or VIEW that was not locked for WRITE. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.2 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:03:45 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.3 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:08:51 +0400 message: WL#3726 "DDL locking for all metadata objects" Fixed failing Windows builds by adding mdl.cc to the lists of files needed to build server/libmysqld on Windows. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.4 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 21:57:58 +0400 message: WL#3726 "DDL locking for all metadata objects". Fix for assert failures in kill.test which occured when one tried to kill ALTER TABLE statement on merge table while it was waiting in wait_while_table_is_used() for other connections to close this table. These assert failures stemmed from the fact that cleanup code in this case assumed that temporary table representing new version of table was open with adding to THD::temporary_tables list while code which were opening this temporary table wasn't always fulfilling this. This patch changes code that opens new version of table to always do this linking in. It also streamlines cleanup process for cases when error occurs while we have new version of table open. ****** WL#3726 "DDL locking for all metadata objects" Add libmysqld/mdl.cc to .bzrignore. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.6 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sun 2008-05-25 00:33:22 +0400 message: WL#3726 "DDL locking for all metadata objects". Addition to the fix of assert failures in kill.test caused by changes for this worklog. Make sure we close the new table only once.
2009-11-30 16:55:03 +01:00
share->used_tables.empty();
share->free_tables.empty();
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
DBUG_VOID_RETURN;
}
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
/*
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Free table share and memory used by it
SYNOPSIS
free_table_share()
share Table share
NOTES
share->mutex must be locked when we come here if it's not a temp table
*/
void free_table_share(TABLE_SHARE *share)
{
MEM_ROOT mem_root;
BUG#39746 - Debug flag breaks struct definition (server crash) Altering a table with fulltext index[es] which use pluggable fulltext parser may cause server crash in debug builds. The problem was that ALTER TABLE code wrongly assigned fulltext parser name. Also fixed that altering a table with fulltext index[es] leave stale fulltext parser locks, which prevent fulltext parsers from being uninstalled after ALTER TABLE.
2008-12-17 14:24:34 +01:00
uint idx;
KEY *key_info;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
DBUG_ENTER("free_table_share");
DBUG_PRINT("enter", ("table: %s.%s", share->db.str, share->table_name.str));
DBUG_ASSERT(share->ref_count == 0);
Backport of: ------------------------------------------------------------------- revno: 2630.6.6 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-3726 timestamp: Tue 2008-05-27 16:15:44 +0400 message: Implement code review fixes for WL#3726 "DDL locking for all metadata objects": cleanup the code from share->mutex acquisitions, which are now obsolete.
2009-11-30 20:38:25 +01:00
/* The mutex is initialized only for shares that are part of the TDC */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (share->tmp_table == NO_TMP_TABLE)
Merge next-mr -> next-4284
2010-02-03 14:43:03 +01:00
mysql_mutex_destroy(&share->LOCK_ha_data);
Backport of: ---------------------------------------------------------- revno: 2617.22.5 committer: Konstantin Osipov <kostja@sun.com> branch nick: mysql-6.0-runtime timestamp: Tue 2009-01-27 05:08:48 +0300 message: Remove non-prefixed use of HASH. Always use my_hash_init(), my_hash_inited(), my_hash_search(), my_hash_element(), my_hash_delete(), my_hash_free() rather than non-prefixed counterparts (hash_init(), etc). Remove the backward-compatible defines.
2009-10-14 18:37:38 +02:00
my_hash_free(&share->name_hash);
Backport of: ------------------------------------------------------------------- revno: 2630.6.6 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-3726 timestamp: Tue 2008-05-27 16:15:44 +0400 message: Implement code review fixes for WL#3726 "DDL locking for all metadata objects": cleanup the code from share->mutex acquisitions, which are now obsolete.
2009-11-30 20:38:25 +01:00
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
plugin_unlock(NULL, share->db_plugin);
share->db_plugin= NULL;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
BUG#39746 - Debug flag breaks struct definition (server crash) Altering a table with fulltext index[es] which use pluggable fulltext parser may cause server crash in debug builds. The problem was that ALTER TABLE code wrongly assigned fulltext parser name. Also fixed that altering a table with fulltext index[es] leave stale fulltext parser locks, which prevent fulltext parsers from being uninstalled after ALTER TABLE.
2008-12-17 14:24:34 +01:00
/* Release fulltext parsers */
key_info= share->key_info;
for (idx= share->keys; idx; idx--, key_info++)
{
if (key_info->flags & HA_USES_PARSER)
{
plugin_unlock(NULL, key_info->parser);
key_info->flags= 0;
}
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/* We must copy mem_root from share because share is allocated through it */
memcpy((char*) &mem_root, (char*) &share->mem_root, sizeof(mem_root));
free_root(&mem_root, MYF(0)); // Free's share
DBUG_VOID_RETURN;
}
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
BUG#9953: CONVERT_TZ requires mysql.time_zone_name to be locked The problem was that some facilities (like CONVERT_TZ() function or server HELP statement) may require implicit access to some tables in 'mysql' database. This access was done by ordinary means of adding such tables to the list of tables the query is going to open. However, if we issued LOCK TABLES before that, we would get "table was not locked" error trying to open such implicit tables. The solution is to treat certain tables as MySQL system tables, like we already do for mysql.proc. Such tables may be opened for reading at any moment regardless of any locks in effect. The cost of this is that system table may be locked for writing only together with other system tables, it is disallowed to lock system tables for writing and have any other lock on any other table. After this patch the following tables are treated as MySQL system tables: mysql.help_category mysql.help_keyword mysql.help_relation mysql.help_topic mysql.proc (it already was) mysql.time_zone mysql.time_zone_leap_second mysql.time_zone_name mysql.time_zone_transition mysql.time_zone_transition_type These tables are now opened with open_system_tables_for_read() and closed with close_system_tables(), or one table may be opened with open_system_table_for_update() and closed with close_thread_tables() (the latter is used for mysql.proc table, which is updated as part of normal MySQL server operation). These functions may be used when some tables were opened and locked already. NOTE: online update of time zone tables is not possible during replication, because there's no time zone cache flush neither on LOCK TABLES, nor on FLUSH TABLES, so the master may serve stale time zone data from cache, while on slave updated data will be loaded from the time zone tables.
2007-03-09 11:12:31 +01:00
/**
Return TRUE if a table name matches one of the system table names.
Currently these are:
help_category, help_keyword, help_relation, help_topic,
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
proc, event
BUG#9953: CONVERT_TZ requires mysql.time_zone_name to be locked The problem was that some facilities (like CONVERT_TZ() function or server HELP statement) may require implicit access to some tables in 'mysql' database. This access was done by ordinary means of adding such tables to the list of tables the query is going to open. However, if we issued LOCK TABLES before that, we would get "table was not locked" error trying to open such implicit tables. The solution is to treat certain tables as MySQL system tables, like we already do for mysql.proc. Such tables may be opened for reading at any moment regardless of any locks in effect. The cost of this is that system table may be locked for writing only together with other system tables, it is disallowed to lock system tables for writing and have any other lock on any other table. After this patch the following tables are treated as MySQL system tables: mysql.help_category mysql.help_keyword mysql.help_relation mysql.help_topic mysql.proc (it already was) mysql.time_zone mysql.time_zone_leap_second mysql.time_zone_name mysql.time_zone_transition mysql.time_zone_transition_type These tables are now opened with open_system_tables_for_read() and closed with close_system_tables(), or one table may be opened with open_system_table_for_update() and closed with close_thread_tables() (the latter is used for mysql.proc table, which is updated as part of normal MySQL server operation). These functions may be used when some tables were opened and locked already. NOTE: online update of time zone tables is not possible during replication, because there's no time zone cache flush neither on LOCK TABLES, nor on FLUSH TABLES, so the master may serve stale time zone data from cache, while on slave updated data will be loaded from the time zone tables.
2007-03-09 11:12:31 +01:00
time_zone, time_zone_leap_second, time_zone_name, time_zone_transition,
time_zone_transition_type
This function trades accuracy for speed, so may return false
positives. Presumably mysql.* database is for internal purposes only
and should not contain user tables.
*/
inline bool is_system_table_name(const char *name, uint length)
{
CHARSET_INFO *ci= system_charset_info;
return (
Merge from 5.0-bugteam
2009-06-17 16:56:44 +02:00
/* mysql.proc table */
(length == 4 &&
my_tolower(ci, name[0]) == 'p' &&
my_tolower(ci, name[1]) == 'r' &&
my_tolower(ci, name[2]) == 'o' &&
my_tolower(ci, name[3]) == 'c') ||
(length > 4 &&
(
/* one of mysql.help* tables */
(my_tolower(ci, name[0]) == 'h' &&
my_tolower(ci, name[1]) == 'e' &&
my_tolower(ci, name[2]) == 'l' &&
my_tolower(ci, name[3]) == 'p') ||
/* one of mysql.time_zone* tables */
(my_tolower(ci, name[0]) == 't' &&
my_tolower(ci, name[1]) == 'i' &&
my_tolower(ci, name[2]) == 'm' &&
my_tolower(ci, name[3]) == 'e') ||
/* mysql.event table */
(my_tolower(ci, name[0]) == 'e' &&
my_tolower(ci, name[1]) == 'v' &&
my_tolower(ci, name[2]) == 'e' &&
my_tolower(ci, name[3]) == 'n' &&
my_tolower(ci, name[4]) == 't')
)
)
BUG#9953: CONVERT_TZ requires mysql.time_zone_name to be locked The problem was that some facilities (like CONVERT_TZ() function or server HELP statement) may require implicit access to some tables in 'mysql' database. This access was done by ordinary means of adding such tables to the list of tables the query is going to open. However, if we issued LOCK TABLES before that, we would get "table was not locked" error trying to open such implicit tables. The solution is to treat certain tables as MySQL system tables, like we already do for mysql.proc. Such tables may be opened for reading at any moment regardless of any locks in effect. The cost of this is that system table may be locked for writing only together with other system tables, it is disallowed to lock system tables for writing and have any other lock on any other table. After this patch the following tables are treated as MySQL system tables: mysql.help_category mysql.help_keyword mysql.help_relation mysql.help_topic mysql.proc (it already was) mysql.time_zone mysql.time_zone_leap_second mysql.time_zone_name mysql.time_zone_transition mysql.time_zone_transition_type These tables are now opened with open_system_tables_for_read() and closed with close_system_tables(), or one table may be opened with open_system_table_for_update() and closed with close_thread_tables() (the latter is used for mysql.proc table, which is updated as part of normal MySQL server operation). These functions may be used when some tables were opened and locked already. NOTE: online update of time zone tables is not possible during replication, because there's no time zone cache flush neither on LOCK TABLES, nor on FLUSH TABLES, so the master may serve stale time zone data from cache, while on slave updated data will be loaded from the time zone tables.
2007-03-09 11:12:31 +01:00
);
}
Bug #53371: COM_FIELD_LIST can be abused to bypass table level grants. This is the 5.1 merge and extension of the fix. The server was happily accepting paths in table name in all places a table name is accepted (e.g. a SELECT). This allowed all users that have some privilege over some database to read all tables in all databases in all mysql server instances that the server file system has access to. Fixed by : 1. making sure no path elements are allowed in quoted table name when constructing the path (note that the path symbols are still valid in table names when they're properly escaped by the server). 2. checking the #mysql50# prefixed names the same way they're checked for path elements in mysql-5.0.
2010-05-04 16:03:28 +02:00
/**
Check if a string contains path elements
*/
static inline bool has_disabled_path_chars(const char *str)
{
for (; *str; str++)
switch (*str)
{
case FN_EXTCHAR:
case '/':
case '\\':
case '~':
case '@':
return TRUE;
}
return FALSE;
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/*
Read table definition from a binary / text based .frm file
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
SYNOPSIS
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
open_table_def()
thd Thread handler
share Fill this with table definition
db_flags Bit mask of the following flags: OPEN_VIEW
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
NOTES
This function is called when the table definition is not cached in
table_def_cache
The data is returned in 'share', which is alloced by
alloc_table_share().. The code assumes that share is initialized.
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
RETURN VALUES
0 ok
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
1 Error (see open_table_error)
2 Error (see open_table_error)
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
3 Wrong data in .frm file
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
4 Error (see open_table_error)
5 Error (see open_table_error: charset unavailable)
Add special error message for .frm files with an un-understood version, like those from 5.0 with new datatypes. (Bug #10435)
2005-05-07 22:57:22 +02:00
6 Unknown .frm version
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
*/
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
int open_table_def(THD *thd, TABLE_SHARE *share, uint db_flags)
{
int error, table_type;
bool error_given;
File file;
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
uchar head[64], *disk_buff;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
char path[FN_REFLEN];
Simpler arena swapping code Now thd->mem_root is a pointer to thd->main_mem_root and THR_MALLOC is a pointer to thd->mem_root. This gives us the following benefits: - Allow us to easily detect if arena has already been swapped before (this fixes a bug in setup_conds() where arena was swaped twice in some cases) - Faster swaps of arenas (as we don't have to copy the whole MEM_ROOT) - We don't anymore have to call my_pthread_setspecific_ptr(THR_MALLOC,...) to change where memory is alloced. Now it's enough to set thd->mem_root
2004-11-08 00:13:54 +01:00
MEM_ROOT **root_ptr, *old_root;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
DBUG_ENTER("open_table_def");
Bug#18775 - Temporary table from alter table visible to other threads Continued implementation of WL#1324 (table name to filename encoding) The intermediate (not temporary) files of the new table during ALTER TABLE was visible for SHOW TABLES. These intermediate files are copies of the original table with the changes done by ALTER TABLE. After all the data is copied over from the original table, these files are renamed to the original tables file names. So they are not temporary files. They persist after ALTER TABLE, but just with another name. In 5.0 the intermediate files are invisible for SHOW TABLES because all file names beginning with "#sql" were suppressed. This failed since 5.1.6 because even temporary table names were converted when making file names from them. The prefix became converted to "@0023sql". Converting the prefix during SHOW TABLES would suppress the listing of user tables that start with "#sql". The solution of the problem is to continue the implementation of the table name to file name conversion feature. One requirement is to suppress the conversion for temporary table names. This change is straightforward for real temporary tables as there is a function that creates temporary file names. But the generated path names are located in TMPDIR and have no relation to the internal table name. This cannot be used for ALTER TABLE. Its intermediate files need to be in the same directory as the old table files. And it is necessary to be able to deduce the same path from the same table name repeatedly. Consequently the intermediate table files must be handled like normal tables. Their internal names shall start with tmp_file_prefix (#sql) and they shall not be converted like normal table names. I added a flags parameter to all relevant functions that are called from ALTER TABLE. It is used to suppress the conversion for the intermediate table files. The outcome is that the suppression of #sql in SHOW TABLES works again. It does not suppress user tables as these are converted to @0023sql on file level. This patch does also fix ALTER TABLE ... RENAME, which could not rename a table with non-ASCII characters in its name. It does also fix the problem that a user could create a table like `#sql-xxxx-yyyy`, where xxxx is mysqld's pid and yyyy is the thread ID of some other thread, which prevented this thread from running ALTER TABLE. Some of the above problems are mentioned in Bug 1405, which can be closed with this patch. This patch does also contain some minor fixes for other forgotten conversions. Still known problems are reported as bugs 21370, 21373, and 21387.
2006-08-02 17:57:06 +02:00
DBUG_PRINT("enter", ("table: '%s'.'%s' path: '%s'", share->db.str,
share->table_name.str, share->normalized_path.str));
Import changeset
2000-07-31 21:29:14 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
error= 1;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
error_given= 0;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
disk_buff= NULL;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
strxmov(path, share->normalized_path.str, reg_ext, NullS);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if ((file= mysql_file_open(key_file_frm,
path, O_RDONLY | O_SHARE, MYF(0))) < 0)
table.cc: Try to open an old 5.0 table name if encoded name doesn't exist. This makes SELECT working without having to rename tables with tricky characters in name into 5.1 format (simpliefies upgrade).
2006-01-24 14:33:39 +01:00
{
Bug#26402 Server crashes with old-style named table Problem: crash on attempt to open a table having "#mysql50#" prefix in db or table name. Fix: This prefix is reserved for "mysql_upgrade" to access 5.0 tables whose file names are not encoded according to "5.1 tablename to filename encoded". Don't try open tables whose db name or table name has this prefix.
2007-06-14 13:28:33 +02:00
/*
We don't try to open 5.0 unencoded name, if
- non-encoded name contains '@' signs,
because '@' can be misinterpreted.
It is not clear if '@' is escape character in 5.1,
or a normal character in 5.0.
- non-encoded db or table name contain "#mysql50#" prefix.
This kind of tables must have been opened only by the
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_open() above.
Bug#26402 Server crashes with old-style named table Problem: crash on attempt to open a table having "#mysql50#" prefix in db or table name. Fix: This prefix is reserved for "mysql_upgrade" to access 5.0 tables whose file names are not encoded according to "5.1 tablename to filename encoded". Don't try open tables whose db name or table name has this prefix.
2007-06-14 13:28:33 +02:00
*/
Bug #53371: COM_FIELD_LIST can be abused to bypass table level grants. This is the 5.1 merge and extension of the fix. The server was happily accepting paths in table name in all places a table name is accepted (e.g. a SELECT). This allowed all users that have some privilege over some database to read all tables in all databases in all mysql server instances that the server file system has access to. Fixed by : 1. making sure no path elements are allowed in quoted table name when constructing the path (note that the path symbols are still valid in table names when they're properly escaped by the server). 2. checking the #mysql50# prefixed names the same way they're checked for path elements in mysql-5.0.
2010-05-04 16:03:28 +02:00
if (has_disabled_path_chars(share->table_name.str) ||
has_disabled_path_chars(share->db.str) ||
Bug#26402 Server crashes with old-style named table Problem: crash on attempt to open a table having "#mysql50#" prefix in db or table name. Fix: This prefix is reserved for "mysql_upgrade" to access 5.0 tables whose file names are not encoded according to "5.1 tablename to filename encoded". Don't try open tables whose db name or table name has this prefix.
2007-06-14 13:28:33 +02:00
!strncmp(share->db.str, MYSQL50_TABLE_NAME_PREFIX,
MYSQL50_TABLE_NAME_PREFIX_LENGTH) ||
!strncmp(share->table_name.str, MYSQL50_TABLE_NAME_PREFIX,
MYSQL50_TABLE_NAME_PREFIX_LENGTH))
Bug#17142 Crash if create with encoded name upgrade.result, upgrade.test: Adding test case. table.cc: Don't try to open a table with old-formatted name, if the table name contains "@" character. This is to avoid mixing two different names to each other. Thanks Monty for suggesting this fix.
2006-03-07 10:05:24 +01:00
goto err_not_open;
Bug#26402 Server crashes with old-style named table Problem: crash on attempt to open a table having "#mysql50#" prefix in db or table name. Fix: This prefix is reserved for "mysql_upgrade" to access 5.0 tables whose file names are not encoded according to "5.1 tablename to filename encoded". Don't try open tables whose db name or table name has this prefix.
2007-06-14 13:28:33 +02:00
/* Try unencoded 5.0 name */
table.cc: Try to open an old 5.0 table name if encoded name doesn't exist. This makes SELECT working without having to rename tables with tricky characters in name into 5.1 format (simpliefies upgrade).
2006-01-24 14:33:39 +01:00
uint length;
strxnmov(path, sizeof(path)-1,
mysql_data_home, "/", share->db.str, "/",
share->table_name.str, reg_ext, NullS);
length= unpack_filename(path, path) - reg_ext_length;
/*
The following is a safety test and should never fail
as the old file name should never be longer than the new one.
*/
DBUG_ASSERT(length <= share->normalized_path.length);
/*
If the old and the new names have the same length,
then table name does not have tricky characters,
so no need to check the old file name.
*/
if (length == share->normalized_path.length ||
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
((file= mysql_file_open(key_file_frm,
path, O_RDONLY | O_SHARE, MYF(0))) < 0))
table.cc: Try to open an old 5.0 table name if encoded name doesn't exist. This makes SELECT working without having to rename tables with tricky characters in name into 5.1 format (simpliefies upgrade).
2006-01-24 14:33:39 +01:00
goto err_not_open;
/* Unencoded 5.0 table name found */
path[length]= '\0'; // Remove .frm extension
strmov(share->normalized_path.str, path);
share->normalized_path.length= length;
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
error= 4;
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if (mysql_file_read(file, head, 64, MYF(MY_NABP)))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (head[0] == (uchar) 254 && head[1] == 1)
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (head[2] == FRM_VER || head[2] == FRM_VER+1 ||
(head[2] >= FRM_VER+3 && head[2] <= FRM_VER+4))
WL#3732 Information schema optimization
2007-08-03 00:14:05 +02:00
{
/* Open view only */
if (db_flags & OPEN_VIEW_ONLY)
{
error_given= 1;
goto err;
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
table_type= 1;
WL#3732 Information schema optimization
2007-08-03 00:14:05 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
else
{
error= 6; // Unkown .frm version
goto err;
}
}
Merge mysql.com:/home/my/mysql-5.1 into mysql.com:/home/my/mysql-5.1-TDC
2005-11-23 21:58:53 +01:00
else if (memcmp(head, STRING_WITH_LEN("TYPE=")) == 0)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
error= 5;
if (memcmp(head+5,"VIEW",4) == 0)
{
share->is_view= 1;
if (db_flags & OPEN_VIEW)
error= 0;
}
goto err;
}
else
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/* No handling of text based files yet */
if (table_type == 1)
{
root_ptr= my_pthread_getspecific_ptr(MEM_ROOT**, THR_MALLOC);
old_root= *root_ptr;
*root_ptr= &share->mem_root;
error= open_binary_frm(thd, share, head, file);
*root_ptr= old_root;
error_given= 1;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log) Bug#25422 (Hang with log tables) Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the thread) Bug 23044 (Warnings on flush of a log table) Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes a deadlock) Prior to this fix, the server would hang when performing concurrent ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES, which are mysql.general_log and mysql.slow_log. The root cause traces to the following code: in sql_base.cc, open_table() if (table->in_use != thd) { /* wait_for_condition will unlock LOCK_open for us */ wait_for_condition(thd, &LOCK_open, &COND_refresh); } The problem with this code is that the current implementation of the LOGGER creates 'fake' THD objects, like - Log_to_csv_event_handler::general_log_thd - Log_to_csv_event_handler::slow_log_thd which are not associated to a real thread running in the server, so that waiting for these non-existing threads to release table locks cause the dead lock. In general, the design of Log_to_csv_event_handler does not fit into the general architecture of the server, so that the concept of general_log_thd and slow_log_thd has to be abandoned: - this implementation does not work with table locking - it will not work with commands like SHOW PROCESSLIST - having the log tables always opened does not integrate well with DDL operations / FLUSH TABLES / SET GLOBAL READ_ONLY With this patch, the fundamental design of the LOGGER has been changed to: - always open and close a log table when writing a log - remove totally the usage of fake THD objects - clarify how locking of log tables is implemented in general. See WL#3984 for details related to the new locking design. Additional changes (misc bugs exposed and fixed): 1) mysqldump which would ignore some tables in dump_all_tables_in_db(), but forget to ignore the same in dump_all_views_in_db(). 2) mysqldump would also issue an empty "LOCK TABLE" command when all the tables to lock are to be ignored (numrows == 0), instead of not issuing the query. 3) Internal errors handlers could intercept errors but not warnings (see sql_error.cc). 4) Implementing a nested call to open tables, for the performance schema tables, exposed an existing bug in remove_table_from_cache(), which would perform: in_use->some_tables_deleted=1; against another thread, without any consideration about thread locking. This call inside remove_table_from_cache() was not required anyway, since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads that might hold a lock on a table. This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
share->table_category= get_table_category(& share->db, & share->table_name);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (!error)
thd->status_var.opened_shares++;
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
err:
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_close(file, MYF(MY_WME));
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
err_not_open:
if (error && !error_given)
Add special error message for .frm files with an un-understood version, like those from 5.0 with new datatypes. (Bug #10435)
2005-05-07 22:57:22 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->error= error;
open_table_error(share, error, (share->open_errno= my_errno), 0);
Add special error message for .frm files with an un-understood version, like those from 5.0 with new datatypes. (Bug #10435)
2005-05-07 22:57:22 +02:00
}
Bug#17181 (mysqlslap test server crash): Moving assignments to table_map_id for thread-safe handling of table shares.
2006-02-23 13:34:03 +01:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
DBUG_RETURN(error);
}
/*
Read data from a binary .frm file from MySQL 3.23 - 5.0 into TABLE_SHARE
*/
static int open_binary_frm(THD *thd, TABLE_SHARE *share, uchar *head,
File file)
{
int error, errarg= 0;
uint new_frm_ver, field_pack_length, new_field_pack_flag;
uint interval_count, interval_parts, read_length, int_length;
uint db_create_options, keys, key_parts, n_length;
uint key_info_length, com_length, null_bit_pos;
uint extra_rec_buf_length;
uint i,j;
bool use_hash;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
char *keynames, *names, *comment_pos;
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
uchar forminfo[288];
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uchar *record;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
uchar *disk_buff, *strpos, *null_flags, *null_pos;
ulong pos, record_offset, *rec_per_key, rec_buff_length;
handler *handler_file= 0;
KEY *keyinfo;
KEY_PART_INFO *key_part;
SQL_CRYPT *crypted=0;
Field **field_ptr, *reg_field;
const char **interval_array;
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
enum legacy_db_type legacy_db_type;
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
my_bitmap_map *bitmaps;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
DBUG_ENTER("open_binary_frm");
new_field_pack_flag= head[27];
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
new_frm_ver= (head[2] - FRM_VER);
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
field_pack_length= new_frm_ver < 2 ? 11 : 17;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
disk_buff= 0;
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
error= 3;
Import changeset
2000-07-31 21:29:14 +02:00
if (!(pos=get_form_pos(file,head,(TYPELIB*) 0)))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
mysql_file_seek(file,pos,MY_SEEK_SET,MYF(0));
if (mysql_file_read(file, forminfo,288,MYF(MY_NABP)))
goto err;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->frm_version= head[2];
Hack to ensure that CHAR's created in 5.0 are not converted to VARCHAR even if the row type is dynamic (For 4.1 tables old 'VARCHAR' fields are converted to true VARCHAR in the next ALTER TABLE) This ensures that one can use MySQL 5.0 privilege tables with MySQL 4.1
2005-03-22 14:48:06 +01:00
/*
Check if .frm file created by MySQL 5.0. In this case we want to
display CHAR fields as CHAR and not as VARCHAR.
We do it this way as we want to keep the old frm version to enable
MySQL 4.1 to read these files.
*/
if (share->frm_version == FRM_VER_TRUE_VARCHAR -1 && head[33] == 5)
share->frm_version= FRM_VER_TRUE_VARCHAR;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
#ifdef WITH_PARTITION_STORAGE_ENGINE
Bug#20871 Partitions: crash if --skip-innodb return error if engine is disabled
2006-08-16 17:07:12 +02:00
if (*(head+61) &&
!(share->default_part_db_type=
ha_checktype(thd, (enum legacy_db_type) (uint) *(head+61), 1, 0)))
goto err;
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
DBUG_PRINT("info", ("default_part_db_type = %u", head[61]));
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
#endif
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
legacy_db_type= (enum legacy_db_type) (uint) *(head+3);
WL#2936 "Server variables for plugins" Post review fixes.
2007-03-23 18:14:46 +01:00
DBUG_ASSERT(share->db_plugin == NULL);
/*
if the storage engine is dynamic, no point in resolving it by its
dynamically allocated legacy_db_type. We will resolve it later by name.
*/
if (legacy_db_type > DB_TYPE_UNKNOWN &&
legacy_db_type < DB_TYPE_FIRST_DYNAMIC)
share->db_plugin= ha_lock_engine(NULL,
ha_checktype(thd, legacy_db_type, 0, 0));
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->db_create_options= db_create_options= uint2korr(head+30);
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->db_options_in_use= share->db_create_options;
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
share->mysql_version= uint4korr(head+51);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->null_field_first= 0;
Import changeset
2000-07-31 21:29:14 +02:00
if (!head[32]) // New frm file in 3.23
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->avg_row_length= uint4korr(head+34);
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
share->row_type= (row_type) head[40];
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
share->table_charset= get_charset((((uint) head[41]) << 8) +
(uint) head[38],MYF(0));
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->null_field_first= 1;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
}
if (!share->table_charset)
{
/* unknown charset in head[38] or pre-3.23 frm */
Print a warning when an old table (with no character set stored) is opened and the default character set is multi-byte, which will result in character column size changes. (Bug #6913)
2005-01-17 21:22:23 +01:00
if (use_mb(default_charset_info))
{
/* Warn that we may be changing the size of character columns */
sql_print_warning("'%s' had no or invalid character set, "
"and default character set is multi-byte, "
"so character column sizes may have changed",
Fixed a lot of compiler warnings (Mainly in mysqld and instance manager) Fixed some possible fatal wrong arguments to printf() style functions Initialized some not initialized variables Fixed bug in stored procedure and continue handlers (Fixes Bug#22150)
2006-11-01 18:41:09 +01:00
share->path.str);
Print a warning when an old table (with no character set stored) is opened and the default character set is multi-byte, which will result in character column size changes. (Bug #6913)
2005-01-17 21:22:23 +01:00
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->table_charset= default_charset_info;
Import changeset
2000-07-31 21:29:14 +02:00
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->db_record_offset= 1;
Import changeset
2000-07-31 21:29:14 +02:00
if (db_create_options & HA_OPTION_LONG_BLOB_PTR)
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->blob_ptr_size= portable_sizeof_char_ptr;
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
/* Set temporarily a good value for db_low_byte_first */
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
share->db_low_byte_first= test(legacy_db_type != DB_TYPE_ISAM);
Import changeset
2000-07-31 21:29:14 +02:00
error=4;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->max_rows= uint4korr(head+18);
share->min_rows= uint4korr(head+22);
Import changeset
2000-07-31 21:29:14 +02:00
/* Read keyinformation */
Update for BTREE keys in HEAP tables Split handler->option_flag() to handler->table_flags() and handler->index_flags()
2002-04-12 20:35:46 +02:00
key_info_length= (uint) uint2korr(head+28);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_seek(file, (ulong) uint2korr(head+6), MY_SEEK_SET, MYF(0));
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (read_string(file,(uchar**) &disk_buff,key_info_length))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
Bitmap post-review fixes
2003-10-30 19:17:57 +01:00
if (disk_buff[0] & 0x80)
>255 keys support in frm
2003-10-21 00:13:17 +02:00
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->keys= keys= (disk_buff[1] << 7) | (disk_buff[0] & 0x7f);
share->key_parts= key_parts= uint2korr(disk_buff+2);
>255 keys support in frm
2003-10-21 00:13:17 +02:00
}
else
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->keys= keys= disk_buff[0];
share->key_parts= key_parts= disk_buff[1];
>255 keys support in frm
2003-10-21 00:13:17 +02:00
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->keys_for_keyread.init(0);
share->keys_in_use.init(keys);
Import changeset
2000-07-31 21:29:14 +02:00
n_length=keys*sizeof(KEY)+key_parts*sizeof(KEY_PART_INFO);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (!(keyinfo = (KEY*) alloc_root(&share->mem_root,
n_length + uint2korr(disk_buff+4))))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
bzero((char*) keyinfo,n_length);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->key_info= keyinfo;
A lot of portability fixes. Added rename table for BDB. This fixes a bug in ALTER TABLE with BDB tables.
2003-01-05 19:18:49 +01:00
key_part= my_reinterpret_cast(KEY_PART_INFO*) (keyinfo+keys);
Import changeset
2000-07-31 21:29:14 +02:00
strpos=disk_buff+6;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (!(rec_per_key= (ulong*) alloc_root(&share->mem_root,
Bug#41212 crash after race condition between merge table and table_cache evictions On 64-bit Windows: querying MERGE table with keys may cause server crash.The problem is generic and may affect any statement accessing MERGE table cardinality values. When MERGE engine was copying cardinality statistics, it was using incorrect size of element in cardinality statistics array (sizeof(ptr)==8 instead of sizeof(ulong)==4), causing access of memory beyond of the allocated bounds.
2009-05-27 12:34:21 +02:00
sizeof(ulong)*key_parts)))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
Import changeset
2000-07-31 21:29:14 +02:00
for (i=0 ; i < keys ; i++, keyinfo++)
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
keyinfo->table= 0; // Updated in open_frm
Add support for up to VARCHAR (size up to 65535) Renamed HA_VAR_LENGTH to HA_VAR_LENGTH_PART Renamed in all files FIELD_TYPE_STRING and FIELD_TYPE_VAR_STRING to MYSQL_TYPE_STRING and MYSQL_TYPE_VAR_STRING to make it easy to catch all possible errors Added support for VARCHAR KEYS to heap Removed support for ISAM Now only long VARCHAR columns are changed to TEXT on demand (not CHAR) Internal temporary files can now use fixed length tables if the used VARCHAR columns are short
2004-12-06 01:00:37 +01:00
if (new_frm_ver >= 3)
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
{
keyinfo->flags= (uint) uint2korr(strpos) ^ HA_NOSAME;
keyinfo->key_length= (uint) uint2korr(strpos+2);
keyinfo->key_parts= (uint) strpos[4];
keyinfo->algorithm= (enum ha_key_alg) strpos[5];
Added support for key_block_size for key and table level (WL#602) Added support for key_block_size to MyISAM. Simplify interface to 'new Key' to make it easier to add new key options. mysqld option --new is used to define where key options are printed. (In 5.3 we should move all key options to after key part definition to avoid problem with reserved names) Fixed some compiler warnings and a memory leak in ssl
2006-05-03 14:59:17 +02:00
keyinfo->block_size= uint2korr(strpos+6);
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
strpos+=8;
}
else
{
keyinfo->flags= ((uint) strpos[0]) ^ HA_NOSAME;
keyinfo->key_length= (uint) uint2korr(strpos+1);
keyinfo->key_parts= (uint) strpos[3];
keyinfo->algorithm= HA_KEY_ALG_UNDEF;
strpos+=4;
}
Added GIS extension
2002-02-22 12:24:42 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
keyinfo->key_part= key_part;
keyinfo->rec_per_key= rec_per_key;
for (j=keyinfo->key_parts ; j-- ; key_part++)
{
*rec_per_key++=0;
key_part->fieldnr= (uint16) (uint2korr(strpos) & FIELD_NR_MASK);
key_part->offset= (uint) uint2korr(strpos+2)-1;
key_part->key_type= (uint) uint2korr(strpos+5);
// key_part->field= (Field*) 0; // Will be fixed later
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
if (new_frm_ver >= 1)
Import changeset
2000-07-31 21:29:14 +02:00
{
key_part->key_part_flag= *(strpos+4);
key_part->length= (uint) uint2korr(strpos+7);
strpos+=9;
}
else
{
key_part->length= *(strpos+4);
key_part->key_part_flag=0;
if (key_part->length > 128)
{
key_part->length&=127; /* purecov: inspected */
key_part->key_part_flag=HA_REVERSE_SORT; /* purecov: inspected */
}
strpos+=7;
}
key_part->store_length=key_part->length;
}
}
Update for BTREE keys in HEAP tables Split handler->option_flag() to handler->table_flags() and handler->index_flags()
2002-04-12 20:35:46 +02:00
keynames=(char*) key_part;
strpos+= (strmov(keynames, (char *) strpos) - keynames)+1;
Made keyread (key scanning) a key specific attribute. This avoids using fulltext keys for table scanning. This also reverts Sinisa's original fix for this problem.
2002-06-18 23:22:30 +02:00
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
//reading index comments
for (keyinfo= share->key_info, i=0; i < keys; i++, keyinfo++)
{
if (keyinfo->flags & HA_USES_COMMENT)
{
keyinfo->comment.length= uint2korr(strpos);
keyinfo->comment.str= strmake_root(&share->mem_root, (char*) strpos+2,
keyinfo->comment.length);
strpos+= 2 + keyinfo->comment.length;
}
DBUG_ASSERT(test(keyinfo->flags & HA_USES_COMMENT) ==
(keyinfo->comment.length > 0));
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->reclength = uint2korr((head+16));
Import changeset
2000-07-31 21:29:14 +02:00
if (*(head+26) == 1)
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->system= 1; /* one-record-database */
Import changeset
2000-07-31 21:29:14 +02:00
#ifdef HAVE_CRYPTED_FRM
else if (*(head+26) == 2)
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
crypted= get_crypt_for_frm();
share->crypted= 1;
Import changeset
2000-07-31 21:29:14 +02:00
}
#endif
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
record_offset= (ulong) (uint2korr(head+6)+
((uint2korr(head+14) == 0xffff ?
uint4korr(head+47) : uint2korr(head+14))));
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
if ((n_length= uint4korr(head+55)))
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
{
/* Read extra data segment */
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uchar *buff, *next_chunk, *buff_end;
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
DBUG_PRINT("info", ("extra segment size is %u bytes", n_length));
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(next_chunk= buff= (uchar*) my_malloc(n_length, MYF(MY_WME))))
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
goto err;
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if (mysql_file_pread(file, buff, n_length, record_offset + share->reclength,
MYF(MY_NABP)))
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
{
my_free(buff, MYF(0));
goto err;
}
Always save/restore connect string.
2005-10-10 20:53:53 +02:00
share->connect_string.length= uint2korr(buff);
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(share->connect_string.str= strmake_root(&share->mem_root,
(char*) next_chunk + 2,
share->connect_string.
length)))
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
{
Always save/restore connect string.
2005-10-10 20:53:53 +02:00
my_free(buff, MYF(0));
goto err;
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
}
Always save/restore connect string.
2005-10-10 20:53:53 +02:00
next_chunk+= share->connect_string.length + 2;
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
buff_end= buff + n_length;
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
if (next_chunk + 2 < buff_end)
{
uint str_db_type_length= uint2korr(next_chunk);
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
LEX_STRING name;
name.str= (char*) next_chunk + 2;
name.length= str_db_type_length;
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
plugin_ref tmp_plugin= ha_resolve_by_name(thd, &name);
WL#2936 "Server variables for plugins" Post review fixes.
2007-03-23 18:14:46 +01:00
if (tmp_plugin != NULL && !plugin_equals(tmp_plugin, share->db_plugin))
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
{
WL#2936 "Server variables for plugins" Post review fixes.
2007-03-23 18:14:46 +01:00
if (legacy_db_type > DB_TYPE_UNKNOWN &&
legacy_db_type < DB_TYPE_FIRST_DYNAMIC &&
legacy_db_type != ha_legacy_type(
plugin_data(tmp_plugin, handlerton *)))
{
/* bad file, legacy_db_type did not match the name */
my_free(buff, MYF(0));
goto err;
}
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
/*
tmp_plugin is locked with a local lock.
we unlock the old value of share->db_plugin before
replacing it with a globally locked version of tmp_plugin
*/
plugin_unlock(NULL, share->db_plugin);
share->db_plugin= my_plugin_lock(NULL, &tmp_plugin);
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
DBUG_PRINT("info", ("setting dbtype to '%.*s' (%d)",
str_db_type_length, next_chunk + 2,
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
ha_legacy_type(share->db_type())));
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
}
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
#ifdef WITH_PARTITION_STORAGE_ENGINE
BUG#22708 - Error message doesn't refer to storage engine unsupported When openning a table with unsupported (disabled or not compiled) storage engine, confusing error message is returned. Return better error message when we're attempting to open a table that uses unsupported engine.
2007-12-14 14:05:27 +01:00
else if (str_db_type_length == 9 &&
!strncmp((char *) next_chunk + 2, "partition", 9))
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
{
BUG#22708 - Error message doesn't refer to storage engine unsupported When openning a table with unsupported (disabled or not compiled) storage engine, confusing error message is returned. Return better error message when we're attempting to open a table that uses unsupported engine.
2007-12-14 14:05:27 +01:00
/*
Use partition handler
tmp_plugin is locked with a local lock.
we unlock the old value of share->db_plugin before
replacing it with a globally locked version of tmp_plugin
*/
Bug#39893: Crash if select on a partitioned table, when partitioning is disabled Problem was that it tried to run partitioning function calls when opening a partitioned table, when it was explicitly disabled. Solution is to check if the partitioning plugin is ready to use before using any partitioning specific calls.
2009-01-08 15:16:44 +01:00
/* Check if the partitioning engine is ready */
if (!plugin_is_ready(&name, MYSQL_STORAGE_ENGINE_PLUGIN))
{
error= 8;
my_error(ER_OPTION_PREVENTS_STATEMENT, MYF(0),
"--skip-partition");
my_free(buff, MYF(0));
goto err;
}
BUG#22708 - Error message doesn't refer to storage engine unsupported When openning a table with unsupported (disabled or not compiled) storage engine, confusing error message is returned. Return better error message when we're attempting to open a table that uses unsupported engine.
2007-12-14 14:05:27 +01:00
plugin_unlock(NULL, share->db_plugin);
share->db_plugin= ha_lock_engine(NULL, partition_hton);
DBUG_PRINT("info", ("setting dbtype to '%.*s' (%d)",
str_db_type_length, next_chunk + 2,
ha_legacy_type(share->db_type())));
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
}
#endif
BUG#22708 - Error message doesn't refer to storage engine unsupported When openning a table with unsupported (disabled or not compiled) storage engine, confusing error message is returned. Return better error message when we're attempting to open a table that uses unsupported engine.
2007-12-14 14:05:27 +01:00
else if (!tmp_plugin)
{
/* purecov: begin inspected */
error= 8;
fixed "engine not found" error in open_binary_frm corrected rotten test
2009-12-23 13:06:03 +01:00
name.str[name.length]=0;
BUG#22708 - Error message doesn't refer to storage engine unsupported When openning a table with unsupported (disabled or not compiled) storage engine, confusing error message is returned. Return better error message when we're attempting to open a table that uses unsupported engine.
2007-12-14 14:05:27 +01:00
my_error(ER_UNKNOWN_STORAGE_ENGINE, MYF(0), name.str);
my_free(buff, MYF(0));
goto err;
/* purecov: end */
}
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
next_chunk+= str_db_type_length + 2;
}
BUG#16068: Plugin related bugs Fixed that --plugin_dir option doesn't work. Fixed that parser name is not restored correctly from the table.
2005-12-29 14:28:59 +01:00
if (next_chunk + 5 < buff_end)
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
{
Mikaels fix for compiling without partition storage engine
2005-11-25 10:33:38 +01:00
uint32 partition_info_len = uint4korr(next_chunk);
#ifdef WITH_PARTITION_STORAGE_ENGINE
Bug #28477 innodb assertion and crash during alter table to add/drop partition. The bug was repeated on MyISAM tables, so isn't InnoDB specific. Reason of the bug is that partition-related members of TABLE_SHARE wasn't properly updated after ALTER command. So if other thread doesn't reread frm file, and just uses cached SHARE, it uses wrong data
2007-06-01 16:44:09 +02:00
if ((share->partition_info_buffer_size=
share->partition_info_len= partition_info_len))
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
{
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(share->partition_info= (char*)
Remove unnecessary casts to uchar. The casts are stemming from the lexer API which internally uses unsigned char variables to address its state map. The implementation of the lexer should be internal to the lexer, and not influence the rest of the code.
2007-03-27 19:09:56 +02:00
memdup_root(&share->mem_root, next_chunk + 4,
partition_info_len + 1)))
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
{
my_free(buff, MYF(0));
goto err;
}
Mikaels fix for compiling without partition storage engine
2005-11-25 10:33:38 +01:00
}
#else
if (partition_info_len)
{
DBUG_PRINT("info", ("WITH_PARTITION_STORAGE_ENGINE is not defined"));
my_free(buff, MYF(0));
goto err;
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
}
Mikaels fix for compiling without partition storage engine
2005-11-25 10:33:38 +01:00
#endif
BUG#16068: Plugin related bugs Fixed that --plugin_dir option doesn't work. Fixed that parser name is not restored correctly from the table.
2005-12-29 14:28:59 +01:00
next_chunk+= 5 + partition_info_len;
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
}
WL 2826: Error handling of ALTER TABLE for partitioning After review changes
2006-04-16 03:49:13 +02:00
#if MYSQL_VERSION_ID < 50200
if (share->mysql_version >= 50106 && share->mysql_version <= 50109)
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
{
/*
WL 2826: Error handling of ALTER TABLE for partitioning After review changes
2006-04-16 03:49:13 +02:00
Partition state array was here in version 5.1.6 to 5.1.9, this code
makes it possible to load a 5.1.6 table in later versions. Can most
likely be removed at some point in time. Will only be used for
upgrades within 5.1 series of versions. Upgrade to 5.2 can only be
done from newer 5.1 versions.
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
*/
WL 2826: Fourth step Remove partition state from frm file Ensure we can still load tables created in 5.1.6 for the time being
2006-02-01 17:05:22 +01:00
next_chunk+= 4;
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
}
a fix for Bug #46652: ndb_restore_compat gives Valgrind error in open_binary_frm(), until it's fixed in a parent tree
2009-08-11 18:47:26 +02:00
else
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
#endif
a fix for Bug #46652: ndb_restore_compat gives Valgrind error in open_binary_frm(), until it's fixed in a parent tree
2009-08-11 18:47:26 +02:00
if (share->mysql_version >= 50110)
BUG#19010: Fix issues with that ALTER TABLE from auto-partitioned NDB table doesn't work unless primary key exists on table.
2006-05-10 18:53:40 +02:00
{
/* New auto_partitioned indicator introduced in 5.1.11 */
#ifdef WITH_PARTITION_STORAGE_ENGINE
share->auto_partitioned= *next_chunk;
#endif
next_chunk++;
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
keyinfo= share->key_info;
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
for (i= 0; i < keys; i++, keyinfo++)
{
if (keyinfo->flags & HA_USES_PARSER)
{
LEX_STRING parser_name;
if (next_chunk >= buff_end)
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
DBUG_PRINT("error",
("fulltext key uses parser that is not defined in .frm"));
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
my_free(buff, MYF(0));
goto err;
}
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
parser_name.str= (char*) next_chunk;
parser_name.length= strlen((char*) next_chunk);
BUG#27040 - Incorrect FTS parser name output When a table uses two (or more) fulltext keys with different fulltext parsers, all fulltext keys (excluding those that use built-in parser) were using first parser that was mentioned in .frm. With this fix fulltext keys use proper (assigned during create table) fulltext parser. This code was contributed by Yoshinori Matsunobu ymatsunobu@mysql.com on 2007-03-12 The copyright was assigned and transferred under the terms of the MySQL Contributor License Agreement.
2007-08-03 00:50:19 +02:00
next_chunk+= parser_name.length + 1;
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
keyinfo->parser= my_plugin_lock_by_name(NULL, &parser_name,
MYSQL_FTPARSER_PLUGIN);
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
if (! keyinfo->parser)
{
my_error(ER_PLUGIN_IS_NOT_LOADED, MYF(0), parser_name.str);
fixed reading of free'd memory when opening a table with non-existing plugin
2006-04-06 15:49:00 +02:00
my_free(buff, MYF(0));
Fix of crashed 5.1 tree (wrong merge + wrong pullout patch forced us to create a new 5.1 tree) This is a merge of 5.0 -> 5.1 + some code from old 5.1 tree to get all tests to work and keep the .frm format the same as the old 5.1 tree.
2005-11-05 12:20:35 +01:00
goto err;
}
}
}
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
if (forminfo[46] == (uchar)255)
{
//reading long table comment
if (next_chunk + 2 > buff_end)
{
DBUG_PRINT("error",
("long table comment is not defined in .frm"));
my_free(buff, MYF(0));
goto err;
}
share->comment.length = uint2korr(next_chunk);
if (! (share->comment.str= strmake_root(&share->mem_root,
(char*)next_chunk + 2, share->comment.length)))
{
my_free(buff, MYF(0));
goto err;
}
next_chunk+= 2 + share->comment.length;
}
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
my_free(buff, MYF(0));
}
Added support for key_block_size for key and table level (WL#602) Added support for key_block_size to MyISAM. Simplify interface to 'new Key' to make it easier to add new key options. mysqld option --new is used to define where key options are printed. (In 5.3 we should move all key options to after key part definition to avoid problem with reserved names) Fixed some compiler warnings and a memory leak in ssl
2006-05-03 14:59:17 +02:00
share->key_block_size= uint2korr(head+62);
Import changeset
2000-07-31 21:29:14 +02:00
error=4;
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
extra_rec_buf_length= uint2korr(head+59);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
rec_buff_length= ALIGN_SIZE(share->reclength + 1 + extra_rec_buf_length);
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->rec_buff_length= rec_buff_length;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(record= (uchar *) alloc_root(&share->mem_root,
rec_buff_length)))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
share->default_values= record;
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if (mysql_file_pread(file, record, (size_t) share->reclength,
record_offset, MYF(MY_NABP)))
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
goto err; /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
mysql_file_seek(file, pos+288, MY_SEEK_SET, MYF(0));
fixes/cleanups according to Coverity report
2005-01-24 15:48:25 +01:00
#ifdef HAVE_CRYPTED_FRM
Import changeset
2000-07-31 21:29:14 +02:00
if (crypted)
{
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
crypted->decode((char*) forminfo+256,288-256);
if (sint2korr(forminfo+284) != 0) // Should be 0
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; // Wrong password
Import changeset
2000-07-31 21:29:14 +02:00
}
fixes/cleanups according to Coverity report
2005-01-24 15:48:25 +01:00
#endif
Import changeset
2000-07-31 21:29:14 +02:00
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
share->fields= uint2korr(forminfo+258);
pos= uint2korr(forminfo+260); /* Length of all screens */
n_length= uint2korr(forminfo+268);
interval_count= uint2korr(forminfo+270);
interval_parts= uint2korr(forminfo+272);
int_length= uint2korr(forminfo+274);
share->null_fields= uint2korr(forminfo+282);
com_length= uint2korr(forminfo+284);
if (forminfo[46] != (uchar)255)
{
share->comment.length= (int) (forminfo[46]);
share->comment.str= strmake_root(&share->mem_root, (char*) forminfo+47,
share->comment.length);
}
Import changeset
2000-07-31 21:29:14 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
DBUG_PRINT("info",("i_count: %d i_parts: %d index: %d n_length: %d int_length: %d com_length: %d", interval_count,interval_parts, share->keys,n_length,int_length, com_length));
Import changeset
2000-07-31 21:29:14 +02:00
if (!(field_ptr = (Field **)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
alloc_root(&share->mem_root,
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
(uint) ((share->fields+1)*sizeof(Field*)+
Import changeset
2000-07-31 21:29:14 +02:00
interval_count*sizeof(TYPELIB)+
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
(share->fields+interval_parts+
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
keys+3)*sizeof(char *)+
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
(n_length+int_length+com_length)))))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->field= field_ptr;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
read_length=(uint) (share->fields * field_pack_length +
Changes for new binary .frm format Fixes after last merge from 4.0. (Code not yet complete, need anoter merge from 4.0)
2002-06-04 07:23:57 +02:00
pos+ (uint) (n_length+int_length+com_length));
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (read_string(file,(uchar**) &disk_buff,read_length))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
fixes/cleanups according to Coverity report
2005-01-24 15:48:25 +01:00
#ifdef HAVE_CRYPTED_FRM
Import changeset
2000-07-31 21:29:14 +02:00
if (crypted)
{
crypted->decode((char*) disk_buff,read_length);
delete crypted;
crypted=0;
}
fixes/cleanups according to Coverity report
2005-01-24 15:48:25 +01:00
#endif
Import changeset
2000-07-31 21:29:14 +02:00
strpos= disk_buff+pos;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->intervals= (TYPELIB*) (field_ptr+share->fields+1);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
interval_array= (const char **) (share->intervals+interval_count);
names= (char*) (interval_array+share->fields+interval_parts+keys+3);
Import changeset
2000-07-31 21:29:14 +02:00
if (!interval_count)
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->intervals= 0; // For better debugging
memcpy((char*) names, strpos+(share->fields*field_pack_length),
Import changeset
2000-07-31 21:29:14 +02:00
(uint) (n_length+int_length));
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
comment_pos= names+(n_length+int_length);
Changes for new binary .frm format Fixes after last merge from 4.0. (Code not yet complete, need anoter merge from 4.0)
2002-06-04 07:23:57 +02:00
memcpy(comment_pos, disk_buff+read_length-com_length, com_length);
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
fix_type_pointers(&interval_array, &share->fieldnames, 1, &names);
Merge mysql.com:/home/svoj/devel/mysql/BUG24358/mysql-4.1-engines into mysql.com:/home/svoj/devel/mysql/BUG24358/mysql-5.0-engines
2006-12-27 12:50:45 +01:00
if (share->fieldnames.count != share->fields)
goto err;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
fix_type_pointers(&interval_array, share->intervals, interval_count,
Import changeset
2000-07-31 21:29:14 +02:00
&names);
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
{
/* Set ENUM and SET lengths */
TYPELIB *interval;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
for (interval= share->intervals;
interval < share->intervals + interval_count;
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
interval++)
{
uint count= (uint) (interval->count + 1) * sizeof(uint);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (!(interval->type_lengths= (uint *) alloc_root(&share->mem_root,
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
count)))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
for (count= 0; count < interval->count; count++)
Bug#20922 mysql removes a name of first column in a table 0xFF is internal separator for SET|ENUM names. If this symbol is present in SET|ENUM names then we replace it with ','(deprecated symbol for SET|ENUM names) during frm creation and restore to 0xFF during frm opening
2006-09-11 11:50:46 +02:00
{
char *val= (char*) interval->type_names[count];
interval->type_lengths[count]= strlen(val);
}
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
interval->type_lengths[count]= 0;
}
}
Import changeset
2000-07-31 21:29:14 +02:00
if (keynames)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
fix_type_pointers(&interval_array, &share->keynames, 1, &keynames);
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/* Allocate handler */
if (!(handler_file= get_new_handler(share, thd->mem_root,
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
share->db_type())))
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
goto err;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
record= share->default_values-1; /* Fieldstart = 1 */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (share->null_field_first)
Import changeset
2000-07-31 21:29:14 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
null_flags= null_pos= (uchar*) record+1;
bit type
2004-12-17 15:06:05 +01:00
null_bit_pos= (db_create_options & HA_OPTION_PACK_RECORD) ? 0 : 1;
Merge bk-internal.mysql.com:/home/bk/mysql-5.0 into mysql.com:/home/my/mysql-5.0
2005-10-08 02:37:23 +02:00
/*
null_bytes below is only correct under the condition that
there are no bit fields. Correct values is set below after the
table struct is initialized
*/
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->null_bytes= (share->null_fields + null_bit_pos + 7) / 8;
Import changeset
2000-07-31 21:29:14 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
#ifndef WE_WANT_TO_SUPPORT_VERY_OLD_FRM_FILES
Import changeset
2000-07-31 21:29:14 +02:00
else
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->null_bytes= (share->null_fields+7)/8;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
null_flags= null_pos= (uchar*) (record + 1 +share->reclength -
share->null_bytes);
bit type
2004-12-17 15:06:05 +01:00
null_bit_pos= 0;
Import changeset
2000-07-31 21:29:14 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
#endif
Import changeset
2000-07-31 21:29:14 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
use_hash= share->fields >= MAX_FIELDS_BEFORE_HASH;
Import changeset
2000-07-31 21:29:14 +02:00
if (use_hash)
Backport of: ---------------------------------------------------------- revno: 2617.22.5 committer: Konstantin Osipov <kostja@sun.com> branch nick: mysql-6.0-runtime timestamp: Tue 2009-01-27 05:08:48 +0300 message: Remove non-prefixed use of HASH. Always use my_hash_init(), my_hash_inited(), my_hash_search(), my_hash_element(), my_hash_delete(), my_hash_free() rather than non-prefixed counterparts (hash_init(), etc). Remove the backward-compatible defines.
2009-10-14 18:37:38 +02:00
use_hash= !my_hash_init(&share->name_hash,
system_charset_info,
share->fields,0,0,
(my_hash_get_key) get_field_name,0,0);
Import changeset
2000-07-31 21:29:14 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
for (i=0 ; i < share->fields; i++, strpos+=field_pack_length, field_ptr++)
Import changeset
2000-07-31 21:29:14 +02:00
{
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
uint pack_flag, interval_nr, unireg_type, recpos, field_length;
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
enum_field_types field_type;
BINARY charset is now used instead of binary_flag
2002-10-25 10:58:32 +02:00
CHARSET_INFO *charset=NULL;
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
Field::geometry_type geom_type= Field::GEOM_GEOMETRY;
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
LEX_STRING comment;
Import changeset
2000-07-31 21:29:14 +02:00
Add support for up to VARCHAR (size up to 65535) Renamed HA_VAR_LENGTH to HA_VAR_LENGTH_PART Renamed in all files FIELD_TYPE_STRING and FIELD_TYPE_VAR_STRING to MYSQL_TYPE_STRING and MYSQL_TYPE_VAR_STRING to make it easy to catch all possible errors Added support for VARCHAR KEYS to heap Removed support for ISAM Now only long VARCHAR columns are changed to TEXT on demand (not CHAR) Internal temporary files can now use fixed length tables if the used VARCHAR columns are short
2004-12-06 01:00:37 +01:00
if (new_frm_ver >= 3)
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
{
/* new frm file in 4.1 */
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
field_length= uint2korr(strpos+3);
recpos= uint3korr(strpos+5);
pack_flag= uint2korr(strpos+8);
unireg_type= (uint) strpos[10];
interval_nr= (uint) strpos[12];
uint comment_length=uint2korr(strpos+15);
field_type=(enum_field_types) (uint) strpos[13];
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
Add support for up to VARCHAR (size up to 65535) Renamed HA_VAR_LENGTH to HA_VAR_LENGTH_PART Renamed in all files FIELD_TYPE_STRING and FIELD_TYPE_VAR_STRING to MYSQL_TYPE_STRING and MYSQL_TYPE_VAR_STRING to make it easy to catch all possible errors Added support for VARCHAR KEYS to heap Removed support for ISAM Now only long VARCHAR columns are changed to TEXT on demand (not CHAR) Internal temporary files can now use fixed length tables if the used VARCHAR columns are short
2004-12-06 01:00:37 +01:00
/* charset and geometry_type share the same byte in frm */
This finishes the work (someone) started to remove FIELD_ types and use the Enum MYSQL types. The second part to this is to actually deprecate the FIELD defines in mysql_com.h
2006-12-02 02:26:52 +01:00
if (field_type == MYSQL_TYPE_GEOMETRY)
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
{
SCRUM: WL#1163 (Making spatial code optional) Pack of changes to do in sql/ code.
2004-01-15 18:06:22 +01:00
#ifdef HAVE_SPATIAL
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
geom_type= (Field::geometry_type) strpos[14];
charset= &my_charset_bin;
SCRUM: WL#1163 (Making spatial code optional) Pack of changes to do in sql/ code.
2004-01-15 18:06:22 +01:00
#else
error= 4; // unsupported field type
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
SCRUM: WL#1163 (Making spatial code optional) Pack of changes to do in sql/ code.
2004-01-15 18:06:22 +01:00
#endif
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
}
else
{
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
uint csid= strpos[14] + (((uint) strpos[11]) << 8);
if (!csid)
Bugs#7278: Don't open a table if it contains columns with non-supported charsets
2004-12-22 15:56:57 +01:00
charset= &my_charset_bin;
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
else if (!(charset= get_charset(csid, MYF(0))))
Bugs#7278: Don't open a table if it contains columns with non-supported charsets
2004-12-22 15:56:57 +01:00
{
error= 5; // Unknown or unavailable charset
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
errarg= (int) csid;
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
Bugs#7278: Don't open a table if it contains columns with non-supported charsets
2004-12-22 15:56:57 +01:00
}
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
}
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
if (!comment_length)
{
comment.str= (char*) "";
comment.length=0;
}
else
{
comment.str= (char*) comment_pos;
comment.length= comment_length;
comment_pos+= comment_length;
}
}
else
{
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
field_length= (uint) strpos[3];
recpos= uint2korr(strpos+4),
pack_flag= uint2korr(strpos+6);
More fixes for strict mode: More tests. Better error messages. Fixed bug when checking if we updated all needed columns for INSERT. Give an error if we encounter a wrong float value during parsing. Don't print DEFAULT for columns without a default value in SHOW CREATE/SHOW FIELDS. Fixed UPDATE IGNORE when using STRICT mode.
2004-10-02 21:20:08 +02:00
pack_flag&= ~FIELDFLAG_NO_DEFAULT; // Safety for old files
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
unireg_type= (uint) strpos[8];
interval_nr= (uint) strpos[10];
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
/* old frm file */
field_type= (enum_field_types) f_packtype(pack_flag);
table.cc: Bug #6802 MySQL 4.0's VARCHAR(NN) BINARY is interpreted as VARBINARY(NN) in 4.1
2004-11-25 13:18:47 +01:00
if (f_is_binary(pack_flag))
{
/*
Try to choose the best 4.1 type:
- for 4.0 "CHAR(N) BINARY" or "VARCHAR(N) BINARY"
try to find a binary collation for character set.
- for other types (e.g. BLOB) just use my_charset_bin.
*/
if (!f_is_blob(pack_flag))
{
// 3.23 or 4.0 string
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
if (!(charset= get_charset_by_csname(share->table_charset->csname,
table.cc: Bug #6802 MySQL 4.0's VARCHAR(NN) BINARY is interpreted as VARBINARY(NN) in 4.1
2004-11-25 13:18:47 +01:00
MY_CS_BINSORT, MYF(0))))
charset= &my_charset_bin;
}
else
charset= &my_charset_bin;
}
else
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
charset= share->table_charset;
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
bzero((char*) &comment, sizeof(comment));
}
UCS2 support in ENUM and SET, which also fixes: Bug #5174 SHOW CREATE TABLE hangs up if the table contains half-with katakana enum values UCS2 values are stored in HEX encoding in FRM file
2004-12-06 17:45:32 +01:00
if (interval_nr && charset->mbminlen > 1)
{
/* Unescape UCS2 intervals from HEX notation */
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
TYPELIB *interval= share->intervals + interval_nr - 1;
Bug#7302: UCS2 data in ENUM field get truncated when new column is added
2004-12-21 14:12:27 +01:00
unhex_type2(interval);
UCS2 support in ENUM and SET, which also fixes: Bug #5174 SHOW CREATE TABLE hangs up if the table contains half-with katakana enum values UCS2 values are stored in HEX encoding in FRM file
2004-12-06 17:45:32 +01:00
}
Moved detection of wrong decimal fields before decimal field was created to avoid ASSERT() for DECIMAL(1)
2005-05-26 13:00:47 +02:00
#ifndef TO_BE_DELETED_ON_PRODUCTION
This finishes the work (someone) started to remove FIELD_ types and use the Enum MYSQL types. The second part to this is to actually deprecate the FIELD defines in mysql_com.h
2006-12-02 02:26:52 +01:00
if (field_type == MYSQL_TYPE_NEWDECIMAL && !share->mysql_version)
Moved detection of wrong decimal fields before decimal field was created to avoid ASSERT() for DECIMAL(1)
2005-05-26 13:00:47 +02:00
{
/*
Fix pack length of old decimal values from 5.0.3 -> 5.0.4
The difference is that in the old version we stored precision
in the .frm table while we now store the display_length
*/
uint decimals= f_decimals(pack_flag);
field_length= my_decimal_precision_to_length(field_length,
decimals,
f_is_dec(pack_flag) == 0);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
sql_print_error("Found incompatible DECIMAL field '%s' in %s; "
"Please do \"ALTER TABLE '%s' FORCE\" to fix it!",
share->fieldnames.type_names[i], share->table_name.str,
share->table_name.str);
WL#2110 (SIGNAL) WL#2265 (RESIGNAL) Manual merge of SIGNAL and RESIGNAL to mysql-trunk-signal, plus required dependencies.
2009-09-10 11:18:29 +02:00
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_WARN,
Moved detection of wrong decimal fields before decimal field was created to avoid ASSERT() for DECIMAL(1)
2005-05-26 13:00:47 +02:00
ER_CRASHED_ON_USAGE,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
"Found incompatible DECIMAL field '%s' in %s; "
"Please do \"ALTER TABLE '%s' FORCE\" to fix it!",
share->fieldnames.type_names[i],
share->table_name.str,
share->table_name.str);
Moved detection of wrong decimal fields before decimal field was created to avoid ASSERT() for DECIMAL(1)
2005-05-26 13:00:47 +02:00
share->crashed= 1; // Marker for CHECK TABLE
}
#endif
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
*field_ptr= reg_field=
make_field(share, record+recpos,
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
(uint32) field_length,
bit type
2004-12-17 15:06:05 +01:00
null_pos, null_bit_pos,
Import changeset
2000-07-31 21:29:14 +02:00
pack_flag,
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
field_type,
BINARY charset is now used instead of binary_flag
2002-10-25 10:58:32 +02:00
charset,
More OpenGIS compatibility: Point, MultiLine, Poligom, etc, are now not aliases for Geometry, but separate field types.
2003-03-27 10:09:09 +01:00
geom_type,
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
(Field::utype) MTYP_TYPENR(unireg_type),
Import changeset
2000-07-31 21:29:14 +02:00
(interval_nr ?
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->intervals+interval_nr-1 :
Import changeset
2000-07-31 21:29:14 +02:00
(TYPELIB*) 0),
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->fieldnames.type_names[i]);
Merge with 4.0
2002-11-21 14:56:48 +01:00
if (!reg_field) // Not supported field type
Put temporary files in binlog cache when using BEGIN/COMMIT Let MySQL 4.0 read 4.1 .frm files without 4.1 specific extensions New variables @@rand_seed1 and @@rand_seed2 (used by replication) DROP TEMPORARY TABLE
2002-11-07 03:02:37 +01:00
{
error= 4;
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
Put temporary files in binlog cache when using BEGIN/COMMIT Let MySQL 4.0 read 4.1 .frm files without 4.1 specific extensions New variables @@rand_seed1 and @@rand_seed2 (used by replication) DROP TEMPORARY TABLE
2002-11-07 03:02:37 +01:00
}
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
reg_field->field_index= i;
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
reg_field->comment=comment;
This finishes the work (someone) started to remove FIELD_ types and use the Enum MYSQL types. The second part to this is to actually deprecate the FIELD defines in mysql_com.h
2006-12-02 02:26:52 +01:00
if (field_type == MYSQL_TYPE_BIT && !f_bit_as_char(pack_flag))
Import changeset
2000-07-31 21:29:14 +02:00
{
bit type
2004-12-17 15:06:05 +01:00
if ((null_bit_pos+= field_length & 7) > 7)
Import changeset
2000-07-31 21:29:14 +02:00
{
bit type
2004-12-17 15:06:05 +01:00
null_pos++;
null_bit_pos-= 8;
Import changeset
2000-07-31 21:29:14 +02:00
}
}
bit type
2004-12-17 15:06:05 +01:00
if (!(reg_field->flags & NOT_NULL_FLAG))
{
if (!(null_bit_pos= (null_bit_pos + 1) & 7))
null_pos++;
}
More fixes for strict mode: More tests. Better error messages. Fixed bug when checking if we updated all needed columns for INSERT. Give an error if we encounter a wrong float value during parsing. Don't print DEFAULT for columns without a default value in SHOW CREATE/SHOW FIELDS. Fixed UPDATE IGNORE when using STRICT mode.
2004-10-02 21:20:08 +02:00
if (f_no_default(pack_flag))
reg_field->flags|= NO_DEFAULT_VALUE_FLAG;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
if (reg_field->unireg_check == Field::NEXT_NUMBER)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->found_next_number_field= field_ptr;
if (share->timestamp_field == reg_field)
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->timestamp_field_offset= i;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
if (use_hash)
Bug#45613 handle failures from my_hash_insert Not all my_hash_insert() calls are checked for return value. This patch adds appropriate checks and failure responses where needed.
2009-11-20 16:18:01 +01:00
if (my_hash_insert(&share->name_hash, (uchar*) field_ptr) )
{
/*
Set return code 8 here to indicate that an error has
occurred but that the error message already has been
sent (OOM).
*/
error= 8;
goto err;
}
Import changeset
2000-07-31 21:29:14 +02:00
}
*field_ptr=0; // End marker
/* Fix key->name and key_part->field */
if (key_parts)
{
fix for the 2361 bug: ALTER TABLE ... DROP PRIMARY KEY drops a non-primary key "PRIMARY" has been replaced by primary_key_name.
2004-01-14 13:01:55 +01:00
uint primary_key=(uint) (find_type((char*) primary_key_name,
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
&share->keynames, 3) - 1);
Fixed compiler warnings (Mostly VC++): - Removed not used variables - Changed some ulong parameters/variables to ulonglong (possible serious bug) - Added casts to get rid of safe assignment from longlong to long (and similar) - Added casts to function parameters - Fixed signed/unsigned compares - Added some constructores to structures - Removed some not portable constructs Better fix for bug Bug #21428 "skipped 9 bytes from file: socket (3)" on "mysqladmin shutdown" (Added new parameter to net_clear() to define when we want the communication buffer to be emptied)
2006-11-30 02:40:42 +01:00
longlong ha_option= handler_file->ha_table_flags();
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
keyinfo= share->key_info;
key_part= keyinfo->key_part;
Import changeset
2000-07-31 21:29:14 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
for (uint key=0 ; key < share->keys ; key++,keyinfo++)
Import changeset
2000-07-31 21:29:14 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
uint usable_parts= 0;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
keyinfo->name=(char*) share->keynames.type_names[key];
Made keyread (key scanning) a key specific attribute. This avoids using fulltext keys for table scanning. This also reverts Sinisa's original fix for this problem.
2002-06-18 23:22:30 +02:00
/* Fix fulltext keys for old .frm files */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (share->key_info[key].flags & HA_FULLTEXT)
share->key_info[key].algorithm= HA_KEY_ALG_FULLTEXT;
Made keyread (key scanning) a key specific attribute. This avoids using fulltext keys for table scanning. This also reverts Sinisa's original fix for this problem.
2002-06-18 23:22:30 +02:00
Reverted wrong bug fix (Bug#11228)
2006-06-30 23:37:20 +02:00
if (primary_key >= MAX_KEY && (keyinfo->flags & HA_NOSAME))
{
/*
If the UNIQUE key doesn't have NULL columns and is not a part key
declare this as a primary key.
*/
primary_key=key;
for (i=0 ; i < keyinfo->key_parts ;i++)
{
uint fieldnr= key_part[i].fieldnr;
if (!fieldnr ||
share->field[fieldnr-1]->null_ptr ||
share->field[fieldnr-1]->key_length() !=
key_part[i].length)
{
primary_key=MAX_KEY; // Can't be used
break;
}
}
}
Import changeset
2000-07-31 21:29:14 +02:00
for (i=0 ; i < keyinfo->key_parts ; key_part++,i++)
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Field *field;
Import changeset
2000-07-31 21:29:14 +02:00
if (new_field_pack_flag <= 1)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
key_part->fieldnr= (uint16) find_field(share->field,
Added --debug-info to most clients to detect memory leaks in mysql-test-run Moved .progress files into the log directory Moved 'cluster' database tables into the MySQL database, to not have 'cluster' beeing a reserved database name Fixed bug where mysqld got a core dump when trying to use a table created by MySQL 3.23 Fixed some compiler warnings Fixed small memory leak in libmysql Note that this doesn't changeset doesn't include the new mysqldump.c code required to run some tests. This will be added when I merge 5.0 to 5.1
2006-11-21 21:32:58 +01:00
share->default_values,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
(uint) key_part->offset,
(uint) key_part->length);
if (!key_part->fieldnr)
{
error= 4; // Wrong file
goto err;
}
field= key_part->field= share->field[key_part->fieldnr-1];
Merge mysql.com:/home/my/mysql-5.0 into mysql.com:/home/my/mysql-5.1
2006-11-29 22:07:15 +01:00
key_part->type= field->key_type();
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (field->null_ptr)
{
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
key_part->null_offset=(uint) ((uchar*) field->null_ptr -
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->default_values);
key_part->null_bit= field->null_bit;
key_part->store_length+=HA_KEY_NULL_LENGTH;
keyinfo->flags|=HA_NULL_PART_KEY;
keyinfo->extra_length+= HA_KEY_NULL_LENGTH;
keyinfo->key_length+= HA_KEY_NULL_LENGTH;
}
This finishes the work (someone) started to remove FIELD_ types and use the Enum MYSQL types. The second part to this is to actually deprecate the FIELD defines in mysql_com.h
2006-12-02 02:26:52 +01:00
if (field->type() == MYSQL_TYPE_BLOB ||
Merge magare.gmz:/home/kgeorge/mysql/autopush/B30825-new-5.0-opt into magare.gmz:/home/kgeorge/mysql/work/B30825-5.1-opt
2007-10-23 10:44:14 +02:00
field->real_type() == MYSQL_TYPE_VARCHAR ||
field->type() == MYSQL_TYPE_GEOMETRY)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
Merge magare.gmz:/home/kgeorge/mysql/autopush/B30825-new-5.0-opt into magare.gmz:/home/kgeorge/mysql/work/B30825-5.1-opt
2007-10-23 10:44:14 +02:00
if (field->type() == MYSQL_TYPE_BLOB ||
field->type() == MYSQL_TYPE_GEOMETRY)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
key_part->key_part_flag|= HA_BLOB_PART;
else
key_part->key_part_flag|= HA_VAR_LENGTH_PART;
keyinfo->extra_length+=HA_KEY_BLOB_LENGTH;
key_part->store_length+=HA_KEY_BLOB_LENGTH;
keyinfo->key_length+= HA_KEY_BLOB_LENGTH;
}
if (field->type() == MYSQL_TYPE_BIT)
key_part->key_part_flag|= HA_BIT_PART;
if (i == 0 && key != primary_key)
field->flags |= (((keyinfo->flags & HA_NOSAME) &&
(keyinfo->key_parts == 1)) ?
UNIQUE_KEY_FLAG : MULTIPLE_KEY_FLAG);
if (i == 0)
field->key_start.set_bit(key);
if (field->key_length() == key_part->length &&
!(field->flags & BLOB_FLAG))
{
if (handler_file->index_flags(key, i, 0) & HA_KEYREAD_ONLY)
{
share->keys_for_keyread.set_bit(key);
field->part_of_key.set_bit(key);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
field->part_of_key_not_clustered.set_bit(key);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
}
if (handler_file->index_flags(key, i, 1) & HA_READ_ORDER)
field->part_of_sortkey.set_bit(key);
}
if (!(key_part->key_part_flag & HA_REVERSE_SORT) &&
usable_parts == i)
usable_parts++; // For FILESORT
field->flags|= PART_KEY_FLAG;
if (key == primary_key)
{
field->flags|= PRI_KEY_FLAG;
/*
If this field is part of the primary key and all keys contains
the primary key, then we can use any key to find this column
*/
if (ha_option & HA_PRIMARY_KEY_IN_READ_INDEX)
Merge gleb.loc:/home/uchum/work/bk/5.0-opt into gleb.loc:/home/uchum/work/bk/5.1-opt
2007-07-25 22:23:39 +02:00
{
Bug #37742: HA_EXTRA_KEYREAD flag is set when key contains only prefix of requested column When the storage engine uses secondary keys clustered with the primary key MySQL was adding the primary key parts to each secondary key. In doing so it was not checking whether the index was on full columns and this resulted in the secondary keys being added to the list of covering keys even if they have partial columns. Fixed by not adding a primary key part to the list of columns that can be used for index read of the secondary keys when the primary key part is a partial key part.
2008-11-29 14:36:17 +01:00
if (field->key_length() == key_part->length &&
!(field->flags & BLOB_FLAG))
field->part_of_key= share->keys_in_use;
Merge gleb.loc:/home/uchum/work/bk/5.0-opt into gleb.loc:/home/uchum/work/bk/5.1-opt
2007-07-27 10:38:11 +02:00
if (field->part_of_sortkey.is_set(key))
Merge gleb.loc:/home/uchum/work/bk/5.0-opt into gleb.loc:/home/uchum/work/bk/5.1-opt
2007-07-25 22:23:39 +02:00
field->part_of_sortkey= share->keys_in_use;
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
}
if (field->key_length() != key_part->length)
{
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
#ifndef TO_BE_DELETED_ON_PRODUCTION
This finishes the work (someone) started to remove FIELD_ types and use the Enum MYSQL types. The second part to this is to actually deprecate the FIELD defines in mysql_com.h
2006-12-02 02:26:52 +01:00
if (field->type() == MYSQL_TYPE_NEWDECIMAL)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
/*
Fix a fatal error in decimal key handling that causes crashes
on Innodb. We fix it by reducing the key length so that
InnoDB never gets a too big key when searching.
This allows the end user to do an ALTER TABLE to fix the
error.
*/
keyinfo->key_length-= (key_part->length - field->key_length());
key_part->store_length-= (uint16)(key_part->length -
field->key_length());
key_part->length= (uint16)field->key_length();
sql_print_error("Found wrong key definition in %s; "
"Please do \"ALTER TABLE '%s' FORCE \" to fix it!",
share->table_name.str,
share->table_name.str);
WL#2110 (SIGNAL) WL#2265 (RESIGNAL) Manual merge of SIGNAL and RESIGNAL to mysql-trunk-signal, plus required dependencies.
2009-09-10 11:18:29 +02:00
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_WARN,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
ER_CRASHED_ON_USAGE,
"Found wrong key definition in %s; "
"Please do \"ALTER TABLE '%s' FORCE\" to fix "
"it!",
share->table_name.str,
share->table_name.str);
share->crashed= 1; // Marker for CHECK TABLE
goto to_be_deleted;
}
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
#endif
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
key_part->key_part_flag|= HA_PART_KEY_SEG;
}
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
to_be_deleted:
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
/*
If the field can be NULL, don't optimize away the test
key_part_column = expression from the WHERE clause
as we need to test for NULL = NULL.
*/
if (field->real_maybe_null())
Merge mysql.com:/usr/home/ram/work/bug20732/my50-bug20732 into mysql.com:/usr/home/ram/work/bug20732/my51-bug20732
2006-10-23 07:32:39 +02:00
key_part->key_part_flag|= HA_NULL_PART;
Import changeset
2000-07-31 21:29:14 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
keyinfo->usable_key_parts= usable_parts; // Filesort
Fix max_key_length and max_unique_length to take into account varchar and null extra lengths This fixes a bug found by bdb.test Fixes a portability issue with bdb tables
2004-12-31 10:56:50 +01:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
set_if_bigger(share->max_key_length,keyinfo->key_length+
Fix max_key_length and max_unique_length to take into account varchar and null extra lengths This fixes a bug found by bdb.test Fixes a portability issue with bdb tables
2004-12-31 10:56:50 +01:00
keyinfo->key_parts);
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->total_key_length+= keyinfo->key_length;
Bug#10400 - Improperly-defined MERGE table crashes with INSERT ... ON DUPLICATE KEY UPDATE After review version. Added a condition for MERGE tables. These do not have unique indexes. But every key could be a unique key on the underlying MyISAM table. So get the maximum key length for MERGE tables instead of the maximum unique key length. This is used for buffer allocation in write_record().
2005-05-18 19:40:39 +02:00
/*
MERGE tables do not have unique indexes. But every key could be
an unique index on the underlying MyISAM table. (Bug #10400)
*/
if ((keyinfo->flags & HA_NOSAME) ||
(ha_option & HA_ANY_INDEX_MAY_BE_UNIQUE))
Resolve merge from 4.1
2005-05-18 22:54:36 +02:00
set_if_bigger(share->max_unique_length,keyinfo->key_length);
Import changeset
2000-07-31 21:29:14 +02:00
}
bugs #701 and #783 - old .frm files were read in incorrectly in 4.1 (each field was assigned a charset, BLOBs were converted to TEXT)
2003-07-28 16:58:52 +02:00
if (primary_key < MAX_KEY &&
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
(share->keys_in_use.is_set(primary_key)))
Import changeset
2000-07-31 21:29:14 +02:00
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->primary_key= primary_key;
Automatic primary key for BDB tables
2000-10-16 01:29:48 +02:00
/*
If we are using an integer as the primary key then allow the user to
refer to it as '_rowid'
*/
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (share->key_info[primary_key].key_parts == 1)
Import changeset
2000-07-31 21:29:14 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Field *field= share->key_info[primary_key].key_part[0].field;
Import changeset
2000-07-31 21:29:14 +02:00
if (field && field->result_type() == INT_RESULT)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
/* note that fieldnr here (and rowid_field_offset) starts from 1 */
share->rowid_field_offset= (share->key_info[primary_key].key_part[0].
fieldnr);
}
Import changeset
2000-07-31 21:29:14 +02:00
}
}
else
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->primary_key = MAX_KEY; // we do not have a primary key
Import changeset
2000-07-31 21:29:14 +02:00
}
Give an error if you use a BLOB(X) where X > 255 Fixes for MIT pthreads
2000-10-22 00:19:05 +02:00
else
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
share->primary_key= MAX_KEY;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
x_free((uchar*) disk_buff);
Automatic primary key for BDB tables
2000-10-16 01:29:48 +02:00
disk_buff=0;
Import changeset
2000-07-31 21:29:14 +02:00
if (new_field_pack_flag <= 1)
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
{
/* Old file format with default as not null */
uint null_length= (share->null_fields+7)/8;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
bfill(share->default_values + (null_flags - (uchar*) record),
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
null_length, 255);
Import changeset
2000-07-31 21:29:14 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (share->found_next_number_field)
New SET syntax & system variables. Made a some new buffers thread specific and changeable. Resize of key_buffer. AUTO_COMMIT -> AUTOCOMMIT Fixed mutex bug in DROP DATABASE Fixed bug when using auto_increment as second part of a key where first part could include NULL. Split handler->extra() to extra() and extra_opt() to be able to support thread specific buffers. Don't write message to error log when slave reconnects becasue of timeout. Fixed possible update problem when using DELETE/UPDATE on small tables (In some cases we used index even if table scanning would be better) A lot of minior code cleanups
2002-07-23 17:31:22 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
reg_field= *share->found_next_number_field;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
if ((int) (share->next_number_index= (uint)
Added --debug-info to most clients to detect memory leaks in mysql-test-run Moved .progress files into the log directory Moved 'cluster' database tables into the MySQL database, to not have 'cluster' beeing a reserved database name Fixed bug where mysqld got a core dump when trying to use a table created by MySQL 3.23 Fixed some compiler warnings Fixed small memory leak in libmysql Note that this doesn't changeset doesn't include the new mysqldump.c code required to run some tests. This will be added when I merge 5.0 to 5.1
2006-11-21 21:32:58 +01:00
find_ref_key(share->key_info, share->keys,
share->default_values, reg_field,
WL#3700: Handler API change: all index search methods - that is, index_read(), index_read_idx(), index_read_last(), and records_in_range() - instead of 'uint keylen' argument take 'ulonglong keypart_map', a bitmap showing which keyparts are present in the key value. Fallback method is provided for handlers that are lagging behind.
2007-01-29 10:40:26 +01:00
&share->next_number_key_offset,
&share->next_number_keypart)) < 0)
New SET syntax & system variables. Made a some new buffers thread specific and changeable. Resize of key_buffer. AUTO_COMMIT -> AUTOCOMMIT Fixed mutex bug in DROP DATABASE Fixed bug when using auto_increment as second part of a key where first part could include NULL. Split handler->extra() to extra() and extra_opt() to be able to support thread specific buffers. Don't write message to error log when slave reconnects becasue of timeout. Fixed possible update problem when using DELETE/UPDATE on small tables (In some cases we used index even if table scanning would be better) A lot of minior code cleanups
2002-07-23 17:31:22 +02:00
{
Added --debug-info to most clients to detect memory leaks in mysql-test-run Moved .progress files into the log directory Moved 'cluster' database tables into the MySQL database, to not have 'cluster' beeing a reserved database name Fixed bug where mysqld got a core dump when trying to use a table created by MySQL 3.23 Fixed some compiler warnings Fixed small memory leak in libmysql Note that this doesn't changeset doesn't include the new mysqldump.c code required to run some tests. This will be added when I merge 5.0 to 5.1
2006-11-21 21:32:58 +01:00
/* Wrong field definition */
WL#3700: Handler API change: all index search methods - that is, index_read(), index_read_idx(), index_read_last(), and records_in_range() - instead of 'uint keylen' argument take 'ulonglong keypart_map', a bitmap showing which keyparts are present in the key value. Fallback method is provided for handlers that are lagging behind.
2007-01-29 10:40:26 +01:00
error= 4;
goto err;
New SET syntax & system variables. Made a some new buffers thread specific and changeable. Resize of key_buffer. AUTO_COMMIT -> AUTOCOMMIT Fixed mutex bug in DROP DATABASE Fixed bug when using auto_increment as second part of a key where first part could include NULL. Split handler->extra() to extra() and extra_opt() to be able to support thread specific buffers. Don't write message to error log when slave reconnects becasue of timeout. Fixed possible update problem when using DELETE/UPDATE on small tables (In some cases we used index even if table scanning would be better) A lot of minior code cleanups
2002-07-23 17:31:22 +02:00
}
else
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
reg_field->flags |= AUTO_INCREMENT_FLAG;
New SET syntax & system variables. Made a some new buffers thread specific and changeable. Resize of key_buffer. AUTO_COMMIT -> AUTOCOMMIT Fixed mutex bug in DROP DATABASE Fixed bug when using auto_increment as second part of a key where first part could include NULL. Split handler->extra() to extra() and extra_opt() to be able to support thread specific buffers. Don't write message to error log when slave reconnects becasue of timeout. Fixed possible update problem when using DELETE/UPDATE on small tables (In some cases we used index even if table scanning would be better) A lot of minior code cleanups
2002-07-23 17:31:22 +02:00
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
if (share->blob_fields)
Import changeset
2000-07-31 21:29:14 +02:00
{
Field **ptr;
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
uint k, *save;
Import changeset
2000-07-31 21:29:14 +02:00
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
/* Store offsets to blob fields to find them fast */
if (!(share->blob_field= save=
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
(uint*) alloc_root(&share->mem_root,
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
(uint) (share->blob_fields* sizeof(uint)))))
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err;
Merge mysql.com:/home/my/mysql-5.0 into mysql.com:/home/my/mysql-5.1 Merge of 'remove compiler warnings when using -Wshadow'
2007-01-27 02:46:45 +01:00
for (k=0, ptr= share->field ; *ptr ; ptr++, k++)
Import changeset
2000-07-31 21:29:14 +02:00
{
if ((*ptr)->flags & BLOB_FLAG)
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
(*save++)= k;
Import changeset
2000-07-31 21:29:14 +02:00
}
}
Merge bk-internal.mysql.com:/home/bk/mysql-5.0 into mysql.com:/home/my/mysql-5.0
2005-10-08 02:37:23 +02:00
/*
the correct null_bytes can now be set, since bitfields have been taken
into account
*/
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->null_bytes= (null_pos - (uchar*) null_flags +
Merge bk-internal.mysql.com:/home/bk/mysql-5.0 into mysql.com:/home/my/mysql-5.0
2005-10-08 02:37:23 +02:00
(null_bit_pos + 7) / 8);
sql_table.cc, table.cc, table.h: fixed CHECKSUM TABLE to be independent from last - undefined in InnoDB - bits in the null bitmask bug#13710
2005-10-04 17:04:20 +02:00
share->last_null_bit_pos= null_bit_pos;
Corrected last fix for Bug #13152 bit fields and fields that follow become corrupted when dumped from NDB tables
2005-09-19 15:52:57 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->db_low_byte_first= handler_file->low_byte_first();
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
share->column_bitmap_size= bitmap_buffer_size(share->fields);
if (!(bitmaps= (my_bitmap_map*) alloc_root(&share->mem_root,
After merge fixes Removed a lot of compiler warnings Removed not used variables, functions and labels Initialize some variables that could be used unitialized (fatal bugs) %ll -> %l
2007-01-29 00:47:35 +01:00
share->column_bitmap_size)))
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
goto err;
bitmap_init(&share->all_set, bitmaps, share->fields, FALSE);
bitmap_set_all(&share->all_set);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
delete handler_file;
#ifndef DBUG_OFF
if (use_hash)
Backport of: ---------------------------------------------------------- revno: 2617.22.5 committer: Konstantin Osipov <kostja@sun.com> branch nick: mysql-6.0-runtime timestamp: Tue 2009-01-27 05:08:48 +0300 message: Remove non-prefixed use of HASH. Always use my_hash_init(), my_hash_inited(), my_hash_search(), my_hash_element(), my_hash_delete(), my_hash_free() rather than non-prefixed counterparts (hash_init(), etc). Remove the backward-compatible defines.
2009-10-14 18:37:38 +02:00
(void) my_hash_check(&share->name_hash);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
#endif
DBUG_RETURN (0);
err:
share->error= error;
share->open_errno= my_errno;
share->errarg= errarg;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
x_free((uchar*) disk_buff);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
delete crypted;
delete handler_file;
Backport of: ---------------------------------------------------------- revno: 2617.22.5 committer: Konstantin Osipov <kostja@sun.com> branch nick: mysql-6.0-runtime timestamp: Tue 2009-01-27 05:08:48 +0300 message: Remove non-prefixed use of HASH. Always use my_hash_init(), my_hash_inited(), my_hash_search(), my_hash_element(), my_hash_delete(), my_hash_free() rather than non-prefixed counterparts (hash_init(), etc). Remove the backward-compatible defines.
2009-10-14 18:37:38 +02:00
my_hash_free(&share->name_hash);
Bug #37433 Deadlock between open_table, close_open_tables, get_table_share, drop_open_table In the partition handler code, LOCK_open and share->LOCK_ha_data are acquired in the wrong order in certain cases. When doing a multi-row INSERT (i.e a INSERT..SELECT) in a table with auto- increment column(s). the increments must be in a monotonically continuous increasing sequence (i.e it can't have "holes"). To achieve this, a lock is held for the duration of the operation. share->LOCK_ha_data was used for this purpose. Whenever there was a need to open a view _during_ the operation (views are not currently pre-opened the way tables are), and LOCK_open was grabbed, a deadlock could occur. share->LOCK_ha_data is other places used _while_ holding LOCK_open. A new mutex was introduced in the HA_DATA_PARTITION structure, for exclusive use of the autoincrement data fields, so we don't need to overload the use of LOCK_ha_data here. A module test case has not been supplied, since the problem occurs as a result of a race condition, and testing for this condition is thus not deterministic. Testing for it could be done by setting up a test case as described in the bug report.
2009-10-15 13:07:04 +02:00
if (share->ha_data_destroy)
share->ha_data_destroy(share->ha_data);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
open_table_error(share, error, share->open_errno, errarg);
DBUG_RETURN(error);
} /* open_binary_frm */
/*
Open a table based on a TABLE_SHARE
SYNOPSIS
open_table_from_share()
thd Thread handler
share Table definition
alias Alias for table
db_stat open flags (for example HA_OPEN_KEYFILE|
HA_OPEN_RNDFILE..) can be 0 (example in
ha_example_table)
prgflag READ_ALL etc..
ha_open_flags HA_OPEN_ABORT_IF_LOCKED etc..
outparam result table
RETURN VALUES
0 ok
1 Error (see open_table_error)
2 Error (see open_table_error)
3 Wrong data in .frm file
4 Error (see open_table_error)
5 Error (see open_table_error: charset unavailable)
7 Table definition has changed in engine
*/
int open_table_from_share(THD *thd, TABLE_SHARE *share, const char *alias,
uint db_stat, uint prgflag, uint ha_open_flags,
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
TABLE *outparam, bool is_create_table)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
int error;
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
uint records, i, bitmap_size;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
bool error_reported= FALSE;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uchar *record, *bitmaps;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
Field **field_ptr;
DBUG_ENTER("open_table_from_share");
DBUG_PRINT("enter",("name: '%s.%s' form: 0x%lx", share->db.str,
Fixed a LOT of compiler warnings Added missing DBUG_RETURN statements (in mysqldump.c) Added missing enums Fixed a lot of wrong DBUG_PRINT() statements, some of which could cause crashes Removed usage of %lld and %p in printf strings as these are not portable or produces different results on different systems.
2006-11-27 00:47:38 +01:00
share->table_name.str, (long) outparam));
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
error= 1;
bzero((char*) outparam, sizeof(*outparam));
outparam->in_use= thd;
outparam->s= share;
outparam->db_stat= db_stat;
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
outparam->write_row_record= NULL;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
init_sql_alloc(&outparam->mem_root, TABLE_ALLOC_BLOCK_SIZE, 0);
if (!(outparam->alias= my_strdup(alias, MYF(MY_WME))))
goto err;
outparam->quick_keys.init();
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
outparam->covering_keys.init();
Bug#36751: Segmentation fault in ctype-bin.c:308; Linux 86_64, with-max-indexes=128 mysqld is optimized for the default case (up to 64-indices); for a greater number of indices it goes through a different code path. As that code-path is a compile-time option and can not easily be covered in standard tests, bitrot occurred. key-fields need an explicit initialization in the non- optimized case; this setup was presumably not added when a new key- vector was added. Changeset adds the necessary initialisations. No test case added due to dependence on compile-time option.
2009-03-11 19:09:56 +01:00
outparam->merge_keys.init();
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
outparam->keys_in_use_for_query.init();
/* Allocate handler */
WL#3732 Information schema optimization
2007-08-03 00:14:05 +02:00
outparam->file= 0;
if (!(prgflag & OPEN_FRM_FILE_ONLY))
{
if (!(outparam->file= get_new_handler(share, &outparam->mem_root,
share->db_type())))
goto err;
}
else
{
DBUG_ASSERT(!db_stat);
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
error= 4;
outparam->reginfo.lock_type= TL_UNLOCK;
outparam->current_lock= F_UNLCK;
records=0;
if ((db_stat & HA_OPEN_KEYFILE) || (prgflag & DELAYED_OPEN))
records=1;
if (prgflag & (READ_ALL+EXTRA_RECORD))
records++;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(record= (uchar*) alloc_root(&outparam->mem_root,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
share->rec_buff_length * records)))
goto err; /* purecov: inspected */
if (records == 0)
{
/* We are probably in hard repair, and the buffers should not be used */
outparam->record[0]= outparam->record[1]= share->default_values;
}
else
{
outparam->record[0]= record;
if (records > 1)
outparam->record[1]= record+ share->rec_buff_length;
else
outparam->record[1]= outparam->record[0]; // Safety
}
#ifdef HAVE_purify
/*
We need this because when we read var-length rows, we are not updating
bytes after end of varchar
*/
if (records > 1)
{
memcpy(outparam->record[0], share->default_values, share->rec_buff_length);
Bug #24487 Valgrind: uninited byte in table->record[1] in binlog code for rbr + inno db The reason of this valgrind's compaint is not a bug but rather a feature of bitwise ops: for any value of the byte x x | 1 -> 1, and x & 0 -> 0. x, being a null_byte part of record[1] can be left unassigned even after ha_innobase::index_read_idx because the above and still be correct. Addding a check memory upon the invocation of the function can detect this fact long before record[1], old record, is eventually passed to my_write. Fixed with initialization of record[1]'s null_bytes part in open_table_from_share.
2006-11-30 18:56:34 +01:00
memcpy(outparam->record[1], share->default_values, share->null_bytes);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (records > 2)
memcpy(outparam->record[1], share->default_values,
share->rec_buff_length);
}
#endif
if (!(field_ptr = (Field **) alloc_root(&outparam->mem_root,
(uint) ((share->fields+1)*
sizeof(Field*)))))
goto err; /* purecov: inspected */
outparam->field= field_ptr;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
record= (uchar*) outparam->record[0]-1; /* Fieldstart = 1 */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (share->null_field_first)
outparam->null_flags= (uchar*) record+1;
else
outparam->null_flags= (uchar*) (record+ 1+ share->reclength -
share->null_bytes);
/* Setup copy of fields from share, but use the right alias and record */
for (i=0 ; i < share->fields; i++, field_ptr++)
{
if (!((*field_ptr)= share->field[i]->clone(&outparam->mem_root, outparam)))
goto err;
}
(*field_ptr)= 0; // End marker
if (share->found_next_number_field)
outparam->found_next_number_field=
outparam->field[(uint) (share->found_next_number_field - share->field)];
if (share->timestamp_field)
outparam->timestamp_field= (Field_timestamp*) outparam->field[share->timestamp_field_offset];
/* Fix key->name and key_part->field */
if (share->key_parts)
{
KEY *key_info, *key_info_end;
KEY_PART_INFO *key_part;
uint n_length;
n_length= share->keys*sizeof(KEY) + share->key_parts*sizeof(KEY_PART_INFO);
if (!(key_info= (KEY*) alloc_root(&outparam->mem_root, n_length)))
goto err;
outparam->key_info= key_info;
key_part= (my_reinterpret_cast(KEY_PART_INFO*) (key_info+share->keys));
memcpy(key_info, share->key_info, sizeof(*key_info)*share->keys);
memcpy(key_part, share->key_info[0].key_part, (sizeof(*key_part) *
share->key_parts));
for (key_info_end= key_info + share->keys ;
key_info < key_info_end ;
key_info++)
{
KEY_PART_INFO *key_part_end;
key_info->table= outparam;
key_info->key_part= key_part;
for (key_part_end= key_part+ key_info->key_parts ;
key_part < key_part_end ;
key_part++)
{
Field *field= key_part->field= outparam->field[key_part->fieldnr-1];
if (field->key_length() != key_part->length &&
!(field->flags & BLOB_FLAG))
{
/*
We are using only a prefix of the column as a key:
Create a new field for the key part that matches the index
*/
field= key_part->field=field->new_field(&outparam->mem_root,
After merge fix.
2006-07-06 20:12:33 +02:00
outparam, 0);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
field->field_length= key_part->length;
}
}
}
}
#ifdef WITH_PARTITION_STORAGE_ENGINE
Bug#30322 Server crashes on selecting from i_s.columns when cluster is running -regression disable partition processing if we open frm file only
2007-08-09 19:05:56 +02:00
if (share->partition_info_len && outparam->file)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
BUG#21658: Crash when creating table with item in prepared statement that allocates memory in fix_fields We need to use an arena to indicate we are preparing a statement when loading partition function and parsing it as part of an open table.
2006-08-22 22:52:25 +02:00
/*
In this execution we must avoid calling thd->change_item_tree since
we might release memory before statement is completed. We do this
by changing to a new statement arena. As part of this arena we also
set the memory root to be the memory root of the table since we
call the parser and fix_fields which both can allocate memory for
item objects. We keep the arena to ensure that we can release the
free_list when closing the table object.
SEE Bug #21658
*/
Query_arena *backup_stmt_arena_ptr= thd->stmt_arena;
Query_arena backup_arena;
Query_arena part_func_arena(&outparam->mem_root, Query_arena::INITIALIZED);
thd->set_n_backup_active_arena(&part_func_arena, &backup_arena);
thd->stmt_arena= &part_func_arena;
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
bool tmp;
Bug #27084 partitioning by list seems failing when using case creation of the partitioned table could fail as we created Item-s for it's list function in thd->mem_root, and then do Item->fix_fields in the context of other table->mem_root (so that memory alloced there was alloced in this table->mem_root). As we freed the table->mem_root before we do thd->free_items, our Item-s had pointers to the freed memory, that caused the crash
2007-03-27 13:15:38 +02:00
bool work_part_info_used;
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
Merge bk-internal.mysql.com:/home/bk/mysql-5.1-new into mysql.com:/home/my/mysql-5.1
2006-06-04 20:05:22 +02:00
tmp= mysql_unpack_partition(thd, share->partition_info,
share->partition_info_len,
Remove unnecessary casts to uchar. The casts are stemming from the lexer API which internally uses unsigned char variables to address its state map. The implementation of the lexer should be internal to the lexer, and not influence the rest of the code.
2007-03-27 19:09:56 +02:00
share->part_state,
Merge bk-internal.mysql.com:/home/bk/mysql-5.1-new into mysql.com:/home/my/mysql-5.1
2006-06-04 20:05:22 +02:00
share->part_state_len,
outparam, is_create_table,
Bug #27084 partitioning by list seems failing when using case creation of the partitioned table could fail as we created Item-s for it's list function in thd->mem_root, and then do Item->fix_fields in the context of other table->mem_root (so that memory alloced there was alloced in this table->mem_root). As we freed the table->mem_root before we do thd->free_items, our Item-s had pointers to the freed memory, that caused the crash
2007-03-27 13:15:38 +02:00
share->default_part_db_type,
&work_part_info_used);
Bug#32158 Crash in open_table_from_share, on mysql_unpack_partition errors added check for the result of mysql_unpack_partition()
2007-11-19 09:42:25 +01:00
if (tmp)
{
thd->stmt_arena= backup_stmt_arena_ptr;
thd->restore_active_arena(&part_func_arena, &backup_arena);
goto partititon_err;
}
outparam->part_info->is_auto_partitioned= share->auto_partitioned;
Merge bk-internal.mysql.com:/home/bk/mysql-5.1-new into mysql.com:/home/my/mysql-5.1
2006-06-04 20:05:22 +02:00
DBUG_PRINT("info", ("autopartitioned: %u", share->auto_partitioned));
Bug #27084 partitioning by list seems failing when using case creation of the partitioned table could fail as we created Item-s for it's list function in thd->mem_root, and then do Item->fix_fields in the context of other table->mem_root (so that memory alloced there was alloced in this table->mem_root). As we freed the table->mem_root before we do thd->free_items, our Item-s had pointers to the freed memory, that caused the crash
2007-03-27 13:15:38 +02:00
/* we should perform the fix_partition_func in either local or
caller's arena depending on work_part_info_used value
*/
Bug#32158 Crash in open_table_from_share, on mysql_unpack_partition errors added check for the result of mysql_unpack_partition()
2007-11-19 09:42:25 +01:00
if (!work_part_info_used)
Merge c-0409e253.1238-1-64736c10.cust.bredbandsbolaget.se:/home/pappa/clean-mysql-5.1 into c-0409e253.1238-1-64736c10.cust.bredbandsbolaget.se:/home/pappa/bug16002
2006-06-14 15:12:07 +02:00
tmp= fix_partition_func(thd, outparam, is_create_table);
BUG#21658: Crash when creating table with item in prepared statement that allocates memory in fix_fields We need to use an arena to indicate we are preparing a statement when loading partition function and parsing it as part of an open table.
2006-08-22 22:52:25 +02:00
thd->stmt_arena= backup_stmt_arena_ptr;
thd->restore_active_arena(&part_func_arena, &backup_arena);
BUG#21658: Crash partition test in prepared statement protocol Review fixes
2006-08-31 10:16:22 +02:00
if (!tmp)
Bug #27084 partitioning by list seems failing when using case creation of the partitioned table could fail as we created Item-s for it's list function in thd->mem_root, and then do Item->fix_fields in the context of other table->mem_root (so that memory alloced there was alloced in this table->mem_root). As we freed the table->mem_root before we do thd->free_items, our Item-s had pointers to the freed memory, that caused the crash
2007-03-27 13:15:38 +02:00
{
if (work_part_info_used)
tmp= fix_partition_func(thd, outparam, is_create_table);
BUG#21658: Crash partition test in prepared statement protocol Review fixes
2006-08-31 10:16:22 +02:00
outparam->part_info->item_free_list= part_func_arena.free_list;
Bug #27084 partitioning by list seems failing when using case creation of the partitioned table could fail as we created Item-s for it's list function in thd->mem_root, and then do Item->fix_fields in the context of other table->mem_root (so that memory alloced there was alloced in this table->mem_root). As we freed the table->mem_root before we do thd->free_items, our Item-s had pointers to the freed memory, that caused the crash
2007-03-27 13:15:38 +02:00
}
Bug#32158 Crash in open_table_from_share, on mysql_unpack_partition errors added check for the result of mysql_unpack_partition()
2007-11-19 09:42:25 +01:00
partititon_err:
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
if (tmp)
BUG#16000: .par file not removed plus errors to error log for normal errors
2006-06-20 19:24:30 +02:00
{
if (is_create_table)
{
/*
During CREATE/ALTER TABLE it is ok to receive errors here.
It is not ok if it happens during the opening of an frm
file as part of a normal query.
*/
error_reported= TRUE;
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
goto err;
BUG#16000: .par file not removed plus errors to error log for normal errors
2006-06-20 19:24:30 +02:00
}
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
}
#endif
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
/* Allocate bitmaps */
bitmap_size= share->column_bitmap_size;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(bitmaps= (uchar*) alloc_root(&outparam->mem_root, bitmap_size*3)))
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
goto err;
bitmap_init(&outparam->def_read_set,
(my_bitmap_map*) bitmaps, share->fields, FALSE);
bitmap_init(&outparam->def_write_set,
(my_bitmap_map*) (bitmaps+bitmap_size), share->fields, FALSE);
bitmap_init(&outparam->tmp_set,
(my_bitmap_map*) (bitmaps+bitmap_size*2), share->fields, FALSE);
outparam->default_column_bitmaps();
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
/* The table struct is now initialized; Open the table */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
error= 2;
Automatic primary key for BDB tables
2000-10-16 01:29:48 +02:00
if (db_stat)
{
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
int ha_err;
if ((ha_err= (outparam->file->
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
ha_open(outparam, share->normalized_path.str,
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
(db_stat & HA_READ_ONLY ? O_RDONLY : O_RDWR),
(db_stat & HA_OPEN_TEMPORARY ? HA_OPEN_TMP_TABLE :
((db_stat & HA_WAIT_IF_LOCKED) ||
(specialflag & SPECIAL_WAIT_IF_LOCKED)) ?
HA_OPEN_WAIT_IF_LOCKED :
(db_stat & (HA_ABORT_IF_LOCKED | HA_GET_INFO)) ?
HA_OPEN_ABORT_IF_LOCKED :
HA_OPEN_IGNORE_IF_LOCKED) | ha_open_flags))))
Automatic primary key for BDB tables
2000-10-16 01:29:48 +02:00
{
/* Set a flag if the table is crashed and it can be auto. repaired */
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
share->crashed= ((ha_err == HA_ERR_CRASHED_ON_USAGE) &&
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
outparam->file->auto_repair() &&
!(ha_open_flags & HA_OPEN_FOR_REPAIR));
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
ndb - fiexed for bug#15021, binlog_index table become inconsistent if errors during purge of binlogs. improved the original patch, changed if/else to switch/case.
2006-11-28 04:14:36 +01:00
switch (ha_err)
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
{
ndb - fiexed for bug#15021, binlog_index table become inconsistent if errors during purge of binlogs. improved the original patch, changed if/else to switch/case.
2006-11-28 04:14:36 +01:00
case HA_ERR_NO_SUCH_TABLE:
/*
The table did not exists in storage engine, use same error message
as if the .frm file didn't exist
*/
error= 1;
my_errno= ENOENT;
break;
case EMFILE:
/*
Too many files opened, use same error message as if the .frm
file can't open
*/
DBUG_PRINT("error", ("open file: %s failed, too many files opened (errno: %d)",
share->normalized_path.str, ha_err));
error= 1;
my_errno= EMFILE;
break;
default:
outparam->file->print_error(ha_err, MYF(0));
error_reported= TRUE;
if (ha_err == HA_ERR_TABLE_DEF_CHANGED)
error= 7;
break;
WL#1895 - Print message to error log in case of detected MyISAM corruption Changed my_error() to print error messages, which come from arbitrary registered ranges of error messages. Messages can be unregistered (and should be at end of the program). Added registration of handler error messages. Added a new mi_print_error() macro and a new mi_report_error() function, which supply error messages with a table name. Added calls to mi_print_error() or mi_report_error() at all places in MyISAM, where table corruption is detected.
2004-12-23 20:11:38 +01:00
}
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
goto err; /* purecov: inspected */
Automatic primary key for BDB tables
2000-10-16 01:29:48 +02:00
}
}
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
#if defined(HAVE_purify) && !defined(DBUG_OFF)
bzero((char*) bitmaps, bitmap_size*3);
#endif
after-merge fix
2007-08-03 22:02:19 +02:00
outparam->no_replicate= outparam->file &&
test(outparam->file->ha_table_flags() &
Simplify logging code a bit (to make code smaller and faster) Moved duplicated code to inline function store_timestamp() Save thd->time_zone_used when logging to table as CSV internally cases it to be changed Added MYSQL_LOCK_IGNORE_FLUSH to log tables to avoid deadlock in case of flush tables. Mark log tables with TIMESTAMP_NO_AUTO_SET to avoid automatic timestamping Set TABLE->no_replicate on open
2007-08-03 00:14:27 +02:00
HA_HAS_OWN_BINLOGGING);
Removed wrong warnings in test suite (This was because select_insert / select_create results was not freed. Added thd to openfrm() for initialization of TABLE->in_use. This fixed a bug in BDB handling where table->in_use was used early
2004-09-17 02:08:23 +02:00
thd->status_var.opened_tables++;
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN (0);
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
err:
WL#1895 - Print message to error log in case of detected MyISAM corruption Changed my_error() to print error messages, which come from arbitrary registered ranges of error messages. Messages can be unregistered (and should be at end of the program). Added registration of handler error messages. Added a new mi_print_error() macro and a new mi_report_error() function, which supply error messages with a table name. Added calls to mi_print_error() or mi_report_error() at all places in MyISAM, where table corruption is detected.
2004-12-23 20:11:38 +01:00
if (! error_reported)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
open_table_error(share, error, my_errno, 0);
Automatic repair of MyISAM tables + portability fixes
2000-10-10 23:06:37 +02:00
delete outparam->file;
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
#ifdef WITH_PARTITION_STORAGE_ENGINE
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (outparam->part_info)
free_items(outparam->part_info->item_free_list);
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
#endif
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
outparam->file= 0; // For easier error checking
Fixed problem with auto-repair of MyISAM tables Fixed bug in ISAM and MyISAM when updating from multiple-processes
2000-11-21 02:43:34 +01:00
outparam->db_stat=0;
Fix accesses to uninitialized memory (found by valgrind)
2005-01-13 02:02:49 +01:00
free_root(&outparam->mem_root, MYF(0)); // Safe to call on bzero'd root
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
my_free((char*) outparam->alias, MYF(MY_ALLOW_ZERO_PTR));
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN (error);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
Fixed compiler and valgrind warnings Added missing DBUG_xxx_RETURN statements Fixed some usage of not initialized variables (as found by valgrind) Ensure that we don't remove locked tables used as name locks from open table cache until unlock_table_names() are called. This was fixed by having drop_locked_name() returning any table used as a name lock so that we can free it in unlock_table_names() This will allow Tomas to continue with his work to use namelocks to syncronize things. Note: valgrind still produces a lot of warnings about using not initialized code and shows memory loss errors when running the ndb tests
2006-03-29 13:27:36 +02:00
/*
Free information allocated by openfrm
SYNOPSIS
closefrm()
table TABLE object to free
free_share Is 1 if we also want to free table_share
*/
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
int closefrm(register TABLE *table, bool free_share)
Import changeset
2000-07-31 21:29:14 +02:00
{
int error=0;
DBUG_ENTER("closefrm");
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
DBUG_PRINT("enter", ("table: 0x%lx", (long) table));
Fixed compiler and valgrind warnings Added missing DBUG_xxx_RETURN statements Fixed some usage of not initialized variables (as found by valgrind) Ensure that we don't remove locked tables used as name locks from open table cache until unlock_table_names() are called. This was fixed by having drop_locked_name() returning any table used as a name lock so that we can free it in unlock_table_names() This will allow Tomas to continue with his work to use namelocks to syncronize things. Note: valgrind still produces a lot of warnings about using not initialized code and shows memory loss errors when running the ndb tests
2006-03-29 13:27:36 +02:00
Import changeset
2000-07-31 21:29:14 +02:00
if (table->db_stat)
error=table->file->close();
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
my_free((char*) table->alias, MYF(MY_ALLOW_ZERO_PTR));
table->alias= 0;
if (table->field)
Import changeset
2000-07-31 21:29:14 +02:00
{
for (Field **ptr=table->field ; *ptr ; ptr++)
delete *ptr;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
table->field= 0;
Import changeset
2000-07-31 21:29:14 +02:00
}
delete table->file;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
table->file= 0; /* For easier errorchecking */
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
#ifdef WITH_PARTITION_STORAGE_ENGINE
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (table->part_info)
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
free_items(table->part_info->item_free_list);
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
table->part_info->item_free_list= 0;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
table->part_info= 0;
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
}
#endif
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (free_share)
{
if (table->s->tmp_table == NO_TMP_TABLE)
Backport of: ------------------------------------------------------------ revno: 2630.6.1 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-3726 timestamp: Tue 2008-05-27 13:45:34 +0400 message: Remove an unused argument from release_table_share(). Remove unused members from TABLE_SHARE struct. Review comments in scope of WL#3726 "DDL locking for all metadata objects"
2009-11-30 20:11:32 +01:00
release_table_share(table->s);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
else
free_table_share(table->s);
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
free_root(&table->mem_root, MYF(0));
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(error);
}
/* Deallocate temporary blob storage */
void free_blobs(register TABLE *table)
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
uint *ptr, *end;
for (ptr= table->s->blob_field, end=ptr + table->s->blob_fields ;
ptr != end ;
ptr++)
((Field_blob*) table->field[*ptr])->free();
Import changeset
2000-07-31 21:29:14 +02:00
}
Bug#38002 table_cache consumes too much memory with blobs Tables in the table definition cache are keeping a cache buffer for blob fields which can consume a lot of memory. This patch introduces a maximum size threshold for these buffers.
2008-07-24 22:38:44 +02:00
/**
Reclaim temporary blob storage which is bigger than
a threshold.
@param table A handle to the TABLE object containing blob fields
@param size The threshold value.
*/
void free_field_buffers_larger_than(TABLE *table, uint32 size)
{
uint *ptr, *end;
for (ptr= table->s->blob_field, end=ptr + table->s->blob_fields ;
ptr != end ;
ptr++)
{
Field_blob *blob= (Field_blob*) table->field[*ptr];
if (blob->get_field_buffer_size() > size)
blob->free();
}
}
Import changeset
2000-07-31 21:29:14 +02:00
/* Find where a form starts */
/* if formname is NullS then only formnames is read */
ulong get_form_pos(File file, uchar *head, TYPELIB *save_names)
{
uint a_length,names,length;
uchar *pos,*buf;
ulong ret_value=0;
DBUG_ENTER("get_form_pos");
names=uint2korr(head+8);
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
a_length=(names+2)*sizeof(char *); /* Room for two extra */
Import changeset
2000-07-31 21:29:14 +02:00
if (!save_names)
a_length=0;
else
save_names->type_names=0; /* Clear if error */
if (names)
{
length=uint2korr(head+4);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_seek(file, 64L, MY_SEEK_SET, MYF(0));
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (!(buf= (uchar*) my_malloc((size_t) length+a_length+names*4,
Import changeset
2000-07-31 21:29:14 +02:00
MYF(MY_WME))) ||
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_read(file, buf+a_length, (size_t) (length+names*4),
MYF(MY_NABP)))
Import changeset
2000-07-31 21:29:14 +02:00
{ /* purecov: inspected */
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
x_free((uchar*) buf); /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(0L); /* purecov: inspected */
}
pos= buf+a_length+length;
ret_value=uint4korr(pos);
}
if (! save_names)
many warnings (practically safe but annoying) corrected
2006-01-03 17:54:54 +01:00
{
if (names)
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
my_free((uchar*) buf,MYF(0));
many warnings (practically safe but annoying) corrected
2006-01-03 17:54:54 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
else if (!names)
bzero((char*) save_names,sizeof(save_names));
else
{
char *str;
str=(char *) (buf+a_length);
fix_type_pointers((const char ***) &buf,save_names,1,&str);
}
DBUG_RETURN(ret_value);
}
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
/*
Read string from a file with malloc
Import changeset
2000-07-31 21:29:14 +02:00
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
NOTES:
We add an \0 at end of the read string to make reading of C strings easier
*/
int read_string(File file, uchar**to, size_t length)
Import changeset
2000-07-31 21:29:14 +02:00
{
DBUG_ENTER("read_string");
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
x_free(*to);
if (!(*to= (uchar*) my_malloc(length+1,MYF(MY_WME))) ||
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_read(file, *to, length, MYF(MY_NABP)))
Import changeset
2000-07-31 21:29:14 +02:00
{
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
x_free(*to); /* purecov: inspected */
*to= 0; /* purecov: inspected */
DBUG_RETURN(1); /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
}
*((char*) *to+length)= '\0';
DBUG_RETURN (0);
} /* read_string */
/* Add a new form to a form file */
ulong make_new_entry(File file, uchar *fileinfo, TYPELIB *formnames,
const char *newname)
{
uint i,bufflength,maxlength,n_length,length,names;
ulong endpos,newpos;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uchar buff[IO_SIZE];
Import changeset
2000-07-31 21:29:14 +02:00
uchar *pos;
DBUG_ENTER("make_new_entry");
FLUSH TABLE table_list Fixes for RENAME TABLE Portability fixes
2000-08-21 23:18:32 +02:00
length=(uint) strlen(newname)+1;
Import changeset
2000-07-31 21:29:14 +02:00
n_length=uint2korr(fileinfo+4);
maxlength=uint2korr(fileinfo+6);
names=uint2korr(fileinfo+8);
newpos=uint4korr(fileinfo+10);
if (64+length+n_length+(names+1)*4 > maxlength)
{ /* Expand file */
newpos+=IO_SIZE;
int4store(fileinfo+10,newpos);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
/* Copy from file-end */
endpos= (ulong) mysql_file_seek(file, 0L, MY_SEEK_END, MYF(0));
Import changeset
2000-07-31 21:29:14 +02:00
bufflength= (uint) (endpos & (IO_SIZE-1)); /* IO_SIZE is a power of 2 */
while (endpos > maxlength)
{
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_seek(file, (ulong) (endpos-bufflength), MY_SEEK_SET, MYF(0));
if (mysql_file_read(file, buff, bufflength, MYF(MY_NABP+MY_WME)))
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(0L);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_seek(file, (ulong) (endpos-bufflength+IO_SIZE), MY_SEEK_SET,
MYF(0));
if ((mysql_file_write(file, buff, bufflength, MYF(MY_NABP+MY_WME))))
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(0);
endpos-=bufflength; bufflength=IO_SIZE;
}
bzero(buff,IO_SIZE); /* Null new block */
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_seek(file, (ulong) maxlength, MY_SEEK_SET, MYF(0));
if (mysql_file_write(file, buff, bufflength, MYF(MY_NABP+MY_WME)))
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(0L);
maxlength+=IO_SIZE; /* Fix old ref */
int2store(fileinfo+6,maxlength);
for (i=names, pos= (uchar*) *formnames->type_names+n_length-1; i-- ;
pos+=4)
{
endpos=uint4korr(pos)+IO_SIZE;
int4store(pos,endpos);
}
}
if (n_length == 1 )
{ /* First name */
length++;
Backport of: ---------------------------------------------------------------------- ChangeSet@1.2571, 2008-04-08 12:30:06+02:00, vvaintroub@wva. +122 -0 Bug#32082 : definition of VOID in my_global.h conflicts with Windows SDK headers VOID macro is now removed. Its usage is replaced with void cast. In some cases, where cast does not make much sense (pthread_*, printf, hash_delete, my_seek), cast is ommited.
2009-11-24 14:54:59 +01:00
(void) strxmov((char*) buff,"/",newname,"/",NullS);
Import changeset
2000-07-31 21:29:14 +02:00
}
else
Backport of: ---------------------------------------------------------------------- ChangeSet@1.2571, 2008-04-08 12:30:06+02:00, vvaintroub@wva. +122 -0 Bug#32082 : definition of VOID in my_global.h conflicts with Windows SDK headers VOID macro is now removed. Its usage is replaced with void cast. In some cases, where cast does not make much sense (pthread_*, printf, hash_delete, my_seek), cast is ommited.
2009-11-24 14:54:59 +01:00
(void) strxmov((char*) buff,newname,"/",NullS); /* purecov: inspected */
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
mysql_file_seek(file, 63L+(ulong) n_length, MY_SEEK_SET, MYF(0));
if (mysql_file_write(file, buff, (size_t) length+1, MYF(MY_NABP+MY_WME)) ||
(names && mysql_file_write(file,
(uchar*) (*formnames->type_names+n_length-1),
names*4, MYF(MY_NABP+MY_WME))) ||
mysql_file_write(file, fileinfo+10, 4, MYF(MY_NABP+MY_WME)))
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(0L); /* purecov: inspected */
int2store(fileinfo+8,names+1);
int2store(fileinfo+4,n_length+length);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
(void) mysql_file_chsize(file, newpos, 0, MYF(MY_WME));/* Append file with '\0' */
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(newpos);
} /* make_new_entry */
/* error message when opening a form file */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
void open_table_error(TABLE_SHARE *share, int error, int db_errno, int errarg)
Import changeset
2000-07-31 21:29:14 +02:00
{
int err_no;
char buff[FN_REFLEN];
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
myf errortype= ME_ERROR+ME_WAITTANG;
DBUG_ENTER("open_table_error");
Import changeset
2000-07-31 21:29:14 +02:00
switch (error) {
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
case 7:
Import changeset
2000-07-31 21:29:14 +02:00
case 1:
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (db_errno == ENOENT)
my_error(ER_NO_SUCH_TABLE, MYF(0), share->db.str, share->table_name.str);
else
Import changeset
2000-07-31 21:29:14 +02:00
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
strxmov(buff, share->normalized_path.str, reg_ext, NullS);
Merge neptunus.(none):/home/msvensson/mysql/bug15020/my50-bug15020 into neptunus.(none):/home/msvensson/mysql/bug15020/my51-bug15020
2006-01-19 11:43:19 +01:00
my_error((db_errno == EMFILE) ? ER_CANT_OPEN_FILE : ER_FILE_NOT_FOUND,
errortype, buff, db_errno);
Import changeset
2000-07-31 21:29:14 +02:00
}
break;
case 2:
{
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
handler *file= 0;
const char *datext= "";
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
if (share->db_type() != NULL)
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
if ((file= get_new_handler(share, current_thd->mem_root,
WL#2936 "Server Variables for Plugins" Implement support for plugins to declare server variables. Demonstrate functionality by removing InnoDB specific code from sql/* New feature for HASH - HASH_UNIQUE flag New feature for DYNAMIC_ARRAY - initializer accepts preallocated ptr. Completed support for plugin reference counting.
2007-03-02 17:43:45 +01:00
share->db_type())))
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
{
if (!(datext= *file->bas_ext()))
datext= "";
}
}
err_no= (db_errno == ENOENT) ? ER_FILE_NOT_FOUND : (db_errno == EAGAIN) ?
Import changeset
2000-07-31 21:29:14 +02:00
ER_FILE_USED : ER_CANT_OPEN_FILE;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
strxmov(buff, share->normalized_path.str, datext, NullS);
my_error(err_no,errortype, buff, db_errno);
delete file;
Import changeset
2000-07-31 21:29:14 +02:00
break;
}
Bugs#7278: Don't open a table if it contains columns with non-supported charsets
2004-12-22 15:56:57 +01:00
case 5:
{
const char *csname= get_charset_name((uint) errarg);
char tmp[10];
if (!csname || csname[0] =='?')
{
my_snprintf(tmp, sizeof(tmp), "#%d", errarg);
csname= tmp;
}
my_printf_error(ER_UNKNOWN_COLLATION,
"Unknown collation '%s' in table '%-.64s' definition",
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
MYF(0), csname, share->table_name.str);
Import changeset
2000-07-31 21:29:14 +02:00
break;
}
Add special error message for .frm files with an un-understood version, like those from 5.0 with new datatypes. (Bug #10435)
2005-05-07 22:57:22 +02:00
case 6:
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
strxmov(buff, share->normalized_path.str, reg_ext, NullS);
Add special error message for .frm files with an un-understood version, like those from 5.0 with new datatypes. (Bug #10435)
2005-05-07 22:57:22 +02:00
my_printf_error(ER_NOT_FORM_FILE,
"Table '%-.64s' was created with a different version "
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
"of MySQL and cannot be read",
MYF(0), buff);
Add special error message for .frm files with an un-understood version, like those from 5.0 with new datatypes. (Bug #10435)
2005-05-07 22:57:22 +02:00
break;
BUG#22708 - Error message doesn't refer to storage engine unsupported When openning a table with unsupported (disabled or not compiled) storage engine, confusing error message is returned. Return better error message when we're attempting to open a table that uses unsupported engine.
2007-12-14 14:05:27 +01:00
case 8:
break;
Import changeset
2000-07-31 21:29:14 +02:00
default: /* Better wrong error than none */
case 4:
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
strxmov(buff, share->normalized_path.str, reg_ext, NullS);
my_error(ER_NOT_FORM_FILE, errortype, buff, 0);
Import changeset
2000-07-31 21:29:14 +02:00
break;
}
DBUG_VOID_RETURN;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
} /* open_table_error */
Import changeset
2000-07-31 21:29:14 +02:00
/*
** fix a str_type to a array type
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
** typeparts separated with some char. differents types are separated
Import changeset
2000-07-31 21:29:14 +02:00
** with a '\0'
*/
static void
fix_type_pointers(const char ***array, TYPELIB *point_to_type, uint types,
char **names)
{
char *type_name, *ptr;
char chr;
ptr= *names;
while (types--)
{
point_to_type->name=0;
point_to_type->type_names= *array;
if ((chr= *ptr)) /* Test if empty type */
{
while ((type_name=strchr(ptr+1,chr)) != NullS)
{
*((*array)++) = ptr+1;
*type_name= '\0'; /* End string */
ptr=type_name;
}
Portability fix Cleanup typos (like SKIPP -> SKIP)
2001-11-06 23:13:29 +01:00
ptr+=2; /* Skip end mark and last 0 */
Import changeset
2000-07-31 21:29:14 +02:00
}
else
ptr++;
point_to_type->count= (uint) (*array - point_to_type->type_names);
point_to_type++;
*((*array)++)= NullS; /* End of type */
}
*names=ptr; /* Update end */
return;
} /* fix_type_pointers */
A fix and a test case for Bug#14410 "Crash in Enum or Set type in CREATE TABLE and PS/SP": make sure that 'typelib' object for ENUM values and 'Item_string' object for DEFAULT clause are created in the statement memory root.
2005-11-25 11:25:31 +01:00
TYPELIB *typelib(MEM_ROOT *mem_root, List<String> &strings)
Import changeset
2000-07-31 21:29:14 +02:00
{
A fix and a test case for Bug#14410 "Crash in Enum or Set type in CREATE TABLE and PS/SP": make sure that 'typelib' object for ENUM values and 'Item_string' object for DEFAULT clause are created in the statement memory root.
2005-11-25 11:25:31 +01:00
TYPELIB *result= (TYPELIB*) alloc_root(mem_root, sizeof(TYPELIB));
Import changeset
2000-07-31 21:29:14 +02:00
if (!result)
return 0;
result->count=strings.elements;
result->name="";
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
uint nbytes= (sizeof(char*) + sizeof(uint)) * (result->count + 1);
A fix and a test case for Bug#14410 "Crash in Enum or Set type in CREATE TABLE and PS/SP": make sure that 'typelib' object for ENUM values and 'Item_string' object for DEFAULT clause are created in the statement memory root.
2005-11-25 11:25:31 +01:00
if (!(result->type_names= (const char**) alloc_root(mem_root, nbytes)))
Import changeset
2000-07-31 21:29:14 +02:00
return 0;
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
result->type_lengths= (uint*) (result->type_names + result->count + 1);
Import changeset
2000-07-31 21:29:14 +02:00
List_iterator<String> it(strings);
String *tmp;
for (uint i=0; (tmp=it++) ; i++)
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
{
result->type_names[i]= tmp->ptr();
result->type_lengths[i]= tmp->length();
}
result->type_names[result->count]= 0; // End marker
result->type_lengths[result->count]= 0;
Import changeset
2000-07-31 21:29:14 +02:00
return result;
}
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
/*
Search after a field with given start & length
If an exact field isn't found, return longest field with starts
at right position.
NOTES
This is needed because in some .frm fields 'fieldnr' was saved wrong
RETURN
0 error
# field number +1
*/
Import changeset
2000-07-31 21:29:14 +02:00
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
static uint find_field(Field **fields, uchar *record, uint start, uint length)
Import changeset
2000-07-31 21:29:14 +02:00
{
Field **field;
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
uint i, pos;
Import changeset
2000-07-31 21:29:14 +02:00
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
pos= 0;
for (field= fields, i=1 ; *field ; i++,field++)
Import changeset
2000-07-31 21:29:14 +02:00
{
Added --debug-info to most clients to detect memory leaks in mysql-test-run Moved .progress files into the log directory Moved 'cluster' database tables into the MySQL database, to not have 'cluster' beeing a reserved database name Fixed bug where mysqld got a core dump when trying to use a table created by MySQL 3.23 Fixed some compiler warnings Fixed small memory leak in libmysql Note that this doesn't changeset doesn't include the new mysqldump.c code required to run some tests. This will be added when I merge 5.0 to 5.1
2006-11-21 21:32:58 +01:00
if ((*field)->offset(record) == start)
Import changeset
2000-07-31 21:29:14 +02:00
{
if ((*field)->key_length() == length)
return (i);
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
if (!pos || fields[pos-1]->pack_length() <
Import changeset
2000-07-31 21:29:14 +02:00
(*field)->pack_length())
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
pos= i;
Import changeset
2000-07-31 21:29:14 +02:00
}
}
return (pos);
}
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
/* Check that the integer is in the internal */
Import changeset
2000-07-31 21:29:14 +02:00
int set_zone(register int nr, int min_zone, int max_zone)
{
if (nr<=min_zone)
return (min_zone);
if (nr>=max_zone)
return (max_zone);
return (nr);
} /* set_zone */
/* Adjust number to next larger disk buffer */
ulong next_io_size(register ulong pos)
{
reg2 ulong offset;
if ((offset= pos & (IO_SIZE-1)))
return pos-offset+IO_SIZE;
return pos;
} /* next_io_size */
proper fix for the bug #2077: accented characters in enum/defaul values are reported incorrectly. note: bar asked me to use res.charset in ::sql_type() functions to be more consistent.
2003-12-16 14:39:33 +01:00
/*
Store an SQL quoted string.
SYNOPSIS
append_unescaped()
res result String
pos string to be quoted
length it's length
NOTE
This function works correctly with utf8 or single-byte charset strings.
May fail with some multibyte charsets though.
*/
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
proper fix for the bug #2077: accented characters in enum/defaul values are reported incorrectly. note: bar asked me to use res.charset in ::sql_type() functions to be more consistent.
2003-12-16 14:39:33 +01:00
void append_unescaped(String *res, const char *pos, uint length)
Import changeset
2000-07-31 21:29:14 +02:00
{
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
const char *end= pos+length;
res->append('\'');
for (; pos != end ; pos++)
Import changeset
2000-07-31 21:29:14 +02:00
{
Code style fixes. Initialize LOG_error_log before get_options to not use an uninitalized mutex in case of an error from handle_options()
2004-08-27 23:49:54 +02:00
#if defined(USE_MB) && MYSQL_VERSION_ID < 40100
table.cc: Bug #4558 Escape handling error for ENUM values in SJIS encoding
2004-08-27 09:09:28 +02:00
uint mblen;
if (use_mb(default_charset_info) &&
(mblen= my_ismbchar(default_charset_info, pos, end)))
{
res->append(pos, mblen);
pos+= mblen;
continue;
}
#endif
Import changeset
2000-07-31 21:29:14 +02:00
switch (*pos) {
case 0: /* Must be escaped for 'mysql' */
res->append('\\');
res->append('0');
break;
case '\n': /* Must be escaped for logs */
res->append('\\');
res->append('n');
break;
case '\r':
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
res->append('\\'); /* This gives better readability */
Import changeset
2000-07-31 21:29:14 +02:00
res->append('r');
break;
case '\\':
res->append('\\'); /* Because of the sql syntax */
res->append('\\');
break;
case '\'':
res->append('\''); /* Because of the sql syntax */
res->append('\'');
break;
default:
res->append(*pos);
break;
}
}
Extension of .frm file (not yet ready for push)
2002-06-02 20:22:20 +02:00
res->append('\'');
Import changeset
2000-07-31 21:29:14 +02:00
}
WL#1324 table name to file name encoding - Encoding itself, implemented as a charset "filename". Originally planned to use '.' as an escape character, but now changed to '@' for two reasons: "ls" does not return file names starting with '.' considering them as a kind of hidden files; some platforms do not allow several dots in a file name. - replacing many calls of my_snprintf() and strnxmov() to the new build_table_filename(). - Adding MY_APPEND_EXT mysys flag, to append an extention rather that replace it. - Replacing all numeric constants in fn_format flag arguments to their mysys definitions, e.g. MY_UNPACK_FILENAME, - Predictability in several function/methods: when a table name can appear with or withot .frm extension. Some functions/methods were changed so accept names strictly with .frm, other - strictly without .frm extensions. Several DBUG_ASSERTs were added to check whether an extension is passed. Many files: table name to file name encoding mysql_priv.h: Prototypes for new table name encoding tools. ctype-utf8.c: Implementing "filename" charset for table name to file name encoding. row0mysql.c: Fixing table name prefix. mf_format.c: Adding MY_APPEND_EXT processing. Many files: Fixing tests. my_sys.h: Adding new flag to append rather than replace an extension. m_ctype.h: Adding "filename" charset definition.
2005-12-31 06:01:26 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
/* Create a .frm file */
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
File create_frm(THD *thd, const char *name, const char *db,
Merge mysql.com:/home/my/mysql-4.1 into mysql.com:/home/my/mysql-5.0
2005-07-28 16:09:54 +02:00
const char *table, uint reclength, uchar *fileinfo,
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
HA_CREATE_INFO *create_info, uint keys, KEY *key_info)
Import changeset
2000-07-31 21:29:14 +02:00
{
register File file;
ulong length;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uchar fill[IO_SIZE];
Fixes for bugs reported by Stefano Di Paola (stefano.dipaola@wisec.it)
2005-03-03 19:51:29 +01:00
int create_flags= O_RDWR | O_TRUNC;
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
ulong key_comment_total_bytes= 0;
uint i;
Fixes for bugs reported by Stefano Di Paola (stefano.dipaola@wisec.it)
2005-03-03 19:51:29 +01:00
if (create_info->options & HA_LEX_CREATE_TMP_TABLE)
create_flags|= O_EXCL | O_NOFOLLOW;
Import changeset
2000-07-31 21:29:14 +02:00
After merge fixes Return NULL if a time argument is given to date_add(). (Warning will be shown after Dimitri's timezone patch is pushed)
2004-06-21 09:21:20 +02:00
/* Fix this when we have new .frm files; Current limit is 4G rows (QQ) */
Fix handling of maximum value for MAX_ROWS on 64-bit platforms. (Bug #14155)
2005-11-24 02:05:59 +01:00
if (create_info->max_rows > UINT_MAX32)
create_info->max_rows= UINT_MAX32;
if (create_info->min_rows > UINT_MAX32)
create_info->min_rows= UINT_MAX32;
Fix for bug #15602: 5.0.17 test case 'create' failure.
2005-12-09 16:55:59 +01:00
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if ((file= mysql_file_create(key_file_frm,
name, CREATE_MODE, create_flags, MYF(0))) >= 0)
Import changeset
2000-07-31 21:29:14 +02:00
{
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
uint key_length, tmp_key_length, tmp, csid;
Import changeset
2000-07-31 21:29:14 +02:00
bzero((char*) fileinfo,64);
Add support for up to VARCHAR (size up to 65535) Renamed HA_VAR_LENGTH to HA_VAR_LENGTH_PART Renamed in all files FIELD_TYPE_STRING and FIELD_TYPE_VAR_STRING to MYSQL_TYPE_STRING and MYSQL_TYPE_VAR_STRING to make it easy to catch all possible errors Added support for VARCHAR KEYS to heap Removed support for ISAM Now only long VARCHAR columns are changed to TEXT on demand (not CHAR) Internal temporary files can now use fixed length tables if the used VARCHAR columns are short
2004-12-06 01:00:37 +01:00
/* header */
fileinfo[0]=(uchar) 254;
fileinfo[1]= 1;
fileinfo[2]= FRM_VER+3+ test(create_info->varchar);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
fileinfo[3]= (uchar) ha_legacy_type(
ha_checktype(thd,ha_legacy_type(create_info->db_type),0,0));
Import changeset
2000-07-31 21:29:14 +02:00
fileinfo[4]=1;
int2store(fileinfo+6,IO_SIZE); /* Next block starts here */
Bug #26642: create index corrupts table definition in .frm Thanks to Martin Friebe for finding and submitting a fix for this bug! A table with maximum number of key segments and maximum length key name would have a corrupted .frm file, due to an incorrect calculation of the complete key length. Now the key length is computed correctly (I hope) :-) MyISAM would reject a table with the maximum number of keys and the maximum number of key segments in all keys. It would allow one less than this total maximum. Now MyISAM accepts a table defined with the maximum. (This is a very minor issue.)
2007-06-27 13:35:49 +02:00
/*
Keep in sync with pack_keys() in unireg.cc
For each key:
8 bytes for the key header
9 bytes for each key-part (MAX_REF_PARTS)
NAME_LEN bytes for the name
1 byte for the NAMES_SEP_CHAR (before the name)
For all keys:
6 bytes for the header
1 byte for the NAMES_SEP_CHAR (after the last name)
9 extra bytes (padding for safety? alignment?)
*/
Patch for WL#3736: Extended Table, Column and Index Comments. The task is to (a) add a comment on indexes and (b) increase the maximum length of column, table and the new index comments. The patch committed on behalf of Yoshinori Matsunobu (Yoshinori.Matsunobu@Sun.COM).
2010-02-20 11:07:32 +01:00
for (i= 0; i < keys; i++)
{
DBUG_ASSERT(test(key_info[i].flags & HA_USES_COMMENT) ==
(key_info[i].comment.length > 0));
if (key_info[i].flags & HA_USES_COMMENT)
key_comment_total_bytes += 2 + key_info[i].comment.length;
}
key_length= keys * (8 + MAX_REF_PARTS * 9 + NAME_LEN + 1) + 16
+ key_comment_total_bytes;
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
length= next_io_size((ulong) (IO_SIZE+key_length+reclength+
create_info->extra_size));
Import changeset
2000-07-31 21:29:14 +02:00
int4store(fileinfo+10,length);
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
tmp_key_length= (key_length < 0xffff) ? key_length : 0xffff;
int2store(fileinfo+14,tmp_key_length);
Import changeset
2000-07-31 21:29:14 +02:00
int2store(fileinfo+16,reclength);
int4store(fileinfo+18,create_info->max_rows);
int4store(fileinfo+22,create_info->min_rows);
Moved a lot of old bug fixes and safe cleanups from Maria 5.1 tree to 5.1 - Reserver namespace and place in frm for TABLE_CHECKSUM and PAGE_CHECKSUM create options - Added syncing of directory when creating .frm files - Portability fixes - Added missing cast that could cause bugs - Code cleanups - Made some bit functions inline - Moved things out of myisam.h to my_handler.h to make them more accessable - Renamed some myisam variables and defines to make them more globaly usable (as they are used outside of MyISAM) - Fixed bugs in error conditions - Use compiler time asserts instead of run time - Fixed indentation HA_EXTRA_PREPARE_FOR_DELETE -> HA_EXTRA_PREPARE_FOR_DROP as the old name was wrong (Added a define for old value to ensure we don't break any old code) Added HA_EXTRA_PREPARE_FOR_RENAME as a signal for rename (before we used a DROP signal which is wrong) - Initialize error messages early to get better errors when mysqld or an engine fails to start - Fix windows bug that query_performance_frequency was not initialized if registry code failed - thread_stack -> my_thread_stack_size
2007-10-11 17:07:40 +02:00
/* fileinfo[26] is set in mysql_create_frm() */
Import changeset
2000-07-31 21:29:14 +02:00
fileinfo[27]=2; // Use long pack-fields
Moved a lot of old bug fixes and safe cleanups from Maria 5.1 tree to 5.1 - Reserver namespace and place in frm for TABLE_CHECKSUM and PAGE_CHECKSUM create options - Added syncing of directory when creating .frm files - Portability fixes - Added missing cast that could cause bugs - Code cleanups - Made some bit functions inline - Moved things out of myisam.h to my_handler.h to make them more accessable - Renamed some myisam variables and defines to make them more globaly usable (as they are used outside of MyISAM) - Fixed bugs in error conditions - Use compiler time asserts instead of run time - Fixed indentation HA_EXTRA_PREPARE_FOR_DELETE -> HA_EXTRA_PREPARE_FOR_DROP as the old name was wrong (Added a define for old value to ensure we don't break any old code) Added HA_EXTRA_PREPARE_FOR_RENAME as a signal for rename (before we used a DROP signal which is wrong) - Initialize error messages early to get better errors when mysqld or an engine fails to start - Fix windows bug that query_performance_frequency was not initialized if registry code failed - thread_stack -> my_thread_stack_size
2007-10-11 17:07:40 +02:00
/* fileinfo[28 & 29] is set to key_info_length in mysql_create_frm() */
Import changeset
2000-07-31 21:29:14 +02:00
create_info->table_options|=HA_OPTION_LONG_BLOB_PTR; // Use portable blob pointers
int2store(fileinfo+30,create_info->table_options);
fileinfo[32]=0; // No filename anymore
Hack to ensure that CHAR's created in 5.0 are not converted to VARCHAR even if the row type is dynamic (For 4.1 tables old 'VARCHAR' fields are converted to true VARCHAR in the next ALTER TABLE) This ensures that one can use MySQL 5.0 privilege tables with MySQL 4.1
2005-03-22 14:48:06 +01:00
fileinfo[33]=5; // Mark for 5.0 frm file
Import changeset
2000-07-31 21:29:14 +02:00
int4store(fileinfo+34,create_info->avg_row_length);
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
csid= (create_info->default_table_charset ?
create_info->default_table_charset->number : 0);
fileinfo[38]= (uchar) csid;
Bug#45829 "CREATE TABLE TRANSACTIONAL PAGE_CHECKSUM ROW_FORMAT=PAGE accepted, does nothing": those keywords do nothing in 5.1 (they are meant for future versions, for example featuring the Maria engine) so they are here removed from the syntax. Adding those keywords to future versions when needed is: - WL#5034 "Add TRANSACTIONA=0|1 and PAGE_CHECKSUM=0|1 clauses to CREATE TABLE" - WL#5037 "New ROW_FORMAT value for CREATE TABLE: PAGE"
2009-07-29 10:54:20 +02:00
/*
In future versions, we will store in fileinfo[39] the values of the
TRANSACTIONAL and PAGE_CHECKSUM clauses of CREATE TABLE.
*/
fileinfo[39]= 0;
Import changeset
2000-07-31 21:29:14 +02:00
fileinfo[40]= (uchar) create_info->row_type;
Moved a lot of old bug fixes and safe cleanups from Maria 5.1 tree to 5.1 - Reserver namespace and place in frm for TABLE_CHECKSUM and PAGE_CHECKSUM create options - Added syncing of directory when creating .frm files - Portability fixes - Added missing cast that could cause bugs - Code cleanups - Made some bit functions inline - Moved things out of myisam.h to my_handler.h to make them more accessable - Renamed some myisam variables and defines to make them more globaly usable (as they are used outside of MyISAM) - Fixed bugs in error conditions - Use compiler time asserts instead of run time - Fixed indentation HA_EXTRA_PREPARE_FOR_DELETE -> HA_EXTRA_PREPARE_FOR_DROP as the old name was wrong (Added a define for old value to ensure we don't break any old code) Added HA_EXTRA_PREPARE_FOR_RENAME as a signal for rename (before we used a DROP signal which is wrong) - Initialize error messages early to get better errors when mysqld or an engine fails to start - Fix windows bug that query_performance_frequency was not initialized if registry code failed - thread_stack -> my_thread_stack_size
2007-10-11 17:07:40 +02:00
/* Next few bytes where for RAID support */
Backporting WL#4164 Two-byte collation IDs
2009-10-15 12:17:32 +02:00
fileinfo[41]= (uchar) (csid >> 8);
This patch is to further remove the RAID code. We removed support for people creating tables with RAID. This patch remove most of the source for this.
2006-02-12 02:21:01 +01:00
fileinfo[42]= 0;
fileinfo[43]= 0;
fileinfo[44]= 0;
fileinfo[45]= 0;
fileinfo[46]= 0;
Fix that we can read tables with the 'older' decimal format used in 5.0.3 & 5.0.4 We will however give a warning when opening such a table that users should use ALTER TABLE ... FORCE to fix the table. In future release we will fix that REPAIR TABLE will be able to handle this case
2005-05-25 17:33:36 +02:00
int4store(fileinfo+47, key_length);
tmp= MYSQL_VERSION_ID; // Store to avoid warning from int4store
int4store(fileinfo+51, tmp);
WL #2604: Partition Management Optimised version of ADD/DROP/REORGANIZE partitions for non-NDB storage engines. New syntax to handle REBUILD/OPTIMIZE/ANALYZE/CHECK/REPAIR partitions Quite a few bug fixes
2006-01-17 08:40:00 +01:00
int4store(fileinfo+55, create_info->extra_size);
Added support for key_block_size for key and table level (WL#602) Added support for key_block_size to MyISAM. Simplify interface to 'new Key' to make it easier to add new key options. mysqld option --new is used to define where key options are printed. (In 5.3 we should move all key options to after key part definition to avoid problem with reserved names) Fixed some compiler warnings and a memory leak in ssl
2006-05-03 14:59:17 +02:00
/*
59-60 is reserved for extra_rec_buf_length,
61 for default_part_db_type
*/
int2store(fileinfo+62, create_info->key_block_size);
Import changeset
2000-07-31 21:29:14 +02:00
bzero(fill,IO_SIZE);
for (; length > IO_SIZE ; length-= IO_SIZE)
{
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
if (mysql_file_write(file, fill, IO_SIZE, MYF(MY_WME | MY_NABP)))
Import changeset
2000-07-31 21:29:14 +02:00
{
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
(void) mysql_file_close(file, MYF(0));
(void) mysql_file_delete(key_file_frm, name, MYF(0));
Import changeset
2000-07-31 21:29:14 +02:00
return(-1);
}
}
}
Fix error message generated when trying to create a table in a non-existent database. (Bug #10407)
2005-07-22 05:08:54 +02:00
else
{
if (my_errno == ENOENT)
my_error(ER_BAD_DB_ERROR,MYF(0),db);
else
my_error(ER_CANT_CREATE_TABLE,MYF(0),table,my_errno);
}
Import changeset
2000-07-31 21:29:14 +02:00
return (file);
} /* create_frm */
void update_create_info_from_table(HA_CREATE_INFO *create_info, TABLE *table)
{
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
TABLE_SHARE *share= table->s;
Fixed that SHOW CREATE TABLE shows all attributes
2001-09-30 04:47:35 +02:00
DBUG_ENTER("update_create_info_from_table");
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
create_info->max_rows= share->max_rows;
create_info->min_rows= share->min_rows;
create_info->table_options= share->db_create_options;
create_info->avg_row_length= share->avg_row_length;
create_info->row_type= share->row_type;
create_info->default_table_charset= share->table_charset;
CLIENT_MULTI_QUERIES -> CLIENT_MULTI_STATEMENTS New multi-key-cache handling. This was needed becasue the old one didn't work reliable with MERGE tables. ALTER TABLE table_name ... CHARACTER SET ... now changes all char/varchar/text columns to the given character set (One must use ALTER TABLE ... DEFAULT CHARACTER SET ... to change the default character set) Fixed that have_compress is detected properly (fixes problems with func_compress.test on platforms without zlib) New syntax for CACHE INDEX ('keys' is optional if no index name is given and one mentions the key cache name only ones) Removed compiler warnings Added mysql_set_server_option() to allow clients like PHP to easaily set/reset the multi-statement flag.
2003-11-18 12:47:27 +01:00
create_info->table_charset= 0;
Apply patch for Bug#27806 table comments not passed in to storage engine during "CREATE ... LIKE ..." Only affects engine writers. No change in server behaviour.
2007-08-08 12:02:07 +02:00
create_info->comment= share->comment;
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
Fixed that SHOW CREATE TABLE shows all attributes
2001-09-30 04:47:35 +02:00
DBUG_VOID_RETURN;
Update copyright Fixed memory leak on shutdown (Affects the embedded version & MyODBC)
2001-12-06 13:10:51 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
int
rename_file_ext(const char * from,const char * to,const char * ext)
{
char from_b[FN_REFLEN],to_b[FN_REFLEN];
Backport of: ---------------------------------------------------------------------- ChangeSet@1.2571, 2008-04-08 12:30:06+02:00, vvaintroub@wva. +122 -0 Bug#32082 : definition of VOID in my_global.h conflicts with Windows SDK headers VOID macro is now removed. Its usage is replaced with void cast. In some cases, where cast does not make much sense (pthread_*, printf, hash_delete, my_seek), cast is ommited.
2009-11-24 14:54:59 +01:00
(void) strxmov(from_b,from,ext,NullS);
(void) strxmov(to_b,to,ext,NullS);
WL#2360 Performance schema Part IV: sql instrumentation
2010-01-07 06:42:07 +01:00
return (mysql_file_rename(key_file_frm, from_b, to_b, MYF(MY_WME)));
Import changeset
2000-07-31 21:29:14 +02:00
}
new version of help
2003-05-29 23:47:31 +02:00
/*
Allocate string field in MEM_ROOT and return it as String
SYNOPSIS
get_field()
mem MEM_ROOT for allocating
field Field for retrieving of string
res result String
RETURN VALUES
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
1 string is empty
0 all ok
new version of help
2003-05-29 23:47:31 +02:00
*/
bool get_field(MEM_ROOT *mem, Field *field, String *res)
{
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
char buff[MAX_FIELD_WIDTH], *to;
new version of help
2003-05-29 23:47:31 +02:00
String str(buff,sizeof(buff),&my_charset_bin);
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
uint length;
::reset(), HA_FAST_KEY_READ, disable_indexes(), enable_indexes(), start_bulk_insert(), end_bulk_insert() Field::val_str simplification, comment
2004-04-06 21:35:26 +02:00
field->val_str(&str);
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
if (!(length= str.length()))
Code cleanups during review of pushed code
2005-03-15 15:07:28 +01:00
{
res->length(0);
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
return 1;
Code cleanups during review of pushed code
2005-03-15 15:07:28 +01:00
}
if (!(to= strmake_root(mem, str.ptr(), length)))
length= 0; // Safety fix
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
res->set(to, length, ((Field_str*)field)->charset());
return 0;
new version of help
2003-05-29 23:47:31 +02:00
}
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
Import changeset
2000-07-31 21:29:14 +02:00
/*
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
Allocate string field in MEM_ROOT and return it as NULL-terminated string
SYNOPSIS
get_field()
mem MEM_ROOT for allocating
field Field for retrieving of string
RETURN VALUES
NullS string is empty
# pointer to NULL-terminated string value of field
Import changeset
2000-07-31 21:29:14 +02:00
*/
Fixed a lot of wrong memory references as reported by valgrind Portability fixes Added new client function: mysql_get_server_version() New server help code (From Victor Vagin) Fixed wrong usage of binary() Disabled RTREE usage for now.
2003-02-12 20:55:37 +01:00
char *get_field(MEM_ROOT *mem, Field *field)
Import changeset
2000-07-31 21:29:14 +02:00
{
Fixes after merge with 4.0 Cleaned up embedded library access and query cache handling Changed min stack size to 128K (to allow longer MyISAM keys) Fixed wrong priority for XOR (should be less than NEG to get -1^1 to work)
2003-12-19 15:25:50 +01:00
char buff[MAX_FIELD_WIDTH], *to;
item_func.cc, item_cmpfunc.cc, unireg.cc, table.cc, sql_base.cc: some default_charset_info were removed
2003-03-07 10:39:53 +01:00
String str(buff,sizeof(buff),&my_charset_bin);
After merge fixes + bugs from last merge
2003-08-19 15:00:12 +02:00
uint length;
::reset(), HA_FAST_KEY_READ, disable_indexes(), enable_indexes(), start_bulk_insert(), end_bulk_insert() Field::val_str simplification, comment
2004-04-06 21:35:26 +02:00
field->val_str(&str);
Fixes after merge with 4.0 Cleaned up embedded library access and query cache handling Changed min stack size to 128K (to allow longer MyISAM keys) Fixed wrong priority for XOR (should be less than NEG to get -1^1 to work)
2003-12-19 15:25:50 +01:00
length= str.length();
code cleanup after some reasoning
2003-12-09 20:49:48 +01:00
if (!length || !(to= (char*) alloc_root(mem,length+1)))
Import changeset
2000-07-31 21:29:14 +02:00
return NullS;
memcpy(to,str.ptr(),(uint) length);
to[length]=0;
return to;
}
WL#3700: Handler API change: all index search methods - that is, index_read(), index_read_idx(), index_read_last(), and records_in_range() - instead of 'uint keylen' argument take 'ulonglong keypart_map', a bitmap showing which keyparts are present in the key value. Fallback method is provided for handlers that are lagging behind.
2007-01-29 10:40:26 +01:00
/*
DESCRIPTION
given a buffer with a key value, and a map of keyparts
that are present in this value, returns the length of the value
*/
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uint calculate_key_len(TABLE *table, uint key, const uchar *buf,
wl#3700 - post-review fixes: s/ulonglong/key_part_map/, comments
2007-03-17 00:13:25 +01:00
key_part_map keypart_map)
WL#3700: Handler API change: all index search methods - that is, index_read(), index_read_idx(), index_read_last(), and records_in_range() - instead of 'uint keylen' argument take 'ulonglong keypart_map', a bitmap showing which keyparts are present in the key value. Fallback method is provided for handlers that are lagging behind.
2007-01-29 10:40:26 +01:00
{
/* works only with key prefixes */
DBUG_ASSERT(((keypart_map + 1) & keypart_map) == 0);
KEY *key_info= table->s->key_info+key;
KEY_PART_INFO *key_part= key_info->key_part;
KEY_PART_INFO *end_key_part= key_part + key_info->key_parts;
uint length= 0;
while (key_part < end_key_part && keypart_map)
{
length+= key_part->store_length;
keypart_map >>= 1;
key_part++;
}
return length;
}
Fixed handling of lower_case_table_names in SHOW TABLE STATUS, mysql_list_fields() and mysql_table_dump(). This fixes some Errcode 13 errors on Windows when deleting tables.
2003-01-29 17:56:34 +01:00
/*
Check if database name is valid
SYNPOSIS
check_db_name()
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
org_name Name of database and length
Fixed handling of lower_case_table_names in SHOW TABLE STATUS, mysql_list_fields() and mysql_table_dump(). This fixes some Errcode 13 errors on Windows when deleting tables.
2003-01-29 17:56:34 +01:00
NOTES
If lower_case_table_names is set then database is converted to lower case
RETURN
0 ok
1 error
*/
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
bool check_db_name(LEX_STRING *org_name)
Import changeset
2000-07-31 21:29:14 +02:00
{
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
char *name= org_name->str;
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
uint name_length= org_name->length;
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
if (!name_length || name_length > NAME_LEN)
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
return 1;
Fixed handling of lower_case_table_names in SHOW TABLE STATUS, mysql_list_fields() and mysql_table_dump(). This fixes some Errcode 13 errors on Windows when deleting tables.
2003-01-29 17:56:34 +01:00
don't modify constant strings
2004-04-06 12:31:25 +02:00
if (lower_case_table_names && name != any_db)
After merge fixes & remove compiler warnings Added lengths for all MYSQL_FIELD string parameters Changed field length to 2 byte in .frm files
2003-02-07 14:47:24 +01:00
my_casedn_str(files_charset_info, name);
Fixed handling of lower_case_table_names in SHOW TABLE STATUS, mysql_list_fields() and mysql_table_dump(). This fixes some Errcode 13 errors on Windows when deleting tables.
2003-01-29 17:56:34 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
#if defined(USE_MB) && defined(USE_MB_IDENT)
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
if (use_mb(system_charset_info))
{
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
name_length= 0;
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
bool last_char_is_space= TRUE;
char *end= name + org_name->length;
while (name < end)
Import changeset
2000-07-31 21:29:14 +02:00
{
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
int len;
last_char_is_space= my_isspace(system_charset_info, *name);
len= my_ismbchar(system_charset_info, name, end);
if (!len)
len= 1;
name+= len;
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
name_length++;
Import changeset
2000-07-31 21:29:14 +02:00
}
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
return (last_char_is_space || name_length > NAME_CHAR_LEN);
Import changeset
2000-07-31 21:29:14 +02:00
}
Changed several char* to LEX_STRING*.
2006-10-16 18:57:33 +02:00
else
#endif
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
return ((org_name->str[org_name->length - 1] != ' ') ||
(name_length > NAME_CHAR_LEN)); /* purecov: inspected */
Import changeset
2000-07-31 21:29:14 +02:00
}
/*
Allow anything as a table name, as long as it doesn't contain an
WL#1324 table name to file name encoding - Encoding itself, implemented as a charset "filename". Originally planned to use '.' as an escape character, but now changed to '@' for two reasons: "ls" does not return file names starting with '.' considering them as a kind of hidden files; some platforms do not allow several dots in a file name. - replacing many calls of my_snprintf() and strnxmov() to the new build_table_filename(). - Adding MY_APPEND_EXT mysys flag, to append an extention rather that replace it. - Replacing all numeric constants in fn_format flag arguments to their mysys definitions, e.g. MY_UNPACK_FILENAME, - Predictability in several function/methods: when a table name can appear with or withot .frm extension. Some functions/methods were changed so accept names strictly with .frm, other - strictly without .frm extensions. Several DBUG_ASSERTs were added to check whether an extension is passed. Many files: table name to file name encoding mysql_priv.h: Prototypes for new table name encoding tools. ctype-utf8.c: Implementing "filename" charset for table name to file name encoding. row0mysql.c: Fixing table name prefix. mf_format.c: Adding MY_APPEND_EXT processing. Many files: Fixing tests. my_sys.h: Adding new flag to append rather than replace an extension. m_ctype.h: Adding "filename" charset definition.
2005-12-31 06:01:26 +01:00
' ' at the end
Import changeset
2000-07-31 21:29:14 +02:00
returns 1 on error
*/
Bug #53371: COM_FIELD_LIST can be abused to bypass table level grants. This is the 5.1 merge and extension of the fix. The server was happily accepting paths in table name in all places a table name is accepted (e.g. a SELECT). This allowed all users that have some privilege over some database to read all tables in all databases in all mysql server instances that the server file system has access to. Fixed by : 1. making sure no path elements are allowed in quoted table name when constructing the path (note that the path symbols are still valid in table names when they're properly escaped by the server). 2. checking the #mysql50# prefixed names the same way they're checked for path elements in mysql-5.0.
2010-05-04 16:03:28 +02:00
bool check_table_name(const char *name, uint length, bool check_for_path_chars)
Import changeset
2000-07-31 21:29:14 +02:00
{
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
uint name_length= 0; // name length in symbols
Import changeset
2000-07-31 21:29:14 +02:00
const char *end= name+length;
Check for empty table/column names
2003-03-14 16:08:42 +01:00
if (!length || length > NAME_LEN)
return 1;
fixed Bug #2985 "Table truncated when creating another table name with Spaces" added to check_db_name, check_table_name and check_column_name test for end space
2004-03-05 19:13:33 +01:00
#if defined(USE_MB) && defined(USE_MB_IDENT)
extra correction of check_db_name, check_table_name and check_column_name
2004-03-12 14:55:33 +01:00
bool last_char_is_space= FALSE;
fixed Bug #2985 "Table truncated when creating another table name with Spaces" added to check_db_name, check_table_name and check_column_name test for end space
2004-03-05 19:13:33 +01:00
#else
if (name[length-1]==' ')
return 1;
#endif
Import changeset
2000-07-31 21:29:14 +02:00
while (name != end)
{
#if defined(USE_MB) && defined(USE_MB_IDENT)
Bug#18004 Connecting crashes server when default charset is UCS2 table.cc: Fixing to use system_charset_info instead of default_charset_info. Crash happened because the "ctype" array is empty in UCS2, and thus cannot be used with my_isspace(). The reason why UCS2 appeared in this context was because of of default_charset_info variable incorrectly substituted to my_isspace(). As functions check_db_name(), check_table_name() and check_column_name() always get values in utf8, system_charset_info must be used instead. ctype_ucs2_def.test, ctype_ucs2_def-master.opt, ctype_ucs2_def.result: new file
2006-03-20 11:43:02 +01:00
last_char_is_space= my_isspace(system_charset_info, *name);
New ctype functions/macros to support many charsets at a time
2002-03-12 18:37:58 +01:00
if (use_mb(system_charset_info))
Import changeset
2000-07-31 21:29:14 +02:00
{
New ctype functions/macros to support many charsets at a time
2002-03-12 18:37:58 +01:00
int len=my_ismbchar(system_charset_info, name, end);
Import changeset
2000-07-31 21:29:14 +02:00
if (len)
{
name += len;
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
name_length++;
Import changeset
2000-07-31 21:29:14 +02:00
continue;
}
}
Bug #53371: COM_FIELD_LIST can be abused to bypass table level grants. This is the 5.1 merge and extension of the fix. The server was happily accepting paths in table name in all places a table name is accepted (e.g. a SELECT). This allowed all users that have some privilege over some database to read all tables in all databases in all mysql server instances that the server file system has access to. Fixed by : 1. making sure no path elements are allowed in quoted table name when constructing the path (note that the path symbols are still valid in table names when they're properly escaped by the server). 2. checking the #mysql50# prefixed names the same way they're checked for path elements in mysql-5.0.
2010-05-04 16:03:28 +02:00
if (check_for_path_chars &&
(*name == '/' || *name == '\\' || *name == '~' || *name == FN_EXTCHAR))
Import changeset
2000-07-31 21:29:14 +02:00
return 1;
#endif
name++;
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
name_length++;
Import changeset
2000-07-31 21:29:14 +02:00
}
fixed Bug #2985 "Table truncated when creating another table name with Spaces" added to check_db_name, check_table_name and check_column_name test for end space
2004-03-05 19:13:33 +01:00
#if defined(USE_MB) && defined(USE_MB_IDENT)
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
return (last_char_is_space || name_length > NAME_CHAR_LEN) ;
extra correction of check_db_name, check_table_name and check_column_name
2004-03-12 14:55:33 +01:00
#else
Import changeset
2000-07-31 21:29:14 +02:00
return 0;
extra correction of check_db_name, check_table_name and check_column_name
2004-03-12 14:55:33 +01:00
#endif
Import changeset
2000-07-31 21:29:14 +02:00
}
Fixes to newly pushed code
2004-03-12 14:56:28 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
bool check_column_name(const char *name)
{
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
uint name_length= 0; // name length in symbols
merge with 4.0
2004-03-16 21:41:30 +01:00
bool last_char_is_space= TRUE;
extra correction of check_db_name, check_table_name and check_column_name
2004-03-12 14:55:33 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
while (*name)
{
#if defined(USE_MB) && defined(USE_MB_IDENT)
Bug#18004 Connecting crashes server when default charset is UCS2 table.cc: Fixing to use system_charset_info instead of default_charset_info. Crash happened because the "ctype" array is empty in UCS2, and thus cannot be used with my_isspace(). The reason why UCS2 appeared in this context was because of of default_charset_info variable incorrectly substituted to my_isspace(). As functions check_db_name(), check_table_name() and check_column_name() always get values in utf8, system_charset_info must be used instead. ctype_ucs2_def.test, ctype_ucs2_def-master.opt, ctype_ucs2_def.result: new file
2006-03-20 11:43:02 +01:00
last_char_is_space= my_isspace(system_charset_info, *name);
New ctype functions/macros to support many charsets at a time
2002-03-12 18:37:58 +01:00
if (use_mb(system_charset_info))
Import changeset
2000-07-31 21:29:14 +02:00
{
New ctype functions/macros to support many charsets at a time
2002-03-12 18:37:58 +01:00
int len=my_ismbchar(system_charset_info, name,
Add support for up to VARCHAR (size up to 65535) Renamed HA_VAR_LENGTH to HA_VAR_LENGTH_PART Renamed in all files FIELD_TYPE_STRING and FIELD_TYPE_VAR_STRING to MYSQL_TYPE_STRING and MYSQL_TYPE_VAR_STRING to make it easy to catch all possible errors Added support for VARCHAR KEYS to heap Removed support for ISAM Now only long VARCHAR columns are changed to TEXT on demand (not CHAR) Internal temporary files can now use fixed length tables if the used VARCHAR columns are short
2004-12-06 01:00:37 +01:00
name+system_charset_info->mbmaxlen);
Import changeset
2000-07-31 21:29:14 +02:00
if (len)
{
name += len;
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
name_length++;
Import changeset
2000-07-31 21:29:14 +02:00
continue;
}
}
fixed Bug #2985 "Table truncated when creating another table name with Spaces" added to check_db_name, check_table_name and check_column_name test for end space
2004-03-05 19:13:33 +01:00
#else
extra correction of check_db_name, check_table_name and check_column_name
2004-03-12 14:55:33 +01:00
last_char_is_space= *name==' ';
Import changeset
2000-07-31 21:29:14 +02:00
#endif
if (*name == NAMES_SEP_CHAR)
return 1;
name++;
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
name_length++;
Import changeset
2000-07-31 21:29:14 +02:00
}
Check for empty table/column names
2003-03-14 16:08:42 +01:00
/* Error if empty or too long column name */
Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte
2007-04-03 13:13:27 +02:00
return last_char_is_space || (uint) name_length > NAME_CHAR_LEN;
Import changeset
2000-07-31 21:29:14 +02:00
}
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
/**
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
Checks whether a table is intact. Should be done *just* after the table has
been opened.
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
@param[in] table The table to check
@param[in] table_f_count Expected number of columns in the table
@param[in] table_def Expected structure of the table (column name
and type)
@retval FALSE OK
@retval TRUE There was an error. An error message is output
to the error log. We do not push an error
message into the error stack because this
function is currently only called at start up,
and such errors never reach the user.
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
*/
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
bool
Table_check_intact::check(TABLE *table, const TABLE_FIELD_DEF *table_def)
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
{
uint i;
my_bool error= FALSE;
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
const TABLE_FIELD_TYPE *field_def= table_def->field;
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
DBUG_ENTER("table_check_intact");
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
DBUG_PRINT("info",("table: %s expected_count: %d",
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
table->alias, table_def->count));
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
/* Whether the table definition has already been validated. */
if (table->s->table_field_def_cache == table_def)
DBUG_RETURN(FALSE);
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
if (table->s->fields != table_def->count)
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
{
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
DBUG_PRINT("info", ("Column count has changed, checking the definition"));
/* previous MySQL version */
if (MYSQL_VERSION_ID > table->s->mysql_version)
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
{
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(ER_COL_COUNT_DOESNT_MATCH_PLEASE_UPDATE,
ER(ER_COL_COUNT_DOESNT_MATCH_PLEASE_UPDATE),
table->alias, table_def->count, table->s->fields,
table->s->mysql_version, MYSQL_VERSION_ID);
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
DBUG_RETURN(TRUE);
}
else if (MYSQL_VERSION_ID == table->s->mysql_version)
{
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(ER_COL_COUNT_DOESNT_MATCH_CORRUPTED,
ER(ER_COL_COUNT_DOESNT_MATCH_CORRUPTED), table->alias,
table_def->count, table->s->fields);
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
DBUG_RETURN(TRUE);
}
/*
Something has definitely changed, but we're running an older
version of MySQL with new system tables.
Let's check column definitions. If a column was added at
the end of the table, then we don't care much since such change
is backward compatible.
*/
}
char buffer[STRING_BUFFER_USUAL_SIZE];
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
for (i=0 ; i < table_def->count; i++, field_def++)
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
{
String sql_type(buffer, sizeof(buffer), system_charset_info);
sql_type.length(0);
if (i < table->s->fields)
{
Field *field= table->field[i];
reduce error log spamming by old privilege tables. If they're old, report it once, don't whine about every missing column.
2006-05-30 18:18:13 +02:00
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
if (strncmp(field->field_name, field_def->name.str,
field_def->name.length))
Fixed some issues found by valgrind (one testcase, one memory leak and some accesses to not initialized memory)
2006-06-06 01:47:30 +02:00
{
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
/*
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
Name changes are not fatal, we use ordinal numbers to access columns.
Still this can be a sign of a tampered table, output an error
to the error log.
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
*/
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(0, "Incorrect definition of table %s.%s: "
"expected column '%s' at position %d, found '%s'.",
table->s->db.str, table->alias, field_def->name.str, i,
field->field_name);
Fixed some issues found by valgrind (one testcase, one memory leak and some accesses to not initialized memory)
2006-06-06 01:47:30 +02:00
}
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
field->sql_type(sql_type);
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
/*
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
Generally, if column types don't match, then something is
wrong.
However, we only compare column definitions up to the
length of the original definition, since we consider the
following definitions compatible:
1. DATETIME and DATETIM
2. INT(11) and INT(11
3. SET('one', 'two') and SET('one', 'two', 'more')
For SETs or ENUMs, if the same prefix is there it's OK to
add more elements - they will get higher ordinal numbers and
the new table definition is backward compatible with the
original one.
*/
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
if (strncmp(sql_type.c_ptr_safe(), field_def->type.str,
field_def->type.length - 1))
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
{
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(0, "Incorrect definition of table %s.%s: "
"expected column '%s' at position %d to have type "
"%s, found type %s.", table->s->db.str, table->alias,
field_def->name.str, i, field_def->type.str,
sql_type.c_ptr_safe());
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
error= TRUE;
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
}
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
else if (field_def->cset.str && !field->has_charset())
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
{
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(0, "Incorrect definition of table %s.%s: "
"expected the type of column '%s' at position %d "
"to have character set '%s' but the type has no "
"character set.", table->s->db.str, table->alias,
field_def->name.str, i, field_def->cset.str);
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
error= TRUE;
}
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
else if (field_def->cset.str &&
strcmp(field->charset()->csname, field_def->cset.str))
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
{
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(0, "Incorrect definition of table %s.%s: "
"expected the type of column '%s' at position %d "
"to have character set '%s' but found "
"character set '%s'.", table->s->db.str, table->alias,
field_def->name.str, i, field_def->cset.str,
field->charset()->csname);
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
error= TRUE;
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
}
}
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
else
{
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
report_error(0, "Incorrect definition of table %s.%s: "
"expected column '%s' at position %d to have type %s "
" but the column is not found.",
table->s->db.str, table->alias,
field_def->name.str, i, field_def->type.str);
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
error= TRUE;
}
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
}
Bug#41726: upgrade from 5.0 to 5.1.30 crashes if you didn't run mysql_upgrade The problem is that the server could crash when attempting to access a non-conformant proc system table. One such case was a crash when invoking stored procedure related statements on a 5.1 server with a proc system table in the 5.0 format. The solution is to validate the proc system table format before attempts to access it are made. If the table is not in the format that the server expects, a message is written to the error log and the statement that caused the table to be accessed fails.
2009-11-21 12:18:21 +01:00
if (! error)
table->s->table_field_def_cache= table_def;
A set of changes aiming to make the Event Scheduler more user-friendly when there are no up-to-date system tables to support it: - initialize the scheduler before reporting "Ready for connections". This ensures that warnings, if any, are printed before "Ready for connections", and this message is not mangled. - do not abort the scheduler if there are no system tables - check the tables once at start up, remember the status and disable the scheduler if the tables are not up to date. If one attempts to use the scheduler with bad tables, issue an error message. - clean up the behaviour of the module under LOCK TABLES and pre-locking mode - make sure implicit commit of Events DDL works as expected. - add more tests Collateral clean ups in the events code. This patch fixes Bug#23631 Events: SHOW VARIABLES doesn't work when mysql.event is damaged
2007-04-05 13:24:34 +02:00
DBUG_RETURN(error);
- final fixes for bug#16431 (Events: An event which alters itself disappears) - fix for bug#16423 (Events: SHOW CREATE EVENT doesn't work) - this Changeset commits makes CREATE/UPDATE/DELETE EVENT real DDL statements by committing the currently open transaction before they are executed. - this Changeset also fixes a trailing space problem since the very early days of the internal cron - adds sophisticated checking of whether mysql.event was tampered accidentally or with purpose by an user. - adds a lot of inline function documentation - documents everything left uncodumented - INTERVAL_XXXX to XXX in I_S.EVENTS.INTERVAL_FIELD WL#1034 (Internal CRON)
2006-02-14 16:20:48 +01:00
}
A fix and a test case for Bug#6513 "Test Suite: Values inserted by using cursor is interpreted latin1 character and Bug#9819 "Cursors: Mysql Server Crash while fetching from table with 5 million records." A fix for a possible memory leak when fetching into an SP cursor in a long loop. The patch uses a common implementation of cursors in the binary protocol and in stored procedures and implements materialized cursors. For implementation details, see comments in sql_cursor.cc
2005-09-22 00:11:21 +02:00
/*
Create Item_field for each column in the table.
SYNPOSIS
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
TABLE::fill_item_list()
A fix and a test case for Bug#6513 "Test Suite: Values inserted by using cursor is interpreted latin1 character and Bug#9819 "Cursors: Mysql Server Crash while fetching from table with 5 million records." A fix for a possible memory leak when fetching into an SP cursor in a long loop. The patch uses a common implementation of cursors in the binary protocol and in stored procedures and implements materialized cursors. For implementation details, see comments in sql_cursor.cc
2005-09-22 00:11:21 +02:00
item_list a pointer to an empty list used to store items
DESCRIPTION
Create Item_field object for each column in the table and
initialize it with the corresponding Field. New items are
created in the current THD memory root.
RETURN VALUE
0 success
1 out of memory
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
bool TABLE::fill_item_list(List<Item> *item_list) const
A fix and a test case for Bug#6513 "Test Suite: Values inserted by using cursor is interpreted latin1 character and Bug#9819 "Cursors: Mysql Server Crash while fetching from table with 5 million records." A fix for a possible memory leak when fetching into an SP cursor in a long loop. The patch uses a common implementation of cursors in the binary protocol and in stored procedures and implements materialized cursors. For implementation details, see comments in sql_cursor.cc
2005-09-22 00:11:21 +02:00
{
/*
All Item_field's created using a direct pointer to a field
are fixed in Item_field constructor.
*/
for (Field **ptr= field; *ptr; ptr++)
{
Item_field *item= new Item_field(*ptr);
if (!item || item_list->push_back(item))
return TRUE;
}
return FALSE;
}
/*
Reset an existing list of Item_field items to point to the
Fields of this table.
SYNPOSIS
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
TABLE::fill_item_list()
A fix and a test case for Bug#6513 "Test Suite: Values inserted by using cursor is interpreted latin1 character and Bug#9819 "Cursors: Mysql Server Crash while fetching from table with 5 million records." A fix for a possible memory leak when fetching into an SP cursor in a long loop. The patch uses a common implementation of cursors in the binary protocol and in stored procedures and implements materialized cursors. For implementation details, see comments in sql_cursor.cc
2005-09-22 00:11:21 +02:00
item_list a non-empty list with Item_fields
DESCRIPTION
This is a counterpart of fill_item_list used to redirect
Item_fields to the fields of a newly created table.
The caller must ensure that number of items in the item_list
is the same as the number of columns in the table.
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::reset_item_list(List<Item> *item_list) const
A fix and a test case for Bug#6513 "Test Suite: Values inserted by using cursor is interpreted latin1 character and Bug#9819 "Cursors: Mysql Server Crash while fetching from table with 5 million records." A fix for a possible memory leak when fetching into an SP cursor in a long loop. The patch uses a common implementation of cursors in the binary protocol and in stored procedures and implements materialized cursors. For implementation details, see comments in sql_cursor.cc
2005-09-22 00:11:21 +02:00
{
List_iterator_fast<Item> it(*item_list);
for (Field **ptr= field; *ptr; ptr++)
{
Item_field *item_field= (Item_field*) it++;
DBUG_ASSERT(item_field != 0);
item_field->reset_field(*ptr);
}
}
Import changeset
2000-07-31 21:29:14 +02:00
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
/*
calculate md5 of query
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::calc_md5()
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
buffer buffer for md5 writing
*/
Cleanup of new code pushed into 5.0 since last pull Merged the different find_xxxx_table_in_list functions to one + some inline functions
2004-09-03 20:43:04 +02:00
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
void TABLE_LIST::calc_md5(char *buffer)
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Cleanup of new code pushed into 5.0 since last pull Merged the different find_xxxx_table_in_list functions to one + some inline functions
2004-09-03 20:43:04 +02:00
uchar digest[16];
merge of bug 42434 to 5.1-bugteam
2009-03-11 17:13:42 +01:00
MY_MD5_HASH(digest, (uchar *) select_stmt.str, select_stmt.length);
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
sprintf((char *) buffer,
"%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x%02x",
digest[0], digest[1], digest[2], digest[3],
digest[4], digest[5], digest[6], digest[7],
digest[8], digest[9], digest[10], digest[11],
digest[12], digest[13], digest[14], digest[15]);
}
Bug#36086: SELECT * from views don't check column grants This patch also fixes bugs 36963 and 35600. - In many places a view was confused with an anonymous derived table, i.e. access checking was skipped. Fixed by introducing a predicate to tell the difference between named and anonymous derived tables. - When inserting fields for "SELECT * ", there was no distinction between base tables and views, where one should be made. View privileges are checked elsewhere.
2008-09-03 16:45:40 +02:00
/**
@brief Set underlying table for table place holder of view.
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Bug#36086: SELECT * from views don't check column grants This patch also fixes bugs 36963 and 35600. - In many places a view was confused with an anonymous derived table, i.e. access checking was skipped. Fixed by introducing a predicate to tell the difference between named and anonymous derived tables. - When inserting fields for "SELECT * ", there was no distinction between base tables and views, where one should be made. View privileges are checked elsewhere.
2008-09-03 16:45:40 +02:00
@details
Post review changes
2005-02-23 17:51:10 +01:00
Bug#36086: SELECT * from views don't check column grants This patch also fixes bugs 36963 and 35600. - In many places a view was confused with an anonymous derived table, i.e. access checking was skipped. Fixed by introducing a predicate to tell the difference between named and anonymous derived tables. - When inserting fields for "SELECT * ", there was no distinction between base tables and views, where one should be made. View privileges are checked elsewhere.
2008-09-03 16:45:40 +02:00
Replace all views that only use one table with the table itself. This
allows us to treat the view as a simple table and even update it (it is a
kind of optimization).
@note
This optimization is potentially dangerous as it makes views
masquerade as base tables: Views don't have the pointer TABLE_LIST::table
set to non-@c NULL.
We may have the case where a view accesses tables not normally accessible
in the current Security_context (only in the definer's
Security_context). According to the table's GRANT_INFO (TABLE::grant),
access is fulfilled, but this is implicitly meant in the definer's security
context. Hence we must never look at only a TABLE's GRANT_INFO without
looking at the one of the referring TABLE_LIST.
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
*/
Cleanup of new code pushed into 5.0 since last pull Merged the different find_xxxx_table_in_list functions to one + some inline functions
2004-09-03 20:43:04 +02:00
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
void TABLE_LIST::set_underlying_merge()
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Post review changes
2005-02-23 17:51:10 +01:00
TABLE_LIST *tbl;
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if ((tbl= merge_underlying_list))
merge
2004-11-21 19:08:12 +01:00
{
Post review changes
2005-02-23 17:51:10 +01:00
/* This is a view. Process all tables of view */
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
DBUG_ASSERT(view && effective_algorithm == VIEW_ALGORITHM_MERGE);
Post review changes
2005-02-23 17:51:10 +01:00
do
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if (tbl->merge_underlying_list) // This is a view
Post review changes
2005-02-23 17:51:10 +01:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
DBUG_ASSERT(tbl->view &&
tbl->effective_algorithm == VIEW_ALGORITHM_MERGE);
Post review changes
2005-02-23 17:51:10 +01:00
/*
This is the only case where set_ancestor is called on an object
that may not be a view (in which case ancestor is 0)
*/
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
tbl->merge_underlying_list->set_underlying_merge();
Post review changes
2005-02-23 17:51:10 +01:00
}
} while ((tbl= tbl->next_local));
Many files: Fixed bug #8528. Representation for single-table views was made similar to representation for multi-table views. view.test: Added test case for bug #8528. view.result: Added test case for bug #8528. Fixed other test cases.
2005-05-11 01:31:13 +02:00
if (!multitable_view)
Post review changes
2005-02-23 17:51:10 +01:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
table= merge_underlying_list->table;
schema_table= merge_underlying_list->schema_table;
Post review changes
2005-02-23 17:51:10 +01:00
}
merge
2004-11-21 19:08:12 +01:00
}
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
/*
setup fields of placeholder of merged VIEW
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::setup_underlying()
new behaviour of CHECK option build, for mor efficience and more correct: check option build only according most top VIEW CHECK OPTION TYPE (BUG#5993)
2004-10-21 17:10:59 +02:00
thd - thread handler
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
DESCRIPTION
It is:
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
- preparing translation table for view columns
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
If there are underlying view(s) procedure first will be called for them.
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
RETURN
Post review changes
2005-02-23 17:51:10 +01:00
FALSE - OK
TRUE - error
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
*/
Cleanup of new code pushed into 5.0 since last pull Merged the different find_xxxx_table_in_list functions to one + some inline functions
2004-09-03 20:43:04 +02:00
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::setup_underlying(THD *thd)
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
DBUG_ENTER("TABLE_LIST::setup_underlying");
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if (!field_translation && merge_underlying_list)
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
Field_translator *transl;
SELECT_LEX *select= &view->select_lex;
Item *item;
TABLE_LIST *tbl;
List_iterator_fast<Item> it(select->item_list);
uint field_count= 0;
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if (check_stack_overrun(thd, STACK_MIN_SIZE, (uchar*) &field_count))
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
Post review changes
2005-02-23 17:51:10 +01:00
DBUG_RETURN(TRUE);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
for (tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if (tbl->merge_underlying_list &&
tbl->setup_underlying(thd))
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
DBUG_RETURN(TRUE);
}
}
Post review changes
2005-02-23 17:51:10 +01:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
/* Create view fields translation table */
if (!(transl=
Rename: - current_arena to stmt_arena: the thread may have more than one 'current' arenas: one for runtime data, and one for the parsed tree of a statement. Only one of them is active at any moment. - set_item_arena -> set_query_arena, because Item_arena was renamed to Query_arena a while ago - set_n_backup_item_arena -> set_n_backup_active_arena; the active arena is the arena thd->mem_root and thd->free_list are currently pointing at. - restore_backup_item_arena -> restore_active_arena (with the same rationale) - change_arena_if_needed -> activate_stmt_arena_if_needed; this method sets thd->stmt_arena active if it's not done yet.
2005-09-02 15:21:19 +02:00
(Field_translator*)(thd->stmt_arena->
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
alloc(select->item_list.elements *
sizeof(Field_translator)))))
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(TRUE);
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
while ((item= it++))
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
transl[field_count].name= item->name;
transl[field_count++].item= item;
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
field_translation= transl;
field_translation_end= transl + field_count;
/* TODO: use hash for big number of fields */
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
/* full text function moving to current select */
if (view->select_lex.ftfunc_list->elements)
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
Item_func_match *ifm;
SELECT_LEX *current_select= thd->lex->current_select;
List_iterator_fast<Item_func_match>
li(*(view->select_lex.ftfunc_list));
while ((ifm= li++))
current_select->ftfunc_list->push_front(ifm);
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(FALSE);
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Post review changes
2005-02-23 17:51:10 +01:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
/*
Prepare where expression of view
error message interception (Bug#4598) error handling of view initialization fixed (Bug#4599)
2004-07-21 03:26:20 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::prep_where()
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
thd - thread handler
conds - condition of this JOIN
no_where_clause - do not build WHERE or ON outer qwery do not need it
(it is INSERT), we do not need conds if this flag is set
fixed using VIEW fields (BUG#4617)
2004-07-23 08:20:58 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
NOTE: have to be called befor CHECK OPTION preparation, because it makes
fix_fields for view WHERE clause
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
RETURN
FALSE - OK
TRUE - error
*/
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::prep_where(THD *thd, Item **conds,
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
bool no_where_clause)
{
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
DBUG_ENTER("TABLE_LIST::prep_where");
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
if (tbl->view && tbl->prep_where(thd, conds, no_where_clause))
{
DBUG_RETURN(TRUE);
}
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
}
view with WHERE in nested join (BUG#5511)
2004-09-10 21:39:04 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
if (where)
{
if (!where->fixed && where->fix_fields(thd, &where))
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(TRUE);
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
postmerge fix
2004-10-04 12:45:35 +02:00
/*
check that it is not VIEW in which we insert with INSERT SELECT
(in this case we can't add view WHERE condition to main SELECT_LEX)
*/
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
if (!no_where_clause && !where_processed)
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
TABLE_LIST *tbl= this;
Rename: - current_arena to stmt_arena: the thread may have more than one 'current' arenas: one for runtime data, and one for the parsed tree of a statement. Only one of them is active at any moment. - set_item_arena -> set_query_arena, because Item_arena was renamed to Query_arena a while ago - set_n_backup_item_arena -> set_n_backup_active_arena; the active arena is the arena thd->mem_root and thd->free_list are currently pointing at. - restore_backup_item_arena -> restore_active_arena (with the same rationale) - change_arena_if_needed -> activate_stmt_arena_if_needed; this method sets thd->stmt_arena active if it's not done yet.
2005-09-02 15:21:19 +02:00
Query_arena *arena= thd->stmt_arena, backup;
arena= thd->activate_stmt_arena_if_needed(&backup); // For easier test
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
postmerge fix
2004-10-04 12:45:35 +02:00
/* Go up to join tree and try to find left join */
for (; tbl; tbl= tbl->embedding)
CHECK OPTIONs added (WL#1983)
2004-09-03 14:18:40 +02:00
{
postmerge fix
2004-10-04 12:45:35 +02:00
if (tbl->outer_join)
{
/*
Store WHERE condition to ON expression for outer join, because
fixed & added comments (according to Igor's review and a bit more)
2004-10-07 00:45:06 +02:00
we can't use WHERE to correctly execute left joins on VIEWs and
postmerge fix
2004-10-04 12:45:35 +02:00
this expression will not be moved to WHERE condition (i.e. will
be clean correctly for PS/SP)
*/
bug #16813 (WITH CHECK OPTION fails with UPDATE) We use the condition from CHECK OPTION twice handling UPDATE command. First we construnct 'update_cond' AND 'option_cond' condition to select records to be updated, then we check the 'option_cond' for the updated row. The problem is that first 'AND' condition is optimized during the 'select' which can break 'option_cond' structure, so it will be unusable for the sectond use - to check the updated row. Possible soultion is either use copy of the condition in the first use or to make optimization less traumatic for the operands. I picked the first one.
2006-09-29 09:16:07 +02:00
tbl->on_expr= and_conds(tbl->on_expr,
where->copy_andor_structure(thd));
postmerge fix
2004-10-04 12:45:35 +02:00
break;
}
CHECK OPTIONs added (WL#1983)
2004-09-03 14:18:40 +02:00
}
postmerge fix
2004-10-04 12:45:35 +02:00
if (tbl == 0)
bug #16813 (WITH CHECK OPTION fails with UPDATE) We use the condition from CHECK OPTION twice handling UPDATE command. First we construnct 'update_cond' AND 'option_cond' condition to select records to be updated, then we check the 'option_cond' for the updated row. The problem is that first 'AND' condition is optimized during the 'select' which can break 'option_cond' structure, so it will be unusable for the sectond use - to check the updated row. Possible soultion is either use copy of the condition in the first use or to make optimization less traumatic for the operands. I picked the first one.
2006-09-29 09:16:07 +02:00
*conds= and_conds(*conds, where->copy_andor_structure(thd));
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
if (arena)
Rename: - current_arena to stmt_arena: the thread may have more than one 'current' arenas: one for runtime data, and one for the parsed tree of a statement. Only one of them is active at any moment. - set_item_arena -> set_query_arena, because Item_arena was renamed to Query_arena a while ago - set_n_backup_item_arena -> set_n_backup_active_arena; the active arena is the arena thd->mem_root and thd->free_list are currently pointing at. - restore_backup_item_arena -> restore_active_arena (with the same rationale) - change_arena_if_needed -> activate_stmt_arena_if_needed; this method sets thd->stmt_arena active if it's not done yet.
2005-09-02 15:21:19 +02:00
thd->restore_active_arena(arena, &backup);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
where_processed= TRUE;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
}
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(FALSE);
}
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
/*
Merge ON expressions for a view
SYNOPSIS
merge_on_conds()
thd thread handle
table table for the VIEW
is_cascaded TRUE <=> merge ON expressions from underlying views
DESCRIPTION
This function returns the result of ANDing the ON expressions
of the given view and all underlying views. The ON expressions
of the underlying views are added only if is_cascaded is TRUE.
RETURN
Pointer to the built expression if there is any.
Otherwise and in the case of a failure NULL is returned.
*/
static Item *
merge_on_conds(THD *thd, TABLE_LIST *table, bool is_cascaded)
{
DBUG_ENTER("merge_on_conds");
Item *cond= NULL;
DBUG_PRINT("info", ("alias: %s", table->alias));
if (table->on_expr)
cond= table->on_expr->copy_andor_structure(thd);
if (!table->nested_join)
DBUG_RETURN(cond);
List_iterator<TABLE_LIST> li(table->nested_join->join_list);
while (TABLE_LIST *tbl= li++)
{
if (tbl->view && !is_cascaded)
continue;
cond= and_conds(cond, merge_on_conds(thd, tbl, is_cascaded));
}
DBUG_RETURN(cond);
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
/*
Prepare check option expression of table
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::prep_check_option()
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
thd - thread handler
check_opt_type - WITH CHECK OPTION type (VIEW_CHECK_NONE,
VIEW_CHECK_LOCAL, VIEW_CHECK_CASCADED)
we use this parameter instead of direct check of
effective_with_check to change type of underlying
views to VIEW_CHECK_CASCADED if outer view have
such option and prevent processing of underlying
view check options if outer view have just
VIEW_CHECK_LOCAL option.
NOTE
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
This method builds check option condition to use it later on
every call (usual execution or every SP/PS call).
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
This method have to be called after WHERE preparation
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
(TABLE_LIST::prep_where)
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
RETURN
FALSE - OK
TRUE - error
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::prep_check_option(THD *thd, uint8 check_opt_type)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
DBUG_ENTER("TABLE_LIST::prep_check_option");
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
bool is_cascaded= check_opt_type == VIEW_CHECK_CASCADED;
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
{
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
/* see comment of check_opt_type parameter */
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
if (tbl->view && tbl->prep_check_option(thd, (is_cascaded ?
VIEW_CHECK_CASCADED :
VIEW_CHECK_NONE)))
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(TRUE);
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
if (check_opt_type && !check_option_processed)
full text function moving to current select (BUG#4822)
2004-08-25 15:14:42 +02:00
{
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
Query_arena *arena= thd->stmt_arena, backup;
arena= thd->activate_stmt_arena_if_needed(&backup); // For easier test
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
if (where)
{
DBUG_ASSERT(where->fixed);
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
check_option= where->copy_andor_structure(thd);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
}
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
if (is_cascaded)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
if (tbl->check_option)
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
check_option= and_conds(check_option, tbl->check_option);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
}
}
Fixed bug #27827. ON conditions from JOIN expression were ignored at CHECK OPTION check when updating a multi-table view with CHECK OPTION. The st_table_list::prep_check_option function has been modified to to take into account ON conditions at CHECK OPTION check It was also changed to build the check option condition only once for any update used in PS/SP.
2007-05-31 23:15:40 +02:00
check_option= and_conds(check_option,
merge_on_conds(thd, this, is_cascaded));
if (arena)
thd->restore_active_arena(arena, &backup);
check_option_processed= TRUE;
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
}
if (check_option)
{
const char *save_where= thd->where;
thd->where= "check option";
Bug #43414 Parenthesis (and other) warnings compiling MySQL with gcc 4.3.2 Compiling MySQL with gcc 4.3.2 and later produces a number of warnings, many of which are new with the recent compiler versions. This bug will be resolved in more than one patch to limit the size of changesets. This is the second patch, fixing more of the warnings.
2009-06-10 16:04:07 +02:00
if ((!check_option->fixed &&
check_option->fix_fields(thd, &check_option)) ||
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
check_option->check_cols(1))
{
DBUG_RETURN(TRUE);
}
thd->where= save_where;
full text function moving to current select (BUG#4822)
2004-08-25 15:14:42 +02:00
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(FALSE);
}
full text function moving to current select (BUG#4822)
2004-08-25 15:14:42 +02:00
Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statement", part 1. Review fixes. Do not send OK/EOF packets to the client until we reached the end of the current statement. This is a consolidation, to keep the functionality that is shared by all SQL statements in one place in the server. Currently this functionality includes: - close_thread_tables() - log_slow_statement(). After this patch and the subsequent patch for Bug#12713, it shall also include: - ha_autocommit_or_rollback() - net_end_statement() - query_cache_end_of_result(). In future it may also include: - mysql_reset_thd_for_next_command().
2007-12-12 16:21:01 +01:00
/**
Bug#35996: SELECT + SHOW VIEW should be enough to display view definition During SHOW CREATE VIEW there is no reason to 'anonymize' errors that name objects that a user does not have access to. Moreover it was inconsistently implemented. For example base tables being referenced from a view appear to be ok, but not views. The manual on the other hand is clear: If a user has the privileges SELECT and SHOW VIEW, the view definition is available to that user, period. The fix changes the behavior to support the manual.
2009-09-28 13:25:47 +02:00
Hide errors which show view underlying table information.
There are currently two mechanisms at work that handle errors for views,
this one and a more general mechanism based on an Internal_error_handler,
see Show_create_error_handler. The latter handles errors encountered during
execution of SHOW CREATE VIEW, while the machanism using this method is
handles SELECT from views. The two methods should not clash.
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statement", part 1. Review fixes. Do not send OK/EOF packets to the client until we reached the end of the current statement. This is a consolidation, to keep the functionality that is shared by all SQL statements in one place in the server. Currently this functionality includes: - close_thread_tables() - log_slow_statement(). After this patch and the subsequent patch for Bug#12713, it shall also include: - ha_autocommit_or_rollback() - net_end_statement() - query_cache_end_of_result(). In future it may also include: - mysql_reset_thd_for_next_command().
2007-12-12 16:21:01 +01:00
@param[in,out] thd thread handler
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statement", part 1. Review fixes. Do not send OK/EOF packets to the client until we reached the end of the current statement. This is a consolidation, to keep the functionality that is shared by all SQL statements in one place in the server. Currently this functionality includes: - close_thread_tables() - log_slow_statement(). After this patch and the subsequent patch for Bug#12713, it shall also include: - ha_autocommit_or_rollback() - net_end_statement() - query_cache_end_of_result(). In future it may also include: - mysql_reset_thd_for_next_command().
2007-12-12 16:21:01 +01:00
@pre This method can be called only if there is an error.
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
void TABLE_LIST::hide_view_error(THD *thd)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
Bug#47736 killing a select from a view when the view is processing a function, asserts hide_view_error() does not take into account that thread query may be killed. Added a check for thd->killed. Addon: backported bug32140 fix from 6.0
2010-02-05 10:39:46 +01:00
if (thd->killed || thd->get_internal_handler())
Bug#35996: SELECT + SHOW VIEW should be enough to display view definition During SHOW CREATE VIEW there is no reason to 'anonymize' errors that name objects that a user does not have access to. Moreover it was inconsistently implemented. For example base tables being referenced from a view appear to be ok, but not views. The manual on the other hand is clear: If a user has the privileges SELECT and SHOW VIEW, the view definition is available to that user, period. The fix changes the behavior to support the manual.
2009-09-28 13:25:47 +02:00
return;
Fix for bug#7212: information_schema: "Can't find file" errors if storage engine gone(after review)
2004-12-16 14:31:36 +01:00
/* Hide "Unknown column" or "Unknown function" error */
Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statement", part 1. Review fixes. Do not send OK/EOF packets to the client until we reached the end of the current statement. This is a consolidation, to keep the functionality that is shared by all SQL statements in one place in the server. Currently this functionality includes: - close_thread_tables() - log_slow_statement(). After this patch and the subsequent patch for Bug#12713, it shall also include: - ha_autocommit_or_rollback() - net_end_statement() - query_cache_end_of_result(). In future it may also include: - mysql_reset_thd_for_next_command().
2007-12-12 16:21:01 +01:00
DBUG_ASSERT(thd->is_error());
WL#2110 (SIGNAL) WL#2265 (RESIGNAL) Manual merge of SIGNAL and RESIGNAL to mysql-trunk-signal, plus required dependencies.
2009-09-10 11:18:29 +02:00
if (thd->stmt_da->sql_errno() == ER_BAD_FIELD_ERROR ||
thd->stmt_da->sql_errno() == ER_SP_DOES_NOT_EXIST ||
thd->stmt_da->sql_errno() == ER_FUNC_INEXISTENT_NAME_COLLISION ||
thd->stmt_da->sql_errno() == ER_PROCACCESS_DENIED_ERROR ||
thd->stmt_da->sql_errno() == ER_COLUMNACCESS_DENIED_ERROR ||
thd->stmt_da->sql_errno() == ER_TABLEACCESS_DENIED_ERROR ||
thd->stmt_da->sql_errno() == ER_TABLE_NOT_LOCKED ||
thd->stmt_da->sql_errno() == ER_NO_SUCH_TABLE)
error message interception (Bug#4598) error handling of view initialization fixed (Bug#4599)
2004-07-21 03:26:20 +02:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
TABLE_LIST *top= top_table();
Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statement", part 1. Review fixes. Do not send OK/EOF packets to the client until we reached the end of the current statement. This is a consolidation, to keep the functionality that is shared by all SQL statements in one place in the server. Currently this functionality includes: - close_thread_tables() - log_slow_statement(). After this patch and the subsequent patch for Bug#12713, it shall also include: - ha_autocommit_or_rollback() - net_end_statement() - query_cache_end_of_result(). In future it may also include: - mysql_reset_thd_for_next_command().
2007-12-12 16:21:01 +01:00
thd->clear_error();
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
my_error(ER_VIEW_INVALID, MYF(0), top->view_db.str, top->view_name.str);
error message interception (Bug#4598) error handling of view initialization fixed (Bug#4599)
2004-07-21 03:26:20 +02:00
}
WL#2110 (SIGNAL) WL#2265 (RESIGNAL) Manual merge of SIGNAL and RESIGNAL to mysql-trunk-signal, plus required dependencies.
2009-09-10 11:18:29 +02:00
else if (thd->stmt_da->sql_errno() == ER_NO_DEFAULT_FOR_FIELD)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
TABLE_LIST *top= top_table();
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
thd->clear_error();
// TODO: make correct error message
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
my_error(ER_NO_DEFAULT_FOR_VIEW_FIELD, MYF(0),
top->view_db.str, top->view_name.str);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
- stackoverflow check added for view of view processing - fixed bug in join view processing - postreview fixes (BUG#9398 & BUG#8703)
2005-04-03 00:23:45 +02:00
/*
Find underlying base tables (TABLE_LIST) which represent given
table_to_find (TABLE)
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::find_underlying_table()
- stackoverflow check added for view of view processing - fixed bug in join view processing - postreview fixes (BUG#9398 & BUG#8703)
2005-04-03 00:23:45 +02:00
table_to_find table to find
RETURN
0 table is not found
found table reference
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST *TABLE_LIST::find_underlying_table(TABLE *table_to_find)
- stackoverflow check added for view of view processing - fixed bug in join view processing - postreview fixes (BUG#9398 & BUG#8703)
2005-04-03 00:23:45 +02:00
{
/* is this real table and table which we are looking for? */
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if (table == table_to_find && merge_underlying_list == 0)
- stackoverflow check added for view of view processing - fixed bug in join view processing - postreview fixes (BUG#9398 & BUG#8703)
2005-04-03 00:23:45 +02:00
return this;
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
- stackoverflow check added for view of view processing - fixed bug in join view processing - postreview fixes (BUG#9398 & BUG#8703)
2005-04-03 00:23:45 +02:00
{
TABLE_LIST *result;
if ((result= tbl->find_underlying_table(table_to_find)))
return result;
}
return 0;
}
postreview fixes
2004-11-08 00:54:23 +01:00
/*
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
cleanup items belonged to view fields translation table
postreview fixes
2004-11-08 00:54:23 +01:00
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::cleanup_items()
postreview fixes
2004-11-08 00:54:23 +01:00
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
void TABLE_LIST::cleanup_items()
postreview fixes
2004-11-08 00:54:23 +01:00
{
if (!field_translation)
return;
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
for (Field_translator *transl= field_translation;
transl < field_translation_end;
transl++)
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
transl->item->walk(&Item::cleanup_processor, 0, 0);
postreview fixes
2004-11-08 00:54:23 +01:00
}
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
/*
check CHECK OPTION condition
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::view_check_option()
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
ignore_failure ignore check option fail
RETURN
VIEW_CHECK_OK OK
VIEW_CHECK_ERROR FAILED
VIEW_CHECK_SKIP FAILED, but continue
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
int TABLE_LIST::view_check_option(THD *thd, bool ignore_failure)
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
{
if (check_option && check_option->val_int() == 0)
{
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
TABLE_LIST *main_view= top_table();
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
if (ignore_failure)
{
WL#2110 (SIGNAL) WL#2265 (RESIGNAL) Manual merge of SIGNAL and RESIGNAL to mysql-trunk-signal, plus required dependencies.
2009-09-10 11:18:29 +02:00
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_WARN,
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
ER_VIEW_CHECK_FAILED, ER(ER_VIEW_CHECK_FAILED),
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
main_view->view_db.str, main_view->view_name.str);
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
return(VIEW_CHECK_SKIP);
}
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
my_error(ER_VIEW_CHECK_FAILED, MYF(0), main_view->view_db.str,
main_view->view_name.str);
return(VIEW_CHECK_ERROR);
post review changes: CHECK OPTION moved to one function view name added to error messages
2004-09-29 15:35:01 +02:00
}
return(VIEW_CHECK_OK);
}
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
/*
spelling fixed/comments added (postreview fixes)
2004-11-25 08:28:32 +01:00
Find table in underlying tables by mask and check that only this
post merge
2004-11-21 18:33:49 +01:00
table belong to given mask
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::check_single_table()
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
table_arg reference on variable where to store found table
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
(should be 0 on call, to find table, or point to table for
unique test)
map bit mask of tables
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
view_arg view for which we are looking table
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
RETURN
post merge
2004-11-21 18:33:49 +01:00
FALSE table not found or found only one
TRUE found several tables
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::check_single_table(TABLE_LIST **table_arg,
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
table_map map,
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST *view_arg)
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
{
if (tbl->table)
{
if (tbl->table->map & map)
{
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
if (*table_arg)
post merge
2004-11-21 18:33:49 +01:00
return TRUE;
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
*table_arg= tbl;
tbl->check_option= view_arg->check_option;
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
}
}
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
else if (tbl->check_single_table(table_arg, map, view_arg))
Code cleanups during code reviews Ensure we get error if INSERT IGNORE ... SELECT fails Fixed wrong key_part->key_length usage in index_merge
2005-06-01 15:35:09 +02:00
return TRUE;
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
}
post merge
2004-11-21 18:33:49 +01:00
return FALSE;
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
}
/*
Set insert_values buffer
SYNOPSIS
set_insert_values()
mem_root memory pool for allocating
RETURN
FALSE - OK
TRUE - out of memory
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::set_insert_values(MEM_ROOT *mem_root)
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
{
if (table)
{
if (!table->insert_values &&
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
!(table->insert_values= (uchar *)alloc_root(mem_root,
First stage of table definition cache Split TABLE to TABLE and TABLE_SHARE (TABLE_SHARE is still allocated as part of table, will be fixed soon) Created Field::make_field() and made Field_num::make_field() to call this Added 'TABLE_SHARE->db' that points to database name; Changed all usage of table_cache_key as database name to use this instead Changed field->table_name to point to pointer to alias. This allows us to change alias for a table by just updating one pointer. Renamed TABLE_SHARE->real_name to table_name Renamed TABLE->table_name to alias Renamed TABLE_LIST->real_name to table_name
2005-01-06 12:00:13 +01:00
table->s->rec_buff_length)))
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
return TRUE;
}
else
{
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
DBUG_ASSERT(view && merge_underlying_list);
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
support of join view updateability (WL#1809)
2004-09-15 22:42:56 +02:00
if (tbl->set_insert_values(mem_root))
return TRUE;
}
return FALSE;
}
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
/*
Test if this is a leaf with respect to name resolution.
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::is_leaf_for_name_resolution()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
DESCRIPTION
A table reference is a leaf with respect to name resolution if
it is either a leaf node in a nested join tree (table, view,
schema table, subquery), or an inner node that represents a
NATURAL/USING join, or a nested join with materialized join
columns.
RETURN
TRUE if a leaf, FALSE otherwise.
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::is_leaf_for_name_resolution()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
return (view || is_natural_join || is_join_columns_complete ||
!nested_join);
}
/*
Retrieve the first (left-most) leaf in a nested join tree with
respect to name resolution.
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::first_leaf_for_name_resolution()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
DESCRIPTION
Given that 'this' is a nested table reference, recursively walk
down the left-most children of 'this' until we reach a leaf
table reference with respect to name resolution.
IMPLEMENTATION
The left-most child of a nested table reference is the last element
in the list of children because the children are inserted in
reverse order.
RETURN
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
If 'this' is a nested table reference - the left-most child of
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
the tree rooted in 'this',
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
else return 'this'
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST *TABLE_LIST::first_leaf_for_name_resolution()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
TABLE_LIST *cur_table_ref;
NESTED_JOIN *cur_nested_join;
LINT_INIT(cur_table_ref);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
if (is_leaf_for_name_resolution())
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
return this;
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
DBUG_ASSERT(nested_join);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
for (cur_nested_join= nested_join;
cur_nested_join;
cur_nested_join= cur_table_ref->nested_join)
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
List_iterator_fast<TABLE_LIST> it(cur_nested_join->join_list);
cur_table_ref= it++;
/*
Fix for BUG#13597 - columns in ON condition not resolved if references a table in a nested right join. The problem was in that when finding the last table reference in a nested join tree, the procedure doing the iteration over the right-most branches of a join tree was testing for RIGHT JOINs the table reference that represents the join, and not the second operand of the JOIN. Currently the information whether a join is LEFT/RIGHT is stored not on the join object itself, but on one of its operands.
2005-09-30 09:39:17 +02:00
If the current nested join is a RIGHT JOIN, the operands in
'join_list' are in reverse order, thus the first operand is
already at the front of the list. Otherwise the first operand
is in the end of the list of join operands.
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
*/
if (!(cur_table_ref->outer_join & JOIN_TYPE_RIGHT))
{
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
TABLE_LIST *next;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
while ((next= it++))
cur_table_ref= next;
}
if (cur_table_ref->is_leaf_for_name_resolution())
break;
}
return cur_table_ref;
}
/*
Retrieve the last (right-most) leaf in a nested join tree with
respect to name resolution.
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::last_leaf_for_name_resolution()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
DESCRIPTION
Given that 'this' is a nested table reference, recursively walk
down the right-most children of 'this' until we reach a leaf
table reference with respect to name resolution.
IMPLEMENTATION
The right-most child of a nested table reference is the first
element in the list of children because the children are inserted
in reverse order.
RETURN
- If 'this' is a nested table reference - the right-most child of
the tree rooted in 'this',
- else - 'this'
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST *TABLE_LIST::last_leaf_for_name_resolution()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
TABLE_LIST *cur_table_ref= this;
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
NESTED_JOIN *cur_nested_join;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
if (is_leaf_for_name_resolution())
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
return this;
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
DBUG_ASSERT(nested_join);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
for (cur_nested_join= nested_join;
cur_nested_join;
cur_nested_join= cur_table_ref->nested_join)
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
Fix for BUG#13597 - columns in ON condition not resolved if references a table in a nested right join. The problem was in that when finding the last table reference in a nested join tree, the procedure doing the iteration over the right-most branches of a join tree was testing for RIGHT JOINs the table reference that represents the join, and not the second operand of the JOIN. Currently the information whether a join is LEFT/RIGHT is stored not on the join object itself, but on one of its operands.
2005-09-30 09:39:17 +02:00
cur_table_ref= cur_nested_join->join_list.head();
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
/*
Fix for BUG#13597 - columns in ON condition not resolved if references a table in a nested right join. The problem was in that when finding the last table reference in a nested join tree, the procedure doing the iteration over the right-most branches of a join tree was testing for RIGHT JOINs the table reference that represents the join, and not the second operand of the JOIN. Currently the information whether a join is LEFT/RIGHT is stored not on the join object itself, but on one of its operands.
2005-09-30 09:39:17 +02:00
If the current nested is a RIGHT JOIN, the operands in
'join_list' are in reverse order, thus the last operand is in the
end of the list.
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
*/
if ((cur_table_ref->outer_join & JOIN_TYPE_RIGHT))
{
List_iterator_fast<TABLE_LIST> it(cur_nested_join->join_list);
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
TABLE_LIST *next;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
cur_table_ref= it++;
while ((next= it++))
cur_table_ref= next;
}
if (cur_table_ref->is_leaf_for_name_resolution())
break;
}
return cur_table_ref;
}
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
/*
Register access mode which we need for underlying tables
SYNOPSIS
register_want_access()
want_access Acess which we require
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
void TABLE_LIST::register_want_access(ulong want_access)
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
{
/* Remove SHOW_VIEW_ACL, because it will be checked during making view */
want_access&= ~SHOW_VIEW_ACL;
if (belong_to_view)
{
grant.want_privilege= want_access;
if (table)
table->grant.want_privilege= want_access;
}
for (TABLE_LIST *tbl= merge_underlying_list; tbl; tbl= tbl->next_local)
tbl->register_want_access(want_access);
}
/*
fixe of typos, comments & layout
2005-10-27 23:56:44 +02:00
Load security context information for this view
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::prepare_view_securety_context()
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
thd [in] thread handler
RETURN
FALSE OK
TRUE Error
*/
#ifndef NO_EMBEDDED_ACCESS_CHECKS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::prepare_view_securety_context(THD *thd)
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
{
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
DBUG_ENTER("TABLE_LIST::prepare_view_securety_context");
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
DBUG_PRINT("enter", ("table: %s", alias));
DBUG_ASSERT(!prelocking_placeholder && view);
if (view_suid)
{
DBUG_PRINT("info", ("This table is suid view => load contest"));
DBUG_ASSERT(view && view_sctx);
if (acl_getroot_no_password(view_sctx,
definer.user.str,
definer.host.str,
definer.host.str,
thd->db))
{
Bug #26817: mysqldump fails to backup database containing view with invalid definer give some leeway on required permissions for SHOW FIELDS on views so an unknonwn DEFINER will no longer break mysqldump
2007-03-23 19:24:03 +01:00
if ((thd->lex->sql_command == SQLCOM_SHOW_CREATE) ||
(thd->lex->sql_command == SQLCOM_SHOW_FIELDS))
Bug #14875: Bad view DEFINER makes SHOW CREATE VIEW fail When reading a view definition from a .frm file it was throwing a SQL error if the DEFINER user is not defined. Changed it to a warning to match the (documented) case when a view with undefined DEFINER user is created.
2006-05-26 10:49:39 +02:00
{
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_NOTE,
ER_NO_SUCH_USER,
ER(ER_NO_SUCH_USER),
definer.user.str, definer.host.str);
}
else
{
Bug#17254: Error for DEFINER security on VIEW provides too much info If a view was created with the DEFINER security and later the definer user was dropped then a SELECT from the view throws the error message saying that there is no definer user is registered. This is ok for a root but too much for a mere user. Now the st_table_list::prepare_view_securety_context() function reveals the absence of the definer only to a superuser and throws the 'access denied' error to others.
2006-11-27 14:15:32 +01:00
if (thd->security_ctx->master_access & SUPER_ACL)
{
my_error(ER_NO_SUCH_USER, MYF(0), definer.user.str, definer.host.str);
}
else
{
my_error(ER_ACCESS_DENIED_ERROR, MYF(0),
thd->security_ctx->priv_user,
thd->security_ctx->priv_host,
(thd->password ? ER(ER_YES) : ER(ER_NO)));
}
Bug #14875: Bad view DEFINER makes SHOW CREATE VIEW fail When reading a view definition from a .frm file it was throwing a SQL error if the DEFINER user is not defined. Changed it to a warning to match the (documented) case when a view with undefined DEFINER user is created.
2006-05-26 10:49:39 +02:00
DBUG_RETURN(TRUE);
}
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
}
}
DBUG_RETURN(FALSE);
}
#endif
/*
Find security context of current view
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::find_view_security_context()
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
thd [in] thread handler
*/
#ifndef NO_EMBEDDED_ACCESS_CHECKS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
Security_context *TABLE_LIST::find_view_security_context(THD *thd)
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
{
Security_context *sctx;
TABLE_LIST *upper_view= this;
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
DBUG_ENTER("TABLE_LIST::find_view_security_context");
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
DBUG_ASSERT(view);
while (upper_view && !upper_view->view_suid)
{
DBUG_ASSERT(!upper_view->prelocking_placeholder);
upper_view= upper_view->referencing_view;
}
if (upper_view)
{
DBUG_PRINT("info", ("Securety context of view %s will be used",
upper_view->alias));
sctx= upper_view->view_sctx;
DBUG_ASSERT(sctx);
}
else
{
DBUG_PRINT("info", ("Current global context will be used"));
sctx= thd->security_ctx;
}
DBUG_RETURN(sctx);
}
#endif
/*
Prepare security context and load underlying tables priveleges for view
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::prepare_security()
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
thd [in] thread handler
RETURN
FALSE OK
TRUE Error
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
bool TABLE_LIST::prepare_security(THD *thd)
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
{
List_iterator_fast<TABLE_LIST> tb(*view_tables);
TABLE_LIST *tbl;
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
DBUG_ENTER("TABLE_LIST::prepare_security");
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
#ifndef NO_EMBEDDED_ACCESS_CHECKS
Security_context *save_security_ctx= thd->security_ctx;
DBUG_ASSERT(!prelocking_placeholder);
if (prepare_view_securety_context(thd))
DBUG_RETURN(TRUE);
thd->security_ctx= find_view_security_context(thd);
while ((tbl= tb++))
{
DBUG_ASSERT(tbl->referencing_view);
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
char *local_db, *local_table_name;
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if (tbl->view)
{
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
local_db= tbl->view_db.str;
local_table_name= tbl->view_name.str;
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
}
else
{
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
local_db= tbl->db;
local_table_name= tbl->table_name;
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
}
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
fill_effective_table_privileges(thd, &tbl->grant, local_db,
local_table_name);
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
if (tbl->table)
tbl->table->grant= grant;
}
thd->security_ctx= save_security_ctx;
#else
while ((tbl= tb++))
tbl->grant.privilege= ~NO_ACCESS;
#endif
found problem removed
2005-11-02 14:17:57 +01:00
DBUG_RETURN(FALSE);
support of view underlying tables and SP functions security check added (BUG#9505) (WL#2787)
2005-10-27 23:18:23 +02:00
}
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
Natural_join_column::Natural_join_column(Field_translator *field_param,
TABLE_LIST *tab)
{
DBUG_ASSERT(tab->field_translation);
view_field= field_param;
table_field= NULL;
table_ref= tab;
is_common= FALSE;
}
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
Natural_join_column::Natural_join_column(Item_field *field_param,
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
TABLE_LIST *tab)
{
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
DBUG_ASSERT(tab->table == field_param->field->table);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
table_field= field_param;
view_field= NULL;
table_ref= tab;
is_common= FALSE;
}
const char *Natural_join_column::name()
{
if (view_field)
{
DBUG_ASSERT(table_field == NULL);
return view_field->name;
}
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return table_field->field_name;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
Item *Natural_join_column::create_item(THD *thd)
{
if (view_field)
{
DBUG_ASSERT(table_field == NULL);
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return create_view_field(thd, table_ref, &view_field->item,
view_field->name);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
return table_field;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
Field *Natural_join_column::field()
{
if (view_field)
{
DBUG_ASSERT(table_field == NULL);
return NULL;
}
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
return table_field->field;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
const char *Natural_join_column::table_name()
{
BUG#25575: ERROR 1052 (Column in from clause is ambiguous) with sub-join Two problems here: Problem 1: While constructing the join columns list the optimizer does as follows: 1. Sets the join_using_fields/natural_join members of the right JOIN operand. 2. Makes a "table reference" (TABLE_LIST) to parent the two tables. 3. Assigns the join_using_fields/is_natural_join of the wrapper table using join_using_fields/natural_join of the rightmost table 4. Sets join_using_fields to NULL for the right JOIN operand. 5. Passes the parent table up to the same procedure on the upper level. Step 1 overrides the the join_using_fields that are set for a nested join wrapping table in step 4. Fixed by making a designated variable SELECT_LEX::prev_join_using to pass the data from step 1 to step 4 without destroying the wrapping table data. Problem 2: The optimizer checks for ambiguous columns while transforming NATURAL JOIN/JOIN USING to JOIN ON. While doing that there was no distinction between columns that are used in the generated join condition (where ambiguity can be checked) and the other columns (where ambiguity can be checked only when resolving references coming from outside the JOIN construct itself). Fixed by allowing the non-USING columns to be present in multiple copies in both sides of the join and moving the ambiguity check to the place where unqualified references to the join columns are resolved (find_field_in_natural_join()).
2007-01-31 15:04:38 +01:00
DBUG_ASSERT(table_ref);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
return table_ref->alias;
}
const char *Natural_join_column::db_name()
{
if (view_field)
return table_ref->view_db.str;
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
/*
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
Test that TABLE_LIST::db is the same as TABLE_SHARE::db to
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
ensure consistency. An exception are I_S schema tables, which
are inconsistent in this respect.
*/
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
DBUG_ASSERT(!strcmp(table_ref->db,
Merge mysql.com:/extern/mysql/5.0/bug17476/mysql-5.0 into mysql.com:/extern/mysql/5.1/generic/mysql-5.1-new
2006-03-03 12:03:27 +01:00
table_ref->table->s->db.str) ||
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
(table_ref->schema_table &&
Manual merge from mysql-trunk-merge. Conflicts: - sql/event_data_objects.cc - sql/events.cc - sql/mysql_priv.h - sql/repl_failsafe.cc - sql/sql_parse.cc - sql/sql_show.cc - sql/sql_view.cc - sql/table.cc
2010-01-27 22:10:37 +01:00
is_infoschema_db(table_ref->table->s->db.str,
table_ref->table->s->db.length)));
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return table_ref->db;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
GRANT_INFO *Natural_join_column::grant()
{
if (view_field)
return &(table_ref->grant);
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return &(table_ref->table->grant);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
void Field_iterator_view::set(TABLE_LIST *table)
{
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
DBUG_ASSERT(table->field_translation);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
view= table;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
ptr= table->field_translation;
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
array_end= table->field_translation_end;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
const char *Field_iterator_table::name()
{
return (*ptr)->field_name;
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
Item *Field_iterator_table::create_item(THD *thd)
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
{
Bug#27874: Non-grouped columns are allowed by * in ONLY_FULL_GROUP_BY mode. When fields are inserted instead of * in the select list they were not marked for check for the ONLY_FULL_GROUP_BY mode. The Field_iterator_table::create_item() function now marks newly created items for check when in the ONLY_FULL_GROUP_BY mode. The setup_wild() and the insert_fields() functions now maintain the cur_pos_in_select_list counter for the ONLY_FULL_GROUP_BY mode.
2007-04-24 20:35:57 +02:00
SELECT_LEX *select= thd->lex->current_select;
Item_field *item= new Item_field(thd, &select->context, *ptr);
if (item && thd->variables.sql_mode & MODE_ONLY_FULL_GROUP_BY &&
!thd->lex->in_sum_func && select->cur_pos_in_select_list != UNDEF_POS)
{
select->non_agg_fields.push_back(item);
item->marker= select->cur_pos_in_select_list;
}
return item;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
const char *Field_iterator_view::name()
{
fixed merged view fields names (BUG#5147) support of merged VIEW over several tables added (WL#1809)
2004-09-14 18:28:29 +02:00
return ptr->name;
VIEW two TABLE_LIST copy eliminated
2004-07-16 00:15:55 +02:00
}
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
Item *Field_iterator_view::create_item(THD *thd)
{
return create_view_field(thd, view, &ptr->item, ptr->name);
}
Item *create_view_field(THD *thd, TABLE_LIST *view, Item **field_ref,
const char *name)
{
bool save_wrapper= thd->lex->select_lex.no_wrap_view_item;
Item *field= *field_ref;
DBUG_ENTER("create_view_field");
if (view->schema_table_reformed)
{
/*
WL#2486 - Natural/using join according to SQL:2003. Post-review fixes according to Monty's review.
2005-11-28 20:57:50 +01:00
Translation table items are always Item_fields and already fixed
('mysql_schema_table' function). So we can return directly the
field. This case happens only for 'show & where' commands.
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
*/
DBUG_ASSERT(field && field->fixed);
DBUG_RETURN(field);
}
DBUG_ASSERT(field);
thd->lex->current_select->no_wrap_view_item= TRUE;
if (!field->fixed)
{
if (field->fix_fields(thd, field_ref))
{
thd->lex->current_select->no_wrap_view_item= save_wrapper;
DBUG_RETURN(0);
}
field= *field_ref;
}
thd->lex->current_select->no_wrap_view_item= save_wrapper;
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
if (save_wrapper)
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
{
DBUG_RETURN(field);
}
Bug#41788 mysql_fetch_field returns org_table == table by a view The problem is that Item_direct_view_ref which is inherited from Item_ident updates orig_table_name and table_name with the same values. The fix is introduction of new constructor into Item_ident and up which updates orig_table_name and table_name separately.
2010-03-12 07:33:16 +01:00
Item *item= new Item_direct_view_ref(view, field_ref, name);
Name resolution context added (BUG#6443)
2005-07-01 06:05:42 +02:00
DBUG_RETURN(item);
}
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
void Field_iterator_natural_join::set(TABLE_LIST *table_ref)
{
DBUG_ASSERT(table_ref->join_columns);
WL#2486 - Natural/using join according to SQL:2003. Post-review fixes according to Monty's review.
2005-11-28 20:57:50 +01:00
column_ref_it.init(*(table_ref->join_columns));
cur_column_ref= column_ref_it++;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
WL#2486 - natural and using join according to SQL:2003 - Corrected problem with N-way nested natural joins in PS mode. - Code cleanup - More asserts to check consistency of name resolution contexts - Fixed potential memory leak of name resolution contexts
2005-08-19 14:22:30 +02:00
void Field_iterator_natural_join::next()
{
WL#2486 - Natural/using join according to SQL:2003. Post-review fixes according to Monty's review.
2005-11-28 20:57:50 +01:00
cur_column_ref= column_ref_it++;
Cleanup during review of new pushed code
2005-08-22 00:13:37 +02:00
DBUG_ASSERT(!cur_column_ref || ! cur_column_ref->table_field ||
cur_column_ref->table_ref->table ==
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
cur_column_ref->table_field->field->table);
WL#2486 - natural and using join according to SQL:2003 - Corrected problem with N-way nested natural joins in PS mode. - Code cleanup - More asserts to check consistency of name resolution contexts - Fixed potential memory leak of name resolution contexts
2005-08-19 14:22:30 +02:00
}
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
void Field_iterator_table_ref::set_field_iterator()
{
DBUG_ENTER("Field_iterator_table_ref::set_field_iterator");
/*
If the table reference we are iterating over is a natural join, or it is
an operand of a natural join, and TABLE_LIST::join_columns contains all
the columns of the join operand, then we pick the columns from
TABLE_LIST::join_columns, instead of the orginial container of the
columns of the join operator.
*/
if (table_ref->is_join_columns_complete)
{
/* Necesary, but insufficient conditions. */
DBUG_ASSERT(table_ref->is_natural_join ||
table_ref->nested_join ||
Bug #43414 Parenthesis (and other) warnings compiling MySQL with gcc 4.3.2 Compiling MySQL with gcc 4.3.2 and later produces a number of warnings, many of which are new with the recent compiler versions. This bug will be resolved in more than one patch to limit the size of changesets. This is the second patch, fixing more of the warnings.
2009-06-10 16:04:07 +02:00
(table_ref->join_columns &&
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
/* This is a merge view. */
((table_ref->field_translation &&
table_ref->join_columns->elements ==
(ulong)(table_ref->field_translation_end -
table_ref->field_translation)) ||
/* This is stored table or a tmptable view. */
(!table_ref->field_translation &&
table_ref->join_columns->elements ==
Bug #43414 Parenthesis (and other) warnings compiling MySQL with gcc 4.3.2 Compiling MySQL with gcc 4.3.2 and later produces a number of warnings, many of which are new with the recent compiler versions. This bug will be resolved in more than one patch to limit the size of changesets. This is the second patch, fixing more of the warnings.
2009-06-10 16:04:07 +02:00
table_ref->table->s->fields))));
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
field_it= &natural_join_it;
DBUG_PRINT("info",("field_it for '%s' is Field_iterator_natural_join",
WL#2486 - natural and using join Fixed debug printout.
2005-08-18 10:28:50 +02:00
table_ref->alias));
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
/* This is a merge view, so use field_translation. */
else if (table_ref->field_translation)
{
DBUG_ASSERT(table_ref->view &&
table_ref->effective_algorithm == VIEW_ALGORITHM_MERGE);
field_it= &view_field_it;
DBUG_PRINT("info", ("field_it for '%s' is Field_iterator_view",
WL#2486 - natural and using join Fixed debug printout.
2005-08-18 10:28:50 +02:00
table_ref->alias));
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
/* This is a base table or stored view. */
else
{
DBUG_ASSERT(table_ref->table || table_ref->view);
field_it= &table_field_it;
DBUG_PRINT("info", ("field_it for '%s' is Field_iterator_table",
WL#2486 - natural and using join Fixed debug printout.
2005-08-18 10:28:50 +02:00
table_ref->alias));
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
field_it->set(table_ref);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
DBUG_VOID_RETURN;
}
void Field_iterator_table_ref::set(TABLE_LIST *table)
{
DBUG_ASSERT(table);
first_leaf= table->first_leaf_for_name_resolution();
last_leaf= table->last_leaf_for_name_resolution();
DBUG_ASSERT(first_leaf && last_leaf);
table_ref= first_leaf;
set_field_iterator();
}
void Field_iterator_table_ref::next()
{
/* Move to the next field in the current table reference. */
field_it->next();
/*
If all fields of the current table reference are exhausted, move to
the next leaf table reference.
*/
if (field_it->end_of_fields() && table_ref != last_leaf)
{
table_ref= table_ref->next_name_resolution_table;
DBUG_ASSERT(table_ref);
set_field_iterator();
}
}
Bug#36086: SELECT * from views don't check column grants This patch also fixes bugs 36963 and 35600. - In many places a view was confused with an anonymous derived table, i.e. access checking was skipped. Fixed by introducing a predicate to tell the difference between named and anonymous derived tables. - When inserting fields for "SELECT * ", there was no distinction between base tables and views, where one should be made. View privileges are checked elsewhere.
2008-09-03 16:45:40 +02:00
const char *Field_iterator_table_ref::get_table_name()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
if (table_ref->view)
return table_ref->view_name.str;
else if (table_ref->is_natural_join)
return natural_join_it.column_ref()->table_name();
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
DBUG_ASSERT(!strcmp(table_ref->table_name,
Table definition cache, part 2 The table opening process now works the following way: - Create common TABLE_SHARE object - Read the .frm file and unpack it into the TABLE_SHARE object - Create a TABLE object based on the information in the TABLE_SHARE object and open a handler to the table object Other noteworthy changes: - In TABLE_SHARE the most common strings are now LEX_STRING's - Better error message when table is not found - Variable table_cache is now renamed 'table_open_cache' - New variable 'table_definition_cache' that is the number of table defintions that will be cached - strxnmov() calls are now fixed to avoid overflows - strxnmov() will now always add one end \0 to result - engine objects are now created with a TABLE_SHARE object instead of a TABLE object. - After creating a field object one must call field->init(table) before using it - For a busy system this change will give you: - Less memory usage for table object - Faster opening of tables (if it's has been in use or is in table definition cache) - Allow you to cache many table definitions objects - Faster drop of table
2005-11-23 21:45:02 +01:00
table_ref->table->s->table_name.str));
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return table_ref->table_name;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
Bug#36086: SELECT * from views don't check column grants This patch also fixes bugs 36963 and 35600. - In many places a view was confused with an anonymous derived table, i.e. access checking was skipped. Fixed by introducing a predicate to tell the difference between named and anonymous derived tables. - When inserting fields for "SELECT * ", there was no distinction between base tables and views, where one should be made. View privileges are checked elsewhere.
2008-09-03 16:45:40 +02:00
const char *Field_iterator_table_ref::get_db_name()
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
if (table_ref->view)
return table_ref->view_db.str;
else if (table_ref->is_natural_join)
return natural_join_it.column_ref()->db_name();
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
/*
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
Test that TABLE_LIST::db is the same as TABLE_SHARE::db to
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
ensure consistency. An exception are I_S schema tables, which
are inconsistent in this respect.
*/
Merge mysql.com:/extern/mysql/5.0/bug17476/mysql-5.0 into mysql.com:/extern/mysql/5.1/generic/mysql-5.1-new
2006-03-03 12:03:27 +01:00
DBUG_ASSERT(!strcmp(table_ref->db, table_ref->table->s->db.str) ||
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
(table_ref->schema_table &&
Manual merge from mysql-trunk-merge. Conflicts: - sql/event_data_objects.cc - sql/events.cc - sql/mysql_priv.h - sql/repl_failsafe.cc - sql/sql_parse.cc - sql/sql_show.cc - sql/sql_view.cc - sql/table.cc
2010-01-27 22:10:37 +01:00
is_infoschema_db(table_ref->table->s->db.str,
table_ref->table->s->db.length)));
Fix for BUG#17523: natural join and information schema. The cause of the bug was an ASSERT that checked the consistency of TABLE_SHARE::db and TABLE_LIST::db and failed for I_S tables. The fix relaxes the requirement for consistency for I_S.
2006-02-22 09:04:10 +01:00
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return table_ref->db;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
GRANT_INFO *Field_iterator_table_ref::grant()
{
if (table_ref->view)
return &(table_ref->grant);
else if (table_ref->is_natural_join)
return natural_join_it.column_ref()->grant();
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return &(table_ref->table->grant);
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
/*
Create new or return existing column reference to a column of a
natural/using join.
SYNOPSIS
Field_iterator_table_ref::get_or_create_column_ref()
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
parent_table_ref the parent table reference over which the
iterator is iterating
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
DESCRIPTION
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
Create a new natural join column for the current field of the
iterator if no such column was created, or return an already
created natural join column. The former happens for base tables or
views, and the latter for natural/using joins. If a new field is
created, then the field is added to 'parent_table_ref' if it is
given, or to the original table referene of the field if
parent_table_ref == NULL.
NOTES
This method is designed so that when a Field_iterator_table_ref
walks through the fields of a table reference, all its fields
are created and stored as follows:
- If the table reference being iterated is a stored table, view or
natural/using join, store all natural join columns in a list
attached to that table reference.
- If the table reference being iterated is a nested join that is
not natural/using join, then do not materialize its result
fields. This is OK because for such table references
Field_iterator_table_ref iterates over the fields of the nested
table references (recursively). In this way we avoid the storage
of unnecessay copies of result columns of nested joins.
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
RETURN
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
# Pointer to a column of a natural join (or its operand)
NULL No memory to allocate the column
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
*/
Natural_join_column *
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
Field_iterator_table_ref::get_or_create_column_ref(THD *thd, TABLE_LIST *parent_table_ref)
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
{
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
Natural_join_column *nj_col;
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
bool is_created= TRUE;
uint field_count;
TABLE_LIST *add_table_ref= parent_table_ref ?
parent_table_ref : table_ref;
Fixed compiler and valgrind warnings Added missing DBUG_xxx_RETURN statements Fixed some usage of not initialized variables (as found by valgrind) Ensure that we don't remove locked tables used as name locks from open table cache until unlock_table_names() are called. This was fixed by having drop_locked_name() returning any table used as a name lock so that we can free it in unlock_table_names() This will allow Tomas to continue with his work to use namelocks to syncronize things. Note: valgrind still produces a lot of warnings about using not initialized code and shows memory loss errors when running the ndb tests
2006-03-29 13:27:36 +02:00
LINT_INIT(field_count);
Merge mysql.com:/home/my/mysql-5.0 into mysql.com:/home/my/mysql-5.1 Merge of 'remove compiler warnings when using -Wshadow'
2007-01-27 02:46:45 +01:00
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
if (field_it == &table_field_it)
WL#2486 - natural and using join according to SQL:2003 - Corrected problem with N-way nested natural joins in PS mode. - Code cleanup - More asserts to check consistency of name resolution contexts - Fixed potential memory leak of name resolution contexts
2005-08-19 14:22:30 +02:00
{
/* The field belongs to a stored table. */
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
Field *tmp_field= table_field_it.field();
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
Item_field *tmp_item=
new Item_field(thd, &thd->lex->current_select->context, tmp_field);
if (!tmp_item)
return NULL;
nj_col= new Natural_join_column(tmp_item, table_ref);
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
field_count= table_ref->table->s->fields;
WL#2486 - natural and using join according to SQL:2003 - Corrected problem with N-way nested natural joins in PS mode. - Code cleanup - More asserts to check consistency of name resolution contexts - Fixed potential memory leak of name resolution contexts
2005-08-19 14:22:30 +02:00
}
else if (field_it == &view_field_it)
{
/* The field belongs to a merge view or information schema table. */
Field_translator *translated_field= view_field_it.field_translator();
nj_col= new Natural_join_column(translated_field, table_ref);
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
field_count= table_ref->field_translation_end -
table_ref->field_translation;
WL#2486 - natural and using join according to SQL:2003 - Corrected problem with N-way nested natural joins in PS mode. - Code cleanup - More asserts to check consistency of name resolution contexts - Fixed potential memory leak of name resolution contexts
2005-08-19 14:22:30 +02:00
}
else
{
/*
The field belongs to a NATURAL join, therefore the column reference was
already created via one of the two constructor calls above. In this case
we just return the already created column reference.
*/
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
DBUG_ASSERT(table_ref->is_join_columns_complete);
is_created= FALSE;
WL#2486 - natural and using join according to SQL:2003 - Corrected problem with N-way nested natural joins in PS mode. - Code cleanup - More asserts to check consistency of name resolution contexts - Fixed potential memory leak of name resolution contexts
2005-08-19 14:22:30 +02:00
nj_col= natural_join_it.column_ref();
DBUG_ASSERT(nj_col);
}
Cleanup during review of new pushed code
2005-08-22 00:13:37 +02:00
DBUG_ASSERT(!nj_col->table_field ||
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
nj_col->table_ref->table == nj_col->table_field->field->table);
Fix for BUG#15229. The cause of this bug was a design flaw due to which the list of natural join columns was incorrectly computed and stored for nested joins that are not natural joins, but are operands (possibly indirect) of nested joins. The patch corrects the flaw in a such a way, that the result columns of a table reference are materialized only if it is a leaf table (that is, only if it is a view, stored table, or natural/using join).
2006-03-02 10:50:15 +01:00
/*
If the natural join column was just created add it to the list of
natural join columns of either 'parent_table_ref' or to the table
reference that directly contains the original field.
*/
if (is_created)
{
/* Make sure not all columns were materialized. */
DBUG_ASSERT(!add_table_ref->is_join_columns_complete);
if (!add_table_ref->join_columns)
{
/* Create a list of natural join columns on demand. */
if (!(add_table_ref->join_columns= new List<Natural_join_column>))
return NULL;
add_table_ref->is_join_columns_complete= FALSE;
}
add_table_ref->join_columns->push_back(nj_col);
/*
If new fields are added to their original table reference, mark if
all fields were added. We do it here as the caller has no easy way
of knowing when to do it.
If the fields are being added to parent_table_ref, then the caller
must take care to mark when all fields are created/added.
*/
if (!parent_table_ref &&
add_table_ref->join_columns->elements == field_count)
add_table_ref->is_join_columns_complete= TRUE;
}
Cleanups and optimization during review of new code
2005-08-18 02:12:42 +02:00
return nj_col;
Implementation of WL#2486 - "Process NATURAL and USING joins according to SQL:2003". * Some of the main problems fixed by the patch: - in "select *" queries the * expanded correctly according to ANSI for arbitrary natural/using joins - natural/using joins are correctly transformed into JOIN ... ON for any number/nesting of the joins. - column references are correctly resolved against natural joins of any nesting and combined with arbitrary other joins. * This patch also contains a fix for name resolution of items inside the ON condition of JOIN ... ON - in this case items must be resolved only against the JOIN operands. To support such 'local' name resolution, the patch introduces a stack of name resolution contexts used at parse time. NOTICE: - This patch is not complete in the sense that - there are 2 test cases that still do not pass - one in join.test, one in select.test. Both are marked with a comment "TODO: WL#2486". - it does not include a new test specific for the task
2005-08-12 16:57:19 +02:00
}
WL#2486 - Natural/using join according to SQL:2003. Post-review fixes according to Monty's review.
2005-11-28 20:57:50 +01:00
/*
Return an existing reference to a column of a natural/using join.
SYNOPSIS
Field_iterator_table_ref::get_natural_column_ref()
DESCRIPTION
The method should be called in contexts where it is expected that
all natural join columns are already created, and that the column
being retrieved is a Natural_join_column.
RETURN
# Pointer to a column of a natural join (or its operand)
NULL No memory to allocate the column
*/
Natural_join_column *
Field_iterator_table_ref::get_natural_column_ref()
{
Natural_join_column *nj_col;
DBUG_ASSERT(field_it == &natural_join_it);
/*
The field belongs to a NATURAL join, therefore the column reference was
already created via one of the two constructor calls above. In this case
we just return the already created column reference.
*/
nj_col= natural_join_it.column_ref();
DBUG_ASSERT(nj_col &&
(!nj_col->table_field ||
Bug #38691: segfault/abort in ``UPDATE ...JOIN'' while ``FLUSH TABLES WITH READ LOCK'' Concurrent execution of 1) multitable update with a NATURAL/USING join and 2) a such query as "FLUSH TABLES WITH READ LOCK" or "ALTER TABLE" of updating table led to a server crash. The mysql_multi_update_prepare() function call is optimized to lock updating tables only, so it postpones locking to the last, and if locking fails, it does cleanup of modified syntax structures and repeats a query analysis. However, that cleanup procedure was incomplete for NATURAL/USING join syntax data: 1) some Field_item items pointed into freed table structures, and 2) the TABLE_LIST::join_columns fields was not reset. Major change: short-living Field *Natural_join_column::table_field has been replaced with long-living Item*.
2008-10-07 23:34:00 +02:00
nj_col->table_ref->table == nj_col->table_field->field->table));
WL#2486 - Natural/using join according to SQL:2003. Post-review fixes according to Monty's review.
2005-11-28 20:57:50 +01:00
return nj_col;
}
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
/*****************************************************************************
Functions to handle column usage bitmaps (read_set, write_set etc...)
*****************************************************************************/
/* Reset all columns bitmaps */
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::clear_column_bitmaps()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
/*
Reset column read/write usage. It's identical to:
bitmap_clear_all(&table->def_read_set);
bitmap_clear_all(&table->def_write_set);
*/
bzero((char*) def_read_set.bitmap, s->column_bitmap_size*2);
column_bitmaps_set(&def_read_set, &def_write_set);
}
/*
Tell handler we are going to call position() and rnd_pos() later.
NOTES:
This is needed for handlers that uses the primary key to find the
row. In this case we have to extend the read bitmap with the primary
key fields.
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::prepare_for_position()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
DBUG_ENTER("TABLE::prepare_for_position");
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
if ((file->ha_table_flags() & HA_PRIMARY_KEY_IN_READ_INDEX) &&
s->primary_key < MAX_KEY)
{
mark_columns_used_by_index_no_reset(s->primary_key, read_set);
/* signal change */
file->column_bitmaps_signal();
}
DBUG_VOID_RETURN;
}
/*
Mark that only fields from one key is used
NOTE:
This changes the bitmap to use the tmp bitmap
After this, you can't access any other columns in the table until
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
bitmaps are reset, for example with TABLE::clear_column_bitmaps()
or TABLE::restore_column_maps_after_mark_index()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::mark_columns_used_by_index(uint index)
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
MY_BITMAP *bitmap= &tmp_set;
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
DBUG_ENTER("TABLE::mark_columns_used_by_index");
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
BUG#49902 - SELECT returns incorrect results Queries optimized with GROUP_MIN_MAX didn't cleanup KEYREAD optimization properly. As a result subsequent queries may return incomplete rows (fields are initialized to default values).
2010-02-09 09:53:13 +01:00
set_keyread(TRUE);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
bitmap_clear_all(bitmap);
mark_columns_used_by_index_no_reset(index, bitmap);
column_bitmaps_set(bitmap, bitmap);
DBUG_VOID_RETURN;
}
/*
Restore to use normal column maps after key read
NOTES
This reverse the change done by mark_columns_used_by_index
WARNING
For this to work, one must have the normal table maps in place
when calling mark_columns_used_by_index
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::restore_column_maps_after_mark_index()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
DBUG_ENTER("TABLE::restore_column_maps_after_mark_index");
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
BUG#49902 - SELECT returns incorrect results Queries optimized with GROUP_MIN_MAX didn't cleanup KEYREAD optimization properly. As a result subsequent queries may return incomplete rows (fields are initialized to default values).
2010-02-09 09:53:13 +01:00
set_keyread(FALSE);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
default_column_bitmaps();
file->column_bitmaps_signal();
DBUG_VOID_RETURN;
}
/*
mark columns used by key, but don't reset other fields
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::mark_columns_used_by_index_no_reset(uint index,
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
MY_BITMAP *bitmap)
{
KEY_PART_INFO *key_part= key_info[index].key_part;
KEY_PART_INFO *key_part_end= (key_part +
key_info[index].key_parts);
for (;key_part != key_part_end; key_part++)
bitmap_set_bit(bitmap, key_part->fieldnr-1);
}
/*
Mark auto-increment fields as used fields in both read and write maps
NOTES
This is needed in insert & update as the auto-increment field is
always set and sometimes read.
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::mark_auto_increment_column()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
DBUG_ASSERT(found_next_number_field);
/*
We must set bit in read set as update_auto_increment() is using the
store() to check overflow of auto_increment values
*/
bitmap_set_bit(read_set, found_next_number_field->field_index);
bitmap_set_bit(write_set, found_next_number_field->field_index);
WL#3700: Handler API change: all index search methods - that is, index_read(), index_read_idx(), index_read_last(), and records_in_range() - instead of 'uint keylen' argument take 'ulonglong keypart_map', a bitmap showing which keyparts are present in the key value. Fallback method is provided for handlers that are lagging behind.
2007-01-29 10:40:26 +01:00
if (s->next_number_keypart)
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
mark_columns_used_by_index_no_reset(s->next_number_index, read_set);
file->column_bitmaps_signal();
}
/*
Mark columns needed for doing an delete of a row
DESCRIPTON
Some table engines don't have a cursor on the retrieve rows
so they need either to use the primary key or all columns to
be able to delete a row.
If the engine needs this, the function works as follows:
- If primary key exits, mark the primary key columns to be read.
- If not, mark all columns to be read
If the engine has HA_REQUIRES_KEY_COLUMNS_FOR_DELETE, we will
mark all key columns as 'to-be-read'. This allows the engine to
loop over the given record to find all keys and doesn't have to
retrieve the row again.
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::mark_columns_needed_for_delete()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
if (triggers)
After merge fixes for patch solving bug#18437 "Wrong values inserted with a before update trigger on NDB table". Two main changes: - We use TABLE::read_set/write_set bitmaps for marking fields used by statement instead of Field::query_id in 5.1. - Now when we mark columns used by statement we take into account columns used by table's triggers instead of marking all columns as used if table has triggers.
2006-07-06 11:33:23 +02:00
triggers->mark_fields_used(TRG_EVENT_DELETE);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
if (file->ha_table_flags() & HA_REQUIRES_KEY_COLUMNS_FOR_DELETE)
{
Field **reg_field;
for (reg_field= field ; *reg_field ; reg_field++)
{
if ((*reg_field)->flags & PART_KEY_FLAG)
bitmap_set_bit(read_set, (*reg_field)->field_index);
}
file->column_bitmaps_signal();
}
if (file->ha_table_flags() & HA_PRIMARY_KEY_REQUIRED_FOR_DELETE)
{
/*
If the handler has no cursor capabilites, we have to read either
the primary key, the hidden primary key or all columns to be
able to do an delete
*/
if (s->primary_key == MAX_KEY)
file->use_hidden_primary_key();
else
{
mark_columns_used_by_index_no_reset(s->primary_key, read_set);
file->column_bitmaps_signal();
}
}
}
/*
Mark columns needed for doing an update of a row
DESCRIPTON
Some engines needs to have all columns in an update (to be able to
build a complete row). If this is the case, we mark all not
updated columns to be read.
If this is no the case, we do like in the delete case and mark
if neeed, either the primary key column or all columns to be read.
(see mark_columns_needed_for_delete() for details)
If the engine has HA_REQUIRES_KEY_COLUMNS_FOR_DELETE, we will
mark all USED key columns as 'to-be-read'. This allows the engine to
loop over the given record to find all changed keys and doesn't have to
retrieve the row again.
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::mark_columns_needed_for_update()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
DBUG_ENTER("mark_columns_needed_for_update");
if (triggers)
After merge fixes for patch solving bug#18437 "Wrong values inserted with a before update trigger on NDB table". Two main changes: - We use TABLE::read_set/write_set bitmaps for marking fields used by statement instead of Field::query_id in 5.1. - Now when we mark columns used by statement we take into account columns used by table's triggers instead of marking all columns as used if table has triggers.
2006-07-06 11:33:23 +02:00
triggers->mark_fields_used(TRG_EVENT_UPDATE);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
if (file->ha_table_flags() & HA_REQUIRES_KEY_COLUMNS_FOR_DELETE)
{
/* Mark all used key columns for read */
Field **reg_field;
for (reg_field= field ; *reg_field ; reg_field++)
{
/* Merge keys is all keys that had a column refered to in the query */
if (merge_keys.is_overlapping((*reg_field)->part_of_key))
bitmap_set_bit(read_set, (*reg_field)->field_index);
}
file->column_bitmaps_signal();
}
if (file->ha_table_flags() & HA_PRIMARY_KEY_REQUIRED_FOR_DELETE)
{
/*
If the handler has no cursor capabilites, we have to read either
the primary key, the hidden primary key or all columns to be
able to do an update
*/
if (s->primary_key == MAX_KEY)
file->use_hidden_primary_key();
else
{
mark_columns_used_by_index_no_reset(s->primary_key, read_set);
file->column_bitmaps_signal();
}
}
DBUG_VOID_RETURN;
}
/*
Mark columns the handler needs for doing an insert
For now, this is used to mark fields used by the trigger
as changed.
*/
Backport of: ---------------------------------------------------------- revno: 2630.22.8 committer: Konstantin Osipov <konstantin@mysql.com> branch nick: mysql-6.0-runtime timestamp: Sun 2008-08-10 18:49:52 +0400 message: Get rid of typedef struct for the most commonly used types: TABLE, TABLE_SHARE, LEX. This simplifies use of tags and forward declarations.
2009-10-14 13:14:58 +02:00
void TABLE::mark_columns_needed_for_insert()
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
{
if (triggers)
{
After merge fixes for patch solving bug#18437 "Wrong values inserted with a before update trigger on NDB table". Two main changes: - We use TABLE::read_set/write_set bitmaps for marking fields used by statement instead of Field::query_id in 5.1. - Now when we mark columns used by statement we take into account columns used by table's triggers instead of marking all columns as used if table has triggers.
2006-07-06 11:33:23 +02:00
/*
We don't need to mark columns which are used by ON DELETE and
ON UPDATE triggers, which may be invoked in case of REPLACE or
INSERT ... ON DUPLICATE KEY UPDATE, since before doing actual
row replacement or update write_record() will mark all table
fields as used.
*/
triggers->mark_fields_used(TRG_EVENT_INSERT);
This changeset is largely a handler cleanup changeset (WL#3281), but includes fixes and cleanups that was found necessary while testing the handler changes Changes that requires code changes in other code of other storage engines. (Note that all changes are very straightforward and one should find all issues by compiling a --debug build and fixing all compiler errors and all asserts in field.cc while running the test suite), - New optional handler function introduced: reset() This is called after every DML statement to make it easy for a handler to statement specific cleanups. (The only case it's not called is if force the file to be closed) - handler::extra(HA_EXTRA_RESET) is removed. Code that was there before should be moved to handler::reset() - table->read_set contains a bitmap over all columns that are needed in the query. read_row() and similar functions only needs to read these columns - table->write_set contains a bitmap over all columns that will be updated in the query. write_row() and update_row() only needs to update these columns. The above bitmaps should now be up to date in all context (including ALTER TABLE, filesort()). The handler is informed of any changes to the bitmap after fix_fields() by calling the virtual function handler::column_bitmaps_signal(). If the handler does caching of these bitmaps (instead of using table->read_set, table->write_set), it should redo the caching in this code. as the signal() may be sent several times, it's probably best to set a variable in the signal and redo the caching on read_row() / write_row() if the variable was set. - Removed the read_set and write_set bitmap objects from the handler class - Removed all column bit handling functions from the handler class. (Now one instead uses the normal bitmap functions in my_bitmap.c instead of handler dedicated bitmap functions) - field->query_id is removed. One should instead instead check table->read_set and table->write_set if a field is used in the query. - handler::extra(HA_EXTRA_RETRIVE_ALL_COLS) and handler::extra(HA_EXTRA_RETRIEVE_PRIMARY_KEY) are removed. One should now instead use table->read_set to check for which columns to retrieve. - If a handler needs to call Field->val() or Field->store() on columns that are not used in the query, one should install a temporary all-columns-used map while doing so. For this, we provide the following functions: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->read_set); field->val(); dbug_tmp_restore_column_map(table->read_set, old_map); and similar for the write map: my_bitmap_map *old_map= dbug_tmp_use_all_columns(table, table->write_set); field->val(); dbug_tmp_restore_column_map(table->write_set, old_map); If this is not done, you will sooner or later hit a DBUG_ASSERT in the field store() / val() functions. (For not DBUG binaries, the dbug_tmp_restore_column_map() and dbug_tmp_restore_column_map() are inline dummy functions and should be optimized away be the compiler). - If one needs to temporary set the column map for all binaries (and not just to avoid the DBUG_ASSERT() in the Field::store() / Field::val() methods) one should use the functions tmp_use_all_columns() and tmp_restore_column_map() instead of the above dbug_ variants. - All 'status' fields in the handler base class (like records, data_file_length etc) are now stored in a 'stats' struct. This makes it easier to know what status variables are provided by the base handler. This requires some trivial variable names in the extra() function. - New virtual function handler::records(). This is called to optimize COUNT(*) if (handler::table_flags() & HA_HAS_RECORDS()) is true. (stats.records is not supposed to be an exact value. It's only has to be 'reasonable enough' for the optimizer to be able to choose a good optimization path). - Non virtual handler::init() function added for caching of virtual constants from engine. - Removed has_transactions() virtual method. Now one should instead return HA_NO_TRANSACTIONS in table_flags() if the table handler DOES NOT support transactions. - The 'xxxx_create_handler()' function now has a MEM_ROOT_root argument that is to be used with 'new handler_name()' to allocate the handler in the right area. The xxxx_create_handler() function is also responsible for any initialization of the object before returning. For example, one should change: static handler *myisam_create_handler(TABLE_SHARE *table) { return new ha_myisam(table); } -> static handler *myisam_create_handler(TABLE_SHARE *table, MEM_ROOT *mem_root) { return new (mem_root) ha_myisam(table); } - New optional virtual function: use_hidden_primary_key(). This is called in case of an update/delete when (table_flags() and HA_PRIMARY_KEY_REQUIRED_FOR_DELETE) is defined but we don't have a primary key. This allows the handler to take precisions in remembering any hidden primary key to able to update/delete any found row. The default handler marks all columns to be read. - handler::table_flags() now returns a ulonglong (to allow for more flags). - New/changed table_flags() - HA_HAS_RECORDS Set if ::records() is supported - HA_NO_TRANSACTIONS Set if engine doesn't support transactions - HA_PRIMARY_KEY_REQUIRED_FOR_DELETE Set if we should mark all primary key columns for read when reading rows as part of a DELETE statement. If there is no primary key, all columns are marked for read. - HA_PARTIAL_COLUMN_READ Set if engine will not read all columns in some cases (based on table->read_set) - HA_PRIMARY_KEY_ALLOW_RANDOM_ACCESS Renamed to HA_PRIMARY_KEY_REQUIRED_FOR_POSITION. - HA_DUPP_POS Renamed to HA_DUPLICATE_POS - HA_REQUIRES_KEY_COLUMNS_FOR_DELETE Set this if we should mark ALL key columns for read when when reading rows as part of a DELETE statement. In case of an update we will mark all keys for read for which key part changed value. - HA_STATS_RECORDS_IS_EXACT Set this if stats.records is exact. (This saves us some extra records() calls when optimizing COUNT(*)) - Removed table_flags() - HA_NOT_EXACT_COUNT Now one should instead use HA_HAS_RECORDS if handler::records() gives an exact count() and HA_STATS_RECORDS_IS_EXACT if stats.records is exact. - HA_READ_RND_SAME Removed (no one supported this one) - Removed not needed functions ha_retrieve_all_cols() and ha_retrieve_all_pk() - Renamed handler::dupp_pos to handler::dup_pos - Removed not used variable handler::sortkey Upper level handler changes: - ha_reset() now does some overall checks and calls ::reset() - ha_table_flags() added. This is a cached version of table_flags(). The cache is updated on engine creation time and updated on open. MySQL level changes (not obvious from the above): - DBUG_ASSERT() added to check that column usage matches what is set in the column usage bit maps. (This found a LOT of bugs in current column marking code). - In 5.1 before, all used columns was marked in read_set and only updated columns was marked in write_set. Now we only mark columns for which we need a value in read_set. - Column bitmaps are created in open_binary_frm() and open_table_from_share(). (Before this was in table.cc) - handler::table_flags() calls are replaced with handler::ha_table_flags() - For calling field->val() you must have the corresponding bit set in table->read_set. For calling field->store() you must have the corresponding bit set in table->write_set. (There are asserts in all store()/val() functions to catch wrong usage) - thd->set_query_id is renamed to thd->mark_used_columns and instead of setting this to an integer value, this has now the values: MARK_COLUMNS_NONE, MARK_COLUMNS_READ, MARK_COLUMNS_WRITE Changed also all variables named 'set_query_id' to mark_used_columns. - In filesort() we now inform the handler of exactly which columns are needed doing the sort and choosing the rows. - The TABLE_SHARE object has a 'all_set' column bitmap one can use when one needs a column bitmap with all columns set. (This is used for table->use_all_columns() and other places) - The TABLE object has 3 column bitmaps: - def_read_set Default bitmap for columns to be read - def_write_set Default bitmap for columns to be written - tmp_set Can be used as a temporary bitmap when needed. The table object has also two pointer to bitmaps read_set and write_set that the handler should use to find out which columns are used in which way. - count() optimization now calls handler::records() instead of using handler->stats.records (if (table_flags() & HA_HAS_RECORDS) is true). - Added extra argument to Item::walk() to indicate if we should also traverse sub queries. - Added TABLE parameter to cp_buffer_from_ref() - Don't close tables created with CREATE ... SELECT but keep them in the table cache. (Faster usage of newly created tables). New interfaces: - table->clear_column_bitmaps() to initialize the bitmaps for tables at start of new statements. - table->column_bitmaps_set() to set up new column bitmaps and signal the handler about this. - table->column_bitmaps_set_no_signal() for some few cases where we need to setup new column bitmaps but don't signal the handler (as the handler has already been signaled about these before). Used for the momement only in opt_range.cc when doing ROR scans. - table->use_all_columns() to install a bitmap where all columns are marked as use in the read and the write set. - table->default_column_bitmaps() to install the normal read and write column bitmaps, but not signaling the handler about this. This is mainly used when creating TABLE instances. - table->mark_columns_needed_for_delete(), table->mark_columns_needed_for_delete() and table->mark_columns_needed_for_insert() to allow us to put additional columns in column usage maps if handler so requires. (The handler indicates what it neads in handler->table_flags()) - table->prepare_for_position() to allow us to tell handler that it needs to read primary key parts to be able to store them in future table->position() calls. (This replaces the table->file->ha_retrieve_all_pk function) - table->mark_auto_increment_column() to tell handler are going to update columns part of any auto_increment key. - table->mark_columns_used_by_index() to mark all columns that is part of an index. It will also send extra(HA_EXTRA_KEYREAD) to handler to allow it to quickly know that it only needs to read colums that are part of the key. (The handler can also use the column map for detecting this, but simpler/faster handler can just monitor the extra() call). - table->mark_columns_used_by_index_no_reset() to in addition to other columns, also mark all columns that is used by the given key. - table->restore_column_maps_after_mark_index() to restore to default column maps after a call to table->mark_columns_used_by_index(). - New item function register_field_in_read_map(), for marking used columns in table->read_map. Used by filesort() to mark all used columns - Maintain in TABLE->merge_keys set of all keys that are used in query. (Simplices some optimization loops) - Maintain Field->part_of_key_not_clustered which is like Field->part_of_key but the field in the clustered key is not assumed to be part of all index. (used in opt_range.cc for faster loops) - dbug_tmp_use_all_columns(), dbug_tmp_restore_column_map() tmp_use_all_columns() and tmp_restore_column_map() functions to temporally mark all columns as usable. The 'dbug_' version is primarily intended inside a handler when it wants to just call Field:store() & Field::val() functions, but don't need the column maps set for any other usage. (ie:: bitmap_is_set() is never called) - We can't use compare_records() to skip updates for handlers that returns a partial column set and the read_set doesn't cover all columns in the write set. The reason for this is that if we have a column marked only for write we can't in the MySQL level know if the value changed or not. The reason this worked before was that MySQL marked all to be written columns as also to be read. The new 'optimal' bitmaps exposed this 'hidden bug'. - open_table_from_share() does not anymore setup temporary MEM_ROOT object as a thread specific variable for the handler. Instead we send the to-be-used MEMROOT to get_new_handler(). (Simpler, faster code) Bugs fixed: - Column marking was not done correctly in a lot of cases. (ALTER TABLE, when using triggers, auto_increment fields etc) (Could potentially result in wrong values inserted in table handlers relying on that the old column maps or field->set_query_id was correct) Especially when it comes to triggers, there may be cases where the old code would cause lost/wrong values for NDB and/or InnoDB tables. - Split thd->options flag OPTION_STATUS_NO_TRANS_UPDATE to two flags: OPTION_STATUS_NO_TRANS_UPDATE and OPTION_KEEP_LOG. This allowed me to remove some wrong warnings about: "Some non-transactional changed tables couldn't be rolled back" - Fixed handling of INSERT .. SELECT and CREATE ... SELECT that wrongly reset (thd->options & OPTION_STATUS_NO_TRANS_UPDATE) which caused us to loose some warnings about "Some non-transactional changed tables couldn't be rolled back") - Fixed use of uninitialized memory in ha_ndbcluster.cc::delete_table() which could cause delete_table to report random failures. - Fixed core dumps for some tests when running with --debug - Added missing FN_LIBCHAR in mysql_rm_tmp_tables() (This has probably caused us to not properly remove temporary files after crash) - slow_logs was not properly initialized, which could maybe cause extra/lost entries in slow log. - If we get an duplicate row on insert, change column map to read and write all columns while retrying the operation. This is required by the definition of REPLACE and also ensures that fields that are only part of UPDATE are properly handled. This fixed a bug in NDB and REPLACE where REPLACE wrongly copied some column values from the replaced row. - For table handler that doesn't support NULL in keys, we would give an error when creating a primary key with NULL fields, even after the fields has been automaticly converted to NOT NULL. - Creating a primary key on a SPATIAL key, would fail if field was not declared as NOT NULL. Cleanups: - Removed not used condition argument to setup_tables - Removed not needed item function reset_query_id_processor(). - Field->add_index is removed. Now this is instead maintained in (field->flags & FIELD_IN_ADD_INDEX) - Field->fieldnr is removed (use field->field_index instead) - New argument to filesort() to indicate that it should return a set of row pointers (not used columns). This allowed me to remove some references to sql_command in filesort and should also enable us to return column results in some cases where we couldn't before. - Changed column bitmap handling in opt_range.cc to be aligned with TABLE bitmap, which allowed me to use bitmap functions instead of looping over all fields to create some needed bitmaps. (Faster and smaller code) - Broke up found too long lines - Moved some variable declaration at start of function for better code readability. - Removed some not used arguments from functions. (setup_fields(), mysql_prepare_insert_check_table()) - setup_fields() now takes an enum instead of an int for marking columns usage. - For internal temporary tables, use handler::write_row(), handler::delete_row() and handler::update_row() instead of handler::ha_xxxx() for faster execution. - Changed some constants to enum's and define's. - Using separate column read and write sets allows for easier checking of timestamp field was set by statement. - Remove calls to free_io_cache() as this is now done automaticly in ha_reset() - Don't build table->normalized_path as this is now identical to table->path (after bar's fixes to convert filenames) - Fixed some missed DBUG_PRINT(.."%lx") to use "0x%lx" to make it easier to do comparision with the 'convert-dbug-for-diff' tool. Things left to do in 5.1: - We wrongly log failed CREATE TABLE ... SELECT in some cases when using row based logging (as shown by testcase binlog_row_mix_innodb_myisam.result) Mats has promised to look into this. - Test that my fix for CREATE TABLE ... SELECT is indeed correct. (I added several test cases for this, but in this case it's better that someone else also tests this throughly). Lars has promosed to do this.
2006-06-04 17:52:22 +02:00
}
if (found_next_number_field)
mark_auto_increment_column();
}
Bug#26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Bug 26377 - Deadlock with MERGE and FLUSH TABLE Bug 25038 - Waiting TRUNCATE Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Bug 19627 - temporary merge table locking Bug 27660 - Falcon: merge table possible Bug 30273 - merge tables: Can't lock file (errno: 155) The problems were: Bug 26379 - Combination of FLUSH TABLE and REPAIR TABLE corrupts a MERGE table 1. A thread trying to lock a MERGE table performs busy waiting while REPAIR TABLE or a similar table administration task is ongoing on one or more of its MyISAM tables. 2. A thread trying to lock a MERGE table performs busy waiting until all threads that did REPAIR TABLE or similar table administration tasks on one or more of its MyISAM tables in LOCK TABLES segments do UNLOCK TABLES. The difference against problem #1 is that the busy waiting takes place *after* the administration task. It is terminated by UNLOCK TABLES only. 3. Two FLUSH TABLES within a LOCK TABLES segment can invalidate the lock. This does *not* require a MERGE table. The first FLUSH TABLES can be replaced by any statement that requires other threads to reopen the table. In 5.0 and 5.1 a single FLUSH TABLES can provoke the problem. Bug 26867 - LOCK TABLES + REPAIR + merge table result in memory/cpu hogging Trying DML on a MERGE table, which has a child locked and repaired by another thread, made an infinite loop in the server. Bug 26377 - Deadlock with MERGE and FLUSH TABLE Locking a MERGE table and its children in parent-child order and flushing the child deadlocked the server. Bug 25038 - Waiting TRUNCATE Truncating a MERGE child, while the MERGE table was in use, let the truncate fail instead of waiting for the table to become free. Bug 25700 - merge base tables get corrupted by optimize/analyze/repair table Repairing a child of an open MERGE table corrupted the child. It was necessary to FLUSH the child first. Bug 30275 - Merge tables: flush tables or unlock tables causes server to crash Flushing and optimizing locked MERGE children crashed the server. Bug 19627 - temporary merge table locking Use of a temporary MERGE table with non-temporary children could corrupt the children. Temporary tables are never locked. So we do now prohibit non-temporary chidlren of a temporary MERGE table. Bug 27660 - Falcon: merge table possible It was possible to create a MERGE table with non-MyISAM children. Bug 30273 - merge tables: Can't lock file (errno: 155) This was a Windows-only bug. Table administration statements sometimes failed with "Can't lock file (errno: 155)". These bugs are fixed by a new implementation of MERGE table open. When opening a MERGE table in open_tables() we do now add the child tables to the list of tables to be opened by open_tables() (the "query_list"). The children are not opened in the handler at this stage. After opening the parent, open_tables() opens each child from the now extended query_list. When the last child is opened, we remove the children from the query_list again and attach the children to the parent. This behaves similar to the old open. However it does not open the MyISAM tables directly, but grabs them from the already open children. When closing a MERGE table in close_thread_table() we detach the children only. Closing of the children is done implicitly because they are in thd->open_tables. For more detail see the comment at the top of ha_myisammrg.cc. Changed from open_ltable() to open_and_lock_tables() in all places that can be relevant for MERGE tables. The latter can handle tables added to the list on the fly. When open_ltable() was used in a loop over a list of tables, the list must be temporarily terminated after every table for open_and_lock_tables(). table_list->required_type is set to FRMTYPE_TABLE to avoid open of special tables. Handling of derived tables is suppressed. These details are handled by the new function open_n_lock_single_table(), which has nearly the same signature as open_ltable() and can replace it in most cases. In reopen_tables() some of the tables open by a thread can be closed and reopened. When a MERGE child is affected, the parent must be closed and reopened too. Closing of the parent is forced before the first child is closed. Reopen happens in the order of thd->open_tables. MERGE parents do not attach their children automatically at open. This is done after all tables are reopened. So all children are open when attaching them. Special lock handling like mysql_lock_abort() or mysql_lock_remove() needs to be suppressed for MERGE children or forwarded to the parent. This depends on the situation. In loops over all open tables one suppresses child lock handling. When a single table is touched, forwarding is done. Behavioral changes: =================== This patch changes the behavior of temporary MERGE tables. Temporary MERGE must have temporary children. The old behavior was wrong. A temporary table is not locked. Hence even non-temporary children were not locked. See Bug 19627 - temporary merge table locking. You cannot change the union list of a non-temporary MERGE table when LOCK TABLES is in effect. The following does *not* work: CREATE TABLE m1 ... ENGINE=MRG_MYISAM ...; LOCK TABLES t1 WRITE, t2 WRITE, m1 WRITE; ALTER TABLE m1 ... UNION=(t1,t2) ...; However, you can do this with a temporary MERGE table. You cannot create a MERGE table with CREATE ... SELECT, neither as a temporary MERGE table, nor as a non-temporary MERGE table. CREATE TABLE m1 ... ENGINE=MRG_MYISAM ... SELECT ...; Gives error message: table is not BASE TABLE.
2007-11-15 20:25:43 +01:00
A fix and a test case for Bug#19399 "res 'Lost Connection' when dropping/creating tables". The bug could lead to a crash when multi-delete statements were prepared and used with temporary tables. The bug was caused by lack of clean-up of multi-delete tables before re-execution of a prepared statement. In a statement like DELETE t1 FROM t1, t2 WHERE ... the first table list (t1) is moved to lex->auxilliary_table_list and excluded from lex->query_tables or select_lex->tables. Thus it was unaccessible to reinit_stmt_before_use and not cleaned up before re-execution of a prepared statement.
2006-07-06 21:59:04 +02:00
/*
Cleanup this table for re-execution.
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::reinit_before_use()
A fix and a test case for Bug#19399 "res 'Lost Connection' when dropping/creating tables". The bug could lead to a crash when multi-delete statements were prepared and used with temporary tables. The bug was caused by lack of clean-up of multi-delete tables before re-execution of a prepared statement. In a statement like DELETE t1 FROM t1, t2 WHERE ... the first table list (t1) is moved to lex->auxilliary_table_list and excluded from lex->query_tables or select_lex->tables. Thus it was unaccessible to reinit_stmt_before_use and not cleaned up before re-execution of a prepared statement.
2006-07-06 21:59:04 +02:00
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
void TABLE_LIST::reinit_before_use(THD *thd)
A fix and a test case for Bug#19399 "res 'Lost Connection' when dropping/creating tables". The bug could lead to a crash when multi-delete statements were prepared and used with temporary tables. The bug was caused by lack of clean-up of multi-delete tables before re-execution of a prepared statement. In a statement like DELETE t1 FROM t1, t2 WHERE ... the first table list (t1) is moved to lex->auxilliary_table_list and excluded from lex->query_tables or select_lex->tables. Thus it was unaccessible to reinit_stmt_before_use and not cleaned up before re-execution of a prepared statement.
2006-07-06 21:59:04 +02:00
{
/*
Reset old pointers to TABLEs: they are not valid since the tables
were closed in the end of previous prepare or execute call.
*/
table= 0;
Post-merge fixes for Bug#19399 "Stored Procedures 'Lost Connection' when dropping/creating tables"
2006-07-11 21:39:51 +02:00
/* Reset is_schema_table_processed value(needed for I_S tables */
Bug#24630 Subselect query crashes mysqld The crash happens because second filling of the same I_S table happens in case of subselect with order by. table->sort.io_cache previously allocated in create_sort_index() is deleted during second filling (function get_schema_tables_result). There are two places where I_S table can be filled: JOIN::exec and create_sort_index(). To fix the bug we should check if the table was already filled in one of these places and skip processing of the table in second.
2007-02-12 13:06:14 +01:00
schema_table_state= NOT_PROCESSED;
Post-merge fixes for Bug#19399 "Stored Procedures 'Lost Connection' when dropping/creating tables"
2006-07-11 21:39:51 +02:00
TABLE_LIST *embedded; /* The table at the current level of nesting. */
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
TABLE_LIST *parent_embedding= this; /* The parent nested table reference. */
Post-merge fixes for Bug#19399 "Stored Procedures 'Lost Connection' when dropping/creating tables"
2006-07-11 21:39:51 +02:00
do
{
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
embedded= parent_embedding;
Post-merge fixes for Bug#19399 "Stored Procedures 'Lost Connection' when dropping/creating tables"
2006-07-11 21:39:51 +02:00
if (embedded->prep_on_expr)
embedded->on_expr= embedded->prep_on_expr->copy_andor_structure(thd);
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
parent_embedding= embedded->embedding;
Post-merge fixes for Bug#19399 "Stored Procedures 'Lost Connection' when dropping/creating tables"
2006-07-11 21:39:51 +02:00
}
Fixed compiler warnings detected by option -Wshadow and -Wunused: - Removed not used variables and functions - Added #ifdef around code that is not used - Renamed variables and functions to avoid conflicts - Removed some not used arguments Fixed some class/struct warnings in ndb Added define IS_LONGDATA() to simplify code in libmysql.c I did run gcov on the changes and added 'purecov' comments on almost all lines that was not just variable name changes
2006-12-14 23:51:37 +01:00
while (parent_embedding &&
parent_embedding->nested_join->join_list.head() == embedded);
Backport of: ---------------------------------------------------------- revno: 2617.69.20 committer: Konstantin Osipov <kostja@sun.com> branch nick: 5.4-4284-1-assert timestamp: Thu 2009-08-13 18:29:55 +0400 message: WL#4284 "Transactional DDL locking" A review fix. Since WL#4284 implementation separated MDL_request and MDL_ticket, MDL_request becamse a utility object necessary only to get a ticket. Store it by-value in TABLE_LIST with the intent to merge MDL_request::key with table_list->table_name and table_list->db in future. Change the MDL subsystem to not require MDL_requests to stay around till close_thread_tables(). Remove the list of requests from the MDL context. Requests for shared metadata locks acquired in open_tables() are only used as a list in recover_from_failed_open_table_attempt(), which calls mdl_context.wait_for_locks() for this list. To keep such list for recover_from_failed_open_table_attempt(), introduce a context class (Open_table_context), that collects all requests. A lot of minor cleanups and simplications that became possible with this change.
2009-12-08 10:57:07 +01:00
mdl_request.ticket= NULL;
/*
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list.
2010-02-01 12:43:06 +01:00
Since we manipulate with the metadata lock type in open_table(),
we need to reset it to the parser default, to restore things back
to first-execution state.
Backport of: ---------------------------------------------------------- revno: 2617.69.20 committer: Konstantin Osipov <kostja@sun.com> branch nick: 5.4-4284-1-assert timestamp: Thu 2009-08-13 18:29:55 +0400 message: WL#4284 "Transactional DDL locking" A review fix. Since WL#4284 implementation separated MDL_request and MDL_ticket, MDL_request becamse a utility object necessary only to get a ticket. Store it by-value in TABLE_LIST with the intent to merge MDL_request::key with table_list->table_name and table_list->db in future. Change the MDL subsystem to not require MDL_requests to stay around till close_thread_tables(). Remove the list of requests from the MDL context. Requests for shared metadata locks acquired in open_tables() are only used as a list in recover_from_failed_open_table_attempt(), which calls mdl_context.wait_for_locks() for this list. To keep such list for recover_from_failed_open_table_attempt(), introduce a context class (Open_table_context), that collects all requests. A lot of minor cleanups and simplications that became possible with this change.
2009-12-08 10:57:07 +01:00
*/
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list.
2010-02-01 12:43:06 +01:00
mdl_request.set_type((lock_type >= TL_WRITE_ALLOW_WRITE) ?
MDL_SHARED_WRITE : MDL_SHARED_READ);
A fix and a test case for Bug#19399 "res 'Lost Connection' when dropping/creating tables". The bug could lead to a crash when multi-delete statements were prepared and used with temporary tables. The bug was caused by lack of clean-up of multi-delete tables before re-execution of a prepared statement. In a statement like DELETE t1 FROM t1, t2 WHERE ... the first table list (t1) is moved to lex->auxilliary_table_list and excluded from lex->query_tables or select_lex->tables. Thus it was unaccessible to reinit_stmt_before_use and not cleaned up before re-execution of a prepared statement.
2006-07-06 21:59:04 +02:00
}
Fixed bug #21727. This is a performance issue for queries with subqueries evaluation of which requires filesort. Allocation of memory for the sort buffer at each evaluation of a subquery may take a significant amount of time if the buffer is rather big. With the fix we allocate the buffer at the first evaluation of the subquery and reuse it at each subsequent evaluation.
2006-11-01 02:31:56 +01:00
/*
Return subselect that contains the FROM list this table is taken from
SYNOPSIS
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
TABLE_LIST::containing_subselect()
Fixed bug #21727. This is a performance issue for queries with subqueries evaluation of which requires filesort. Allocation of memory for the sort buffer at each evaluation of a subquery may take a significant amount of time if the buffer is rather big. With the fix we allocate the buffer at the first evaluation of the subquery and reuse it at each subsequent evaluation.
2006-11-01 02:31:56 +01:00
RETURN
Subselect item for the subquery that contains the FROM list
this table is taken from if there is any
0 - otherwise
*/
Remove typedef st_table_list TABLE_LIST and always use name 'TABLE_LIST'. The need arose when working on Bug 26141, where it became necessary to replace TABLE_LIST with its forward declaration in a few headers, and this involved a lot of s/TABLE_LIST/st_table_list/. Although other workarounds exist, this patch is in line with our general strategy of moving away from typedef-ed names. Sometime in future we might also rename TABLE_LIST to follow the coding style, but this is a huge change.
2007-07-06 14:18:49 +02:00
Item_subselect *TABLE_LIST::containing_subselect()
Fixed bug #21727. This is a performance issue for queries with subqueries evaluation of which requires filesort. Allocation of memory for the sort buffer at each evaluation of a subquery may take a significant amount of time if the buffer is rather big. With the fix we allocate the buffer at the first evaluation of the subquery and reuse it at each subsequent evaluation.
2006-11-01 02:31:56 +01:00
{
return (select_lex ? select_lex->master_unit()->item : 0);
}
WL#2486 - Natural/using join according to SQL:2003. Post-review fixes according to Monty's review.
2005-11-28 20:57:50 +01:00
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
/*
Compiles the tagged hints list and fills up the bitmasks.
SYNOPSIS
process_index_hints()
table the TABLE to operate on.
DESCRIPTION
The parser collects the index hints for each table in a "tagged list"
Post-merge fixes (merge from the main).
2007-07-16 22:59:21 +02:00
(TABLE_LIST::index_hints). Using the information in this tagged list
Merge from mysql-next-mr.
2009-10-23 13:22:21 +02:00
this function sets the members st_table::keys_in_use_for_query,
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
st_table::keys_in_use_for_group_by, st_table::keys_in_use_for_order_by,
Merge from mysql-next-mr.
2009-10-23 13:22:21 +02:00
st_table::force_index, st_table::force_index_order,
Bug #43029: FORCE INDEX FOR ORDER BY is ignored when join buffering is used FORCE INDEX FOR ORDER BY now prevents the optimizer from using join buffering. As a result the optimizer can use indexed access on the first table and doesn't need to sort the complete resultset at the end of the statement.
2009-10-07 17:03:42 +02:00
st_table::force_index_group and st_table::covering_keys.
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
Current implementation of the runtime does not allow mixing FORCE INDEX
and USE INDEX, so this is checked here. Then the FORCE INDEX list
(if non-empty) is appended to the USE INDEX list and a flag is set.
Multiple hints of the same kind are processed so that each clause
is applied to what is computed in the previous clause.
For example:
USE INDEX (i1) USE INDEX (i2)
is equivalent to
USE INDEX (i1,i2)
and means "consider only i1 and i2".
Similarly
USE INDEX () USE INDEX (i1)
is equivalent to
USE INDEX (i1)
and means "consider only the index i1"
It is OK to have the same index several times, e.g. "USE INDEX (i1,i1)" is
not an error.
Different kind of hints (USE/FORCE/IGNORE) are processed in the following
order:
1. All indexes in USE (or FORCE) INDEX are added to the mask.
2. All IGNORE INDEX
e.g. "USE INDEX i1, IGNORE INDEX i1, USE INDEX i1" will not use i1 at all
as if we had "USE INDEX i1, USE INDEX i1, IGNORE INDEX i1".
As an optimization if there is a covering index, and we have
IGNORE INDEX FOR GROUP/ORDER, and this index is used for the JOIN part,
then we have to ignore the IGNORE INDEX FROM GROUP/ORDER.
RETURN VALUE
FALSE no errors found
TRUE found and reported an error.
*/
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
bool TABLE_LIST::process_index_hints(TABLE *tbl)
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
{
/* initialize the result variables */
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->keys_in_use_for_query= tbl->keys_in_use_for_group_by=
tbl->keys_in_use_for_order_by= tbl->s->keys_in_use;
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
/* index hint list processing */
if (index_hints)
{
key_map index_join[INDEX_HINT_FORCE + 1];
key_map index_order[INDEX_HINT_FORCE + 1];
key_map index_group[INDEX_HINT_FORCE + 1];
coding style fix : Index_hint
2007-07-23 18:09:48 +02:00
Index_hint *hint;
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
int type;
bool have_empty_use_join= FALSE, have_empty_use_order= FALSE,
have_empty_use_group= FALSE;
coding style fix : Index_hint
2007-07-23 18:09:48 +02:00
List_iterator <Index_hint> iter(*index_hints);
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
/* initialize temporary variables used to collect hints of each kind */
for (type= INDEX_HINT_IGNORE; type <= INDEX_HINT_FORCE; type++)
{
index_join[type].clear_all();
index_order[type].clear_all();
index_group[type].clear_all();
}
/* iterate over the hints list */
while ((hint= iter++))
{
uint pos;
/* process empty USE INDEX () */
if (hint->type == INDEX_HINT_USE && !hint->key_name.str)
{
if (hint->clause & INDEX_HINT_MASK_JOIN)
{
index_join[hint->type].clear_all();
have_empty_use_join= TRUE;
}
if (hint->clause & INDEX_HINT_MASK_ORDER)
{
index_order[hint->type].clear_all();
have_empty_use_order= TRUE;
}
if (hint->clause & INDEX_HINT_MASK_GROUP)
{
index_group[hint->type].clear_all();
have_empty_use_group= TRUE;
}
continue;
}
/*
Check if an index with the given name exists and get his offset in
the keys bitmask for the table
*/
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
if (tbl->s->keynames.type_names == 0 ||
(pos= find_type(&tbl->s->keynames, hint->key_name.str,
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
hint->key_name.length, 1)) <= 0)
{
my_error(ER_KEY_DOES_NOT_EXITS, MYF(0), hint->key_name.str, alias);
return 1;
}
pos--;
/* add to the appropriate clause mask */
if (hint->clause & INDEX_HINT_MASK_JOIN)
index_join[hint->type].set_bit (pos);
if (hint->clause & INDEX_HINT_MASK_ORDER)
index_order[hint->type].set_bit (pos);
if (hint->clause & INDEX_HINT_MASK_GROUP)
index_group[hint->type].set_bit (pos);
}
/* cannot mix USE INDEX and FORCE INDEX */
if ((!index_join[INDEX_HINT_FORCE].is_clear_all() ||
!index_order[INDEX_HINT_FORCE].is_clear_all() ||
!index_group[INDEX_HINT_FORCE].is_clear_all()) &&
(!index_join[INDEX_HINT_USE].is_clear_all() || have_empty_use_join ||
!index_order[INDEX_HINT_USE].is_clear_all() || have_empty_use_order ||
!index_group[INDEX_HINT_USE].is_clear_all() || have_empty_use_group))
{
my_error(ER_WRONG_USAGE, MYF(0), index_hint_type_name[INDEX_HINT_USE],
index_hint_type_name[INDEX_HINT_FORCE]);
return 1;
}
/* process FORCE INDEX as USE INDEX with a flag */
Bug #43029: FORCE INDEX FOR ORDER BY is ignored when join buffering is used FORCE INDEX FOR ORDER BY now prevents the optimizer from using join buffering. As a result the optimizer can use indexed access on the first table and doesn't need to sort the complete resultset at the end of the statement.
2009-10-07 17:03:42 +02:00
if (!index_order[INDEX_HINT_FORCE].is_clear_all())
{
tbl->force_index_order= TRUE;
index_order[INDEX_HINT_USE].merge(index_order[INDEX_HINT_FORCE]);
}
if (!index_group[INDEX_HINT_FORCE].is_clear_all())
{
tbl->force_index_group= TRUE;
index_group[INDEX_HINT_USE].merge(index_group[INDEX_HINT_FORCE]);
}
/*
TODO: get rid of tbl->force_index (on if any FORCE INDEX is specified) and
create tbl->force_index_join instead.
Then use the correct force_index_XX instead of the global one.
*/
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
if (!index_join[INDEX_HINT_FORCE].is_clear_all() ||
Bug #43029: FORCE INDEX FOR ORDER BY is ignored when join buffering is used FORCE INDEX FOR ORDER BY now prevents the optimizer from using join buffering. As a result the optimizer can use indexed access on the first table and doesn't need to sort the complete resultset at the end of the statement.
2009-10-07 17:03:42 +02:00
tbl->force_index_group || tbl->force_index_order)
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
{
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->force_index= TRUE;
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
index_join[INDEX_HINT_USE].merge(index_join[INDEX_HINT_FORCE]);
}
/* apply USE INDEX */
if (!index_join[INDEX_HINT_USE].is_clear_all() || have_empty_use_join)
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->keys_in_use_for_query.intersect(index_join[INDEX_HINT_USE]);
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
if (!index_order[INDEX_HINT_USE].is_clear_all() || have_empty_use_order)
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->keys_in_use_for_order_by.intersect (index_order[INDEX_HINT_USE]);
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
if (!index_group[INDEX_HINT_USE].is_clear_all() || have_empty_use_group)
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->keys_in_use_for_group_by.intersect (index_group[INDEX_HINT_USE]);
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
/* apply IGNORE INDEX */
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->keys_in_use_for_query.subtract (index_join[INDEX_HINT_IGNORE]);
tbl->keys_in_use_for_order_by.subtract (index_order[INDEX_HINT_IGNORE]);
tbl->keys_in_use_for_group_by.subtract (index_group[INDEX_HINT_IGNORE]);
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
}
/* make sure covering_keys don't include indexes disabled with a hint */
Fixed a lot of compiler warnings and errors detected by Forte C++ on Solaris Faster thr_alarm() Added 'Opened_files' status variable to track calls to my_open() Don't give warnings when running mysql_install_db Added option --source-install to mysql_install_db I had to do the following renames() as used polymorphism didn't work with Forte compiler on 64 bit systems index_read() -> index_read_map() index_read_idx() -> index_read_idx_map() index_read_last() -> index_read_last_map()
2007-08-13 15:11:25 +02:00
tbl->covering_keys.intersect(tbl->keys_in_use_for_query);
WL#3527: Extend IGNORE INDEX so places where index is ignored can be specified Currently MySQL allows one to specify what indexes to ignore during join optimization. The scope of the current USE/FORCE/IGNORE INDEX statement is only the FROM clause, while all other clauses are not affected. However, in certain cases, the optimizer may incorrectly choose an index for sorting and/or grouping, and produce an inefficient query plan. This task provides the means to specify what indexes are ignored/used for what operation in a more fine-grained manner, thus making it possible to manually force a better plan. We do this by extending the current IGNORE/USE/FORCE INDEX syntax to: IGNORE/USE/FORCE INDEX [FOR {JOIN | ORDER | GROUP BY}] so that: - if no FOR is specified, the index hint will apply everywhere. - if MySQL is started with the compatibility option --old_mode then an index hint without a FOR clause works as in 5.0 (i.e, the index will only be ignored for JOINs, but can still be used to compute ORDER BY). See the WL#3527 for further details.
2007-03-05 18:08:41 +01:00
return 0;
}
WL#3817: Simplify string / memory area types and make things more consistent (first part) The following type conversions was done: - Changed byte to uchar - Changed gptr to uchar* - Change my_string to char * - Change my_size_t to size_t - Change size_s to size_t Removed declaration of byte, gptr, my_string, my_size_t and size_s. Following function parameter changes was done: - All string functions in mysys/strings was changed to use size_t instead of uint for string lengths. - All read()/write() functions changed to use size_t (including vio). - All protocoll functions changed to use size_t instead of uint - Functions that used a pointer to a string length was changed to use size_t* - Changed malloc(), free() and related functions from using gptr to use void * as this requires fewer casts in the code and is more in line with how the standard functions work. - Added extra length argument to dirname_part() to return the length of the created string. - Changed (at least) following functions to take uchar* as argument: - db_dump() - my_net_write() - net_write_command() - net_store_data() - DBUG_DUMP() - decimal2bin() & bin2decimal() - Changed my_compress() and my_uncompress() to use size_t. Changed one argument to my_uncompress() from a pointer to a value as we only return one value (makes function easier to use). - Changed type of 'pack_data' argument to packfrm() to avoid casts. - Changed in readfrm() and writefrom(), ha_discover and handler::discover() the type for argument 'frmdata' to uchar** to avoid casts. - Changed most Field functions to use uchar* instead of char* (reduced a lot of casts). - Changed field->val_xxx(xxx, new_ptr) to take const pointers. Other changes: - Removed a lot of not needed casts - Added a few new cast required by other changes - Added some cast to my_multi_malloc() arguments for safety (as string lengths needs to be uint, not size_t). - Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done explicitely as this conflict was often hided by casting the function to hash_get_key). - Changed some buffers to memory regions to uchar* to avoid casts. - Changed some string lengths from uint to size_t. - Changed field->ptr to be uchar* instead of char*. This allowed us to get rid of a lot of casts. - Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar - Include zlib.h in some files as we needed declaration of crc32() - Changed MY_FILE_ERROR to be (size_t) -1. - Changed many variables to hold the result of my_read() / my_write() to be size_t. This was needed to properly detect errors (which are returned as (size_t) -1). - Removed some very old VMS code - Changed packfrm()/unpackfrm() to not be depending on uint size (portability fix) - Removed windows specific code to restore cursor position as this causes slowdown on windows and we should not mix read() and pread() calls anyway as this is not thread safe. Updated function comment to reflect this. Changed function that depended on original behavior of my_pwrite() to itself restore the cursor position (one such case). - Added some missing checking of return value of malloc(). - Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow. - Changed type of table_def::m_size from my_size_t to ulong to reflect that m_size is the number of elements in the array, not a string/memory length. - Moved THD::max_row_length() to table.cc (as it's not depending on THD). Inlined max_row_length_blob() into this function. - More function comments - Fixed some compiler warnings when compiled without partitions. - Removed setting of LEX_STRING() arguments in declaration (portability fix). - Some trivial indentation/variable name changes. - Some trivial code simplifications: - Replaced some calls to alloc_root + memcpy to use strmake_root()/strdup_root(). - Changed some calls from memdup() to strmake() (Safety fix) - Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
size_t max_row_length(TABLE *table, const uchar *data)
{
TABLE_SHARE *table_s= table->s;
size_t length= table_s->reclength + 2 * table_s->fields;
uint *const beg= table_s->blob_field;
uint *const end= beg + table_s->blob_fields;
for (uint *ptr= beg ; ptr != end ; ++ptr)
{
Field_blob* const blob= (Field_blob*) table->field[*ptr];
length+= blob->get_length((const uchar*)
(data + blob->offset(table->record[0]))) +
HA_KEY_BLOB_LENGTH;
}
return length;
}
Initial import of WL#3726 "DDL locking for all metadata objects". Backport of: ------------------------------------------------------------ revno: 2630.4.1 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Fri 2008-05-23 17:54:03 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ------------------------------------------------------------ This is the first patch in series. It transforms the metadata locking subsystem to use a dedicated module (mdl.h,cc). No significant changes in the locking protocol. The import passes the test suite with the exception of deprecated/removed 6.0 features, and MERGE tables. The latter are subject to a fix by WL#4144. Unfortunately, the original changeset comments got lost in a merge, thus this import has its own (largely insufficient) comments. This patch fixes Bug#25144 "replication / binlog with view breaks". Warning: this patch introduces an incompatible change: Under LOCK TABLES, it's no longer possible to FLUSH a table that was not locked for WRITE. Under LOCK TABLES, it's no longer possible to DROP a table or VIEW that was not locked for WRITE. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.2 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:03:45 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.3 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:08:51 +0400 message: WL#3726 "DDL locking for all metadata objects" Fixed failing Windows builds by adding mdl.cc to the lists of files needed to build server/libmysqld on Windows. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.4 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 21:57:58 +0400 message: WL#3726 "DDL locking for all metadata objects". Fix for assert failures in kill.test which occured when one tried to kill ALTER TABLE statement on merge table while it was waiting in wait_while_table_is_used() for other connections to close this table. These assert failures stemmed from the fact that cleanup code in this case assumed that temporary table representing new version of table was open with adding to THD::temporary_tables list while code which were opening this temporary table wasn't always fulfilling this. This patch changes code that opens new version of table to always do this linking in. It also streamlines cleanup process for cases when error occurs while we have new version of table open. ****** WL#3726 "DDL locking for all metadata objects" Add libmysqld/mdl.cc to .bzrignore. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.6 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sun 2008-05-25 00:33:22 +0400 message: WL#3726 "DDL locking for all metadata objects". Addition to the fix of assert failures in kill.test caused by changes for this worklog. Make sure we close the new table only once.
2009-11-30 16:55:03 +01:00
/**
Helper function which allows to allocate metadata lock request
objects for all elements of table list.
*/
Backport of: ---------------------------------------------------------- revno: 2617.69.20 committer: Konstantin Osipov <kostja@sun.com> branch nick: 5.4-4284-1-assert timestamp: Thu 2009-08-13 18:29:55 +0400 message: WL#4284 "Transactional DDL locking" A review fix. Since WL#4284 implementation separated MDL_request and MDL_ticket, MDL_request becamse a utility object necessary only to get a ticket. Store it by-value in TABLE_LIST with the intent to merge MDL_request::key with table_list->table_name and table_list->db in future. Change the MDL subsystem to not require MDL_requests to stay around till close_thread_tables(). Remove the list of requests from the MDL context. Requests for shared metadata locks acquired in open_tables() are only used as a list in recover_from_failed_open_table_attempt(), which calls mdl_context.wait_for_locks() for this list. To keep such list for recover_from_failed_open_table_attempt(), introduce a context class (Open_table_context), that collects all requests. A lot of minor cleanups and simplications that became possible with this change.
2009-12-08 10:57:07 +01:00
void init_mdl_requests(TABLE_LIST *table_list)
Initial import of WL#3726 "DDL locking for all metadata objects". Backport of: ------------------------------------------------------------ revno: 2630.4.1 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Fri 2008-05-23 17:54:03 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ------------------------------------------------------------ This is the first patch in series. It transforms the metadata locking subsystem to use a dedicated module (mdl.h,cc). No significant changes in the locking protocol. The import passes the test suite with the exception of deprecated/removed 6.0 features, and MERGE tables. The latter are subject to a fix by WL#4144. Unfortunately, the original changeset comments got lost in a merge, thus this import has its own (largely insufficient) comments. This patch fixes Bug#25144 "replication / binlog with view breaks". Warning: this patch introduces an incompatible change: Under LOCK TABLES, it's no longer possible to FLUSH a table that was not locked for WRITE. Under LOCK TABLES, it's no longer possible to DROP a table or VIEW that was not locked for WRITE. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.2 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:03:45 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.3 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:08:51 +0400 message: WL#3726 "DDL locking for all metadata objects" Fixed failing Windows builds by adding mdl.cc to the lists of files needed to build server/libmysqld on Windows. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.4 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 21:57:58 +0400 message: WL#3726 "DDL locking for all metadata objects". Fix for assert failures in kill.test which occured when one tried to kill ALTER TABLE statement on merge table while it was waiting in wait_while_table_is_used() for other connections to close this table. These assert failures stemmed from the fact that cleanup code in this case assumed that temporary table representing new version of table was open with adding to THD::temporary_tables list while code which were opening this temporary table wasn't always fulfilling this. This patch changes code that opens new version of table to always do this linking in. It also streamlines cleanup process for cases when error occurs while we have new version of table open. ****** WL#3726 "DDL locking for all metadata objects" Add libmysqld/mdl.cc to .bzrignore. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.6 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sun 2008-05-25 00:33:22 +0400 message: WL#3726 "DDL locking for all metadata objects". Addition to the fix of assert failures in kill.test caused by changes for this worklog. Make sure we close the new table only once.
2009-11-30 16:55:03 +01:00
{
for ( ; table_list ; table_list= table_list->next_global)
Backport of: ------------------------------------------------------------ revno: 2617.68.25 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-next-bg-pre2-2 timestamp: Wed 2009-09-16 18:26:50 +0400 message: Follow-up for one of pre-requisite patches for fixing bug #30977 "Concurrent statement using stored function and DROP FUNCTION breaks SBR". Made enum_mdl_namespace enum part of MDL_key class and removed MDL_ prefix from the names of enum members. In order to do the latter changed name of PROCEDURE symbol to PROCEDURE_SYM (otherwise macro which was automatically generated for this symbol conflicted with MDL_key::PROCEDURE enum member).
2009-12-10 09:21:38 +01:00
table_list->mdl_request.init(MDL_key::TABLE,
table_list->db, table_list->table_name,
Implement new type-of-operation-aware metadata locks. Add a wait-for graph based deadlock detector to the MDL subsystem. Fixes bug #46272 "MySQL 5.4.4, new MDL: unnecessary deadlock" and bug #37346 "innodb does not detect deadlock between update and alter table". The first bug manifested itself as an unwarranted abort of a transaction with ER_LOCK_DEADLOCK error by a concurrent ALTER statement, when this transaction tried to repeat use of a table, which it has already used in a similar fashion before ALTER started. The second bug showed up as a deadlock between table-level locks and InnoDB row locks, which was "detected" only after innodb_lock_wait_timeout timeout. A transaction would start using the table and modify a few rows. Then ALTER TABLE would come in, and start copying rows into a temporary table. Eventually it would stumble on the modified records and get blocked on a row lock. The first transaction would try to do more updates, and get blocked on thr_lock.c lock. This situation of circular wait would only get resolved by a timeout. Both these bugs stemmed from inadequate solutions to the problem of deadlocks occurring between different locking subsystems. In the first case we tried to avoid deadlocks between metadata locking and table-level locking subsystems, when upgrading shared metadata lock to exclusive one. Transactions holding the shared lock on the table and waiting for some table-level lock used to be aborted too aggressively. We also allowed ALTER TABLE to start in presence of transactions that modify the subject table. ALTER TABLE acquires TL_WRITE_ALLOW_READ lock at start, and that block all writes against the table (naturally, we don't want any writes to be lost when switching the old and the new table). TL_WRITE_ALLOW_READ lock, in turn, would block the started transaction on thr_lock.c lock, should they do more updates. This, again, lead to the need to abort such transactions. The second bug occurred simply because we didn't have any mechanism to detect deadlocks between the table-level locks in thr_lock.c and row-level locks in InnoDB, other than innodb_lock_wait_timeout. This patch solves both these problems by moving lock conflicts which are causing these deadlocks into the metadata locking subsystem, thus making it possible to avoid or detect such deadlocks inside MDL. To do this we introduce new type-of-operation-aware metadata locks, which allow MDL subsystem to know not only the fact that transaction has used or is going to use some object but also what kind of operation it has carried out or going to carry out on the object. This, along with the addition of a special kind of upgradable metadata lock, allows ALTER TABLE to wait until all transactions which has updated the table to go away. This solves the second issue. Another special type of upgradable metadata lock is acquired by LOCK TABLE WRITE. This second lock type allows to solve the first issue, since abortion of table-level locks in event of DDL under LOCK TABLES becomes also unnecessary. Below follows the list of incompatible changes introduced by this patch: - From now on, ALTER TABLE and CREATE/DROP TRIGGER SQL (i.e. those statements that acquire TL_WRITE_ALLOW_READ lock) wait for all transactions which has *updated* the table to complete. - From now on, LOCK TABLES ... WRITE, REPAIR/OPTIMIZE TABLE (i.e. all statements which acquire TL_WRITE table-level lock) wait for all transaction which *updated or read* from the table to complete. As a consequence, innodb_table_locks=0 option no longer applies to LOCK TABLES ... WRITE. - DROP DATABASE, DROP TABLE, RENAME TABLE no longer abort statements or transactions which use tables being dropped or renamed, and instead wait for these transactions to complete. - Since LOCK TABLES WRITE now takes a special metadata lock, not compatible with with reads or writes against the subject table and transaction-wide, thr_lock.c deadlock avoidance algorithm that used to ensure absence of deadlocks between LOCK TABLES WRITE and other statements is no longer sufficient, even for MyISAM. The wait-for graph based deadlock detector of MDL subsystem may sometimes be necessary and is involved. This may lead to ER_LOCK_DEADLOCK error produced for multi-statement transactions even if these only use MyISAM: session 1: session 2: begin; update t1 ... lock table t2 write, t1 write; -- gets a lock on t2, blocks on t1 update t2 ... (ER_LOCK_DEADLOCK) - Finally, support of LOW_PRIORITY option for LOCK TABLES ... WRITE was abandoned. LOCK TABLE ... LOW_PRIORITY WRITE from now on has the same priority as the usual LOCK TABLE ... WRITE. SELECT HIGH PRIORITY no longer trumps LOCK TABLE ... WRITE in the wait queue. - We do not take upgradable metadata locks on implicitly locked tables. So if one has, say, a view v1 that uses table t1, and issues: LOCK TABLE v1 WRITE; FLUSH TABLE t1; -- (or just 'FLUSH TABLES'), an error is produced. In order to be able to perform DDL on a table under LOCK TABLES, the table must be locked explicitly in the LOCK TABLES list.
2010-02-01 12:43:06 +01:00
table_list->lock_type >= TL_WRITE_ALLOW_WRITE ?
MDL_SHARED_WRITE : MDL_SHARED_READ);
Initial import of WL#3726 "DDL locking for all metadata objects". Backport of: ------------------------------------------------------------ revno: 2630.4.1 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Fri 2008-05-23 17:54:03 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ------------------------------------------------------------ This is the first patch in series. It transforms the metadata locking subsystem to use a dedicated module (mdl.h,cc). No significant changes in the locking protocol. The import passes the test suite with the exception of deprecated/removed 6.0 features, and MERGE tables. The latter are subject to a fix by WL#4144. Unfortunately, the original changeset comments got lost in a merge, thus this import has its own (largely insufficient) comments. This patch fixes Bug#25144 "replication / binlog with view breaks". Warning: this patch introduces an incompatible change: Under LOCK TABLES, it's no longer possible to FLUSH a table that was not locked for WRITE. Under LOCK TABLES, it's no longer possible to DROP a table or VIEW that was not locked for WRITE. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.2 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:03:45 +0400 message: WL#3726 "DDL locking for all metadata objects". After review fixes in progress. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.3 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 14:08:51 +0400 message: WL#3726 "DDL locking for all metadata objects" Fixed failing Windows builds by adding mdl.cc to the lists of files needed to build server/libmysqld on Windows. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.4 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sat 2008-05-24 21:57:58 +0400 message: WL#3726 "DDL locking for all metadata objects". Fix for assert failures in kill.test which occured when one tried to kill ALTER TABLE statement on merge table while it was waiting in wait_while_table_is_used() for other connections to close this table. These assert failures stemmed from the fact that cleanup code in this case assumed that temporary table representing new version of table was open with adding to THD::temporary_tables list while code which were opening this temporary table wasn't always fulfilling this. This patch changes code that opens new version of table to always do this linking in. It also streamlines cleanup process for cases when error occurs while we have new version of table open. ****** WL#3726 "DDL locking for all metadata objects" Add libmysqld/mdl.cc to .bzrignore. ****** Backport of: ------------------------------------------------------------ revno: 2630.4.6 committer: Dmitry Lenev <dlenev@mysql.com> branch nick: mysql-6.0-3726-w timestamp: Sun 2008-05-25 00:33:22 +0400 message: WL#3726 "DDL locking for all metadata objects". Addition to the fix of assert failures in kill.test caused by changes for this worklog. Make sure we close the new table only once.
2009-11-30 16:55:03 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
/*****************************************************************************
** Instansiate templates
*****************************************************************************/
WL#2286 - Compile MySQL w/YASSL support Fix GCC 4.0 link failure. Better CXX_VERSION guessing.
2005-06-22 11:08:28 +02:00
#ifdef HAVE_EXPLICIT_TEMPLATE_INSTANTIATION
Import changeset
2000-07-31 21:29:14 +02:00
template class List<String>;
template class List_iterator<String>;
#endif
Reference in a new issue Copy permalink