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/handler.cc

4889 lines
141 KiB
C++
Raw Normal View History

Added "Sun Microsystems, Inc." to copyright headers on files modified since Oct 1st
2008-11-10 21:21:49 +01:00
/* Copyright 2000-2008 MySQL AB, 2008 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 */
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @file handler.cc
Import changeset
2000-07-31 21:29:14 +02:00
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
@brief
Handler-calling-functions
*/
Import changeset
2000-07-31 21:29:14 +02:00
Add ifdefs to control when "#pragma implementation" should be used Added some more ifdefs for "#pragma interface"
2005-05-26 12:09:14 +02:00
#ifdef USE_PRAGMA_IMPLEMENTATION
Import changeset
2000-07-31 21:29:14 +02:00
#pragma implementation // gcc: Class implementation
#endif
#include "mysql_priv.h"
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
#include "rpl_filter.h"
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
#include <myisampack.h>
#include <errno.h>
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
#ifdef WITH_PARTITION_STORAGE_ENGINE
#include "ha_partition.h"
First commit to mysql-5.0 tree to include MySQL Federated Storage Handler. This includes both the source and header files, test (results, test, require), and modifications to server and handler base files, and autoconf modifications to properly build federated handler.
2004-12-11 21:03:51 +01:00
#endif
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
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
/*
While we have legacy_db_type, we have this array to
check for dups and to find handlerton from legacy_db_type.
Remove when legacy_db_type is finally gone
*/
after review fixes
2006-05-31 18:07:32 +02:00
st_plugin_int *hton2plugin[MAX_HA];
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
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
static handlerton *installed_htons[128];
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
After merge fixes
2006-06-16 02:52:17 +02:00
#define BITMAP_STACKBUF_SIZE (128/8)
Cleanups after review of WL#602 Fixed warnings from test suite Some fixes in mysql-test-run script to catch more warnings
2006-05-03 18:40:52 +02:00
KEY_CREATE_INFO default_key_create_info= { HA_KEY_ALG_UNDEF, 0, {NullS,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
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +01:00
/* number of entries in handlertons[] */
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
ulong total_ha= 0;
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +01:00
/* number of storage engines (from handlertons[]) that support 2pc */
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
ulong total_ha_2pc= 0;
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +01:00
/* size of savepoint storage area (see ha_init) */
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
ulong savepoint_alloc_size= 0;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
static const LEX_STRING sys_table_aliases[]=
Cleanup for handlerton structure to allow for loadable engine work. This is the first in a series of patches.
2005-09-19 21:06:23 +02:00
{
Cleanup patch. There is an existing macros for initializing LEX_STRINGs with constant strings -> C_STRING_WITH_LEN. Change existing code to use it. (char *) STRING_WITH_LEN -> C_STRING_WITH_LEN
2006-08-17 18:13:45 +02:00
{ C_STRING_WITH_LEN("INNOBASE") }, { C_STRING_WITH_LEN("INNODB") },
{ C_STRING_WITH_LEN("NDB") }, { C_STRING_WITH_LEN("NDBCLUSTER") },
{ C_STRING_WITH_LEN("HEAP") }, { C_STRING_WITH_LEN("MEMORY") },
Merge pchardin@bk-internal.mysql.com:/home/bk/mysql-5.1 into mysql.com:/home/cps/mysql/trees/mysql-5.1-virgin
2006-08-23 14:48:39 +02:00
{ C_STRING_WITH_LEN("MERGE") }, { C_STRING_WITH_LEN("MRG_MYISAM") },
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
{NullS, 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
};
Test case for bug when updating on key Fixed bug in CONCAT_WS() Print the default ISOLATION level. Change lock type for CREATE ... SELECT and INSERT/REPLACE ... SELECT
2001-03-23 19:38:42 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
const char *ha_row_type[] = {
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
"", "FIXED", "DYNAMIC", "COMPRESSED", "REDUNDANT", "COMPACT",
/* Reserved to be "PAGE" in future versions */ "?",
"?","?","?"
Import changeset
2000-07-31 21:29:14 +02:00
};
Added isolation levels Added SELECT .. FOR UPDATE and SELECT .. IN SHARE MODE
2001-03-21 00:02:22 +01:00
const char *tx_isolation_names[] =
Test case for bug when updating on key Fixed bug in CONCAT_WS() Print the default ISOLATION level. Change lock type for CREATE ... SELECT and INSERT/REPLACE ... SELECT
2001-03-23 19:38:42 +01:00
{ "READ-UNCOMMITTED", "READ-COMMITTED", "REPEATABLE-READ", "SERIALIZABLE",
NullS};
TYPELIB tx_isolation_typelib= {array_elements(tx_isolation_names)-1,"",
Bug#6202: ENUMs are not case sensitive even if declared BINARY
2004-10-25 14:51:26 +02:00
tx_isolation_names, NULL};
Import changeset
2000-07-31 21:29:14 +02:00
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
static TYPELIB known_extensions= {0,"known_exts", NULL, NULL};
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
uint known_extensions_id= 0;
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
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
static plugin_ref ha_default_plugin(THD *thd)
{
if (thd->variables.table_plugin)
return thd->variables.table_plugin;
return my_plugin_lock(thd, &global_system_variables.table_plugin);
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
Return the default storage engine handlerton for thread
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
SYNOPSIS
ha_default_handlerton(thd)
thd current thread
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
RETURN
pointer to handlerton
*/
handlerton *ha_default_handlerton(THD *thd)
WorkLog#1280 - Remove fixed table handler from lex/yacc
2003-12-02 21:23:13 +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_ref plugin= ha_default_plugin(thd);
DBUG_ASSERT(plugin);
handlerton *hton= plugin_data(plugin, handlerton*);
DBUG_ASSERT(hton);
return hton;
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
Return the storage engine handlerton for the supplied name
SYNOPSIS
ha_resolve_by_name(thd, name)
thd current thread
name name of storage engine
RETURN
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
pointer to storage engine plugin handle
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
*/
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_ref ha_resolve_by_name(THD *thd, const LEX_STRING *name)
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
{
const LEX_STRING *table_alias;
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 plugin;
Cleanup for handlerton structure to allow for loadable engine work. This is the first in a series of patches.
2005-09-19 21:06:23 +02:00
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
redo:
/* my_strnncoll is a macro and gcc doesn't do early expansion of macro */
if (thd && !my_charset_latin1.coll->strnncoll(&my_charset_latin1,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
(const uchar *)name->str, name->length,
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
(const uchar *)STRING_WITH_LEN("DEFAULT"), 0))
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
return ha_default_plugin(thd);
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +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
if ((plugin= my_plugin_lock_by_name(thd, name, MYSQL_STORAGE_ENGINE_PLUGIN)))
WorkLog#1280 - Remove fixed table handler from lex/yacc
2003-12-02 21:23:13 +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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (!(hton->flags & HTON_NOT_USER_SELECTABLE))
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
return plugin;
/*
unlocking plugin immediately after locking is relatively low cost.
*/
plugin_unlock(thd, plugin);
WorkLog#1280 - Remove fixed table handler from lex/yacc
2003-12-02 21:23:13 +01:00
}
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +01:00
Cleanup for handlerton structure to allow for loadable engine work. This is the first in a series of patches.
2005-09-19 21:06:23 +02:00
/*
This fixed bug 13369. Note, I dislike goto's as well. The alias structure now is a bit more simple and just uses a pointer to replace with the currect name. The giant case statement should go away in the next patch.
2005-09-23 01:58:09 +02:00
We check for the historical aliases.
Cleanup for handlerton structure to allow for loadable engine work. This is the first in a series of patches.
2005-09-19 21:06:23 +02:00
*/
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
for (table_alias= sys_table_aliases; table_alias->str; table_alias+= 2)
WorkLog#1280 - Remove fixed table handler from lex/yacc
2003-12-02 21:23:13 +01:00
{
Store and read engine type string in extra block of .frm.
2005-10-10 20:01:45 +02:00
if (!my_strnncoll(&my_charset_latin1,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
(const uchar *)name->str, name->length,
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
(const uchar *)table_alias->str, table_alias->length))
{
name= table_alias + 1;
goto redo;
}
WorkLog#1280 - Remove fixed table handler from lex/yacc
2003-12-02 21:23:13 +01:00
}
This fixed bug 13369. Note, I dislike goto's as well. The alias structure now is a bit more simple and just uses a pointer to replace with the currect name. The giant case statement should go away in the next patch.
2005-09-23 01:58:09 +02:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return NULL;
WorkLog#1280 - Remove fixed table handler from lex/yacc
2003-12-02 21:23:13 +01:00
}
Reviewing new pushed code - CHAR() now returns binary string as default - CHAR(X*65536+Y*256+Z) is now equal to CHAR(X,Y,Z) independent of the character set for CHAR() - Test for both ETIMEDOUT and ETIME from pthread_cond_timedwait() (Some old systems returns ETIME and it's safer to test for both values than to try to write a wrapper for each old system) - Fixed new introduced bug in NOT BETWEEN X and X - Ensure we call commit_by_xid or rollback_by_xid for all engines, even if one engine has failed - Use octet2hex() for all conversion of string to hex - Simplify and optimize code
2005-10-11 23:58:22 +02:00
BUG#39053 - UNISTALL PLUGIN does not allow the storage engine to cleanup open connections It was possible to UNINSTALL storage engine plugin when binding between THD object and storage engine is still active (e.g. in the middle of transaction). To avoid unclean deactivation (uninstall) of storage engine plugin in the middle of transaction, additional storage engine plugin lock is acquired by thd_set_ha_data(). If ha_data is not null and storage engine plugin was not locked by thd_set_ha_data() in this connection before, storage engine plugin gets locked. If ha_data is null and storage engine plugin was locked by thd_set_ha_data() in this connection before, storage engine plugin lock gets released. If handlerton::close_connection() didn't reset ha_data, server does it immediately after calling handlerton::close_connection(). Note that this is just a framework fix, storage engines must switch to thd_set_ha_data() from thd_ha_data() if they want to see fit.
2010-04-14 11:53:59 +02:00
plugin_ref ha_lock_engine(THD *thd, const handlerton *hton)
Move handler specific options into handlerton flag check BUG#13108
2005-10-01 01:26:48 +02:00
{
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
if (hton)
{
st_plugin_int **plugin= hton2plugin + hton->slot;
WL#2936 "Plugin Server Variables" Post review cleanups.
2007-03-29 06:29:16 +02:00
#ifdef DBUG_OFF
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
return my_plugin_lock(thd, plugin);
#else
return my_plugin_lock(thd, &plugin);
#endif
Move handler specific options into handlerton flag check BUG#13108
2005-10-01 01:26:48 +02:00
}
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
return NULL;
Move handler specific options into handlerton flag check BUG#13108
2005-10-01 01:26:48 +02:00
}
Import changeset
2000-07-31 21:29:14 +02:00
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
#ifdef NOT_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
static handler *create_default(TABLE_SHARE *table, MEM_ROOT *mem_root)
Import changeset
2000-07-31 21:29:14 +02:00
{
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
handlerton *hton= ha_default_handlerton(current_thd);
This removes the passing of global hton to engine instance.
2006-09-30 02:19:02 +02:00
return (hton && hton->create) ? hton->create(hton, table, mem_root) : NULL;
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
}
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
#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
handlerton *ha_resolve_by_legacy_type(THD *thd, enum legacy_db_type db_type)
{
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 plugin;
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
switch (db_type) {
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
case DB_TYPE_DEFAULT:
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
return ha_default_handlerton(thd);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
default:
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 (db_type > DB_TYPE_UNKNOWN && db_type < DB_TYPE_DEFAULT &&
(plugin= ha_lock_engine(thd, installed_htons[db_type])))
return plugin_data(plugin, handlerton*);
/* fall through */
case DB_TYPE_UNKNOWN:
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
return NULL;
Cleanup ha_checktype() and redundant variables.
2004-03-02 11:08:50 +01:00
}
Check that the default storage engine is really available, and refuse to start up if it is not. (Bug #9815)
2005-04-14 03:25:31 +02:00
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Use other database handler if databasehandler is not compiled in.
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
handlerton *ha_checktype(THD *thd, enum legacy_db_type database_type,
Bug#6877 MySQL should give an error if the requested table type is not available Implement new SQL mode - NO_ENGINE_SUBSTITUTION
2005-06-17 23:14:44 +02:00
bool no_substitute, bool report_error)
Import changeset
2000-07-31 21:29:14 +02:00
{
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
handlerton *hton= ha_resolve_by_legacy_type(thd, database_type);
if (ha_storage_engine_is_enabled(hton))
return hton;
Bug#6877 MySQL should give an error if the requested table type is not available Implement new SQL mode - NO_ENGINE_SUBSTITUTION
2005-06-17 23:14:44 +02:00
if (no_substitute)
{
if (report_error)
{
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
const char *engine_name= ha_resolve_storage_engine_name(hton);
Bug#6877 MySQL should give an error if the requested table type is not available Implement new SQL mode - NO_ENGINE_SUBSTITUTION
2005-06-17 23:14:44 +02:00
my_error(ER_FEATURE_DISABLED,MYF(0),engine_name,engine_name);
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return NULL;
Bug#6877 MySQL should give an error if the requested table type is not available Implement new SQL mode - NO_ENGINE_SUBSTITUTION
2005-06-17 23:14:44 +02:00
}
Import changeset
2000-07-31 21:29:14 +02:00
switch (database_type) {
#ifndef NO_HASH
case DB_TYPE_HASH:
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return ha_resolve_by_legacy_type(thd, DB_TYPE_HASH);
Cleanup ha_checktype() and redundant variables.
2004-03-02 11:08:50 +01:00
#endif
Give warning if MySQL doesn't honor given storage engine Allow syntax CREATE TABLE t1 (LIKE t2)
2003-05-13 10:15:11 +02:00
case DB_TYPE_MRG_ISAM:
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return ha_resolve_by_legacy_type(thd, DB_TYPE_MRG_MYISAM);
Import changeset
2000-07-31 21:29:14 +02:00
default:
break;
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
Bug#20168 "Change in behavior --default-storage-engine=ndb or ndbcluster" Reduce use of legacy_db_type, some code cleanup (serg read my mind and implemented desired mysqld.cc changes)
2006-06-12 15:50:11 +02:00
return ha_default_handlerton(thd);
Import changeset
2000-07-31 21:29:14 +02:00
} /* ha_checktype */
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 *get_new_handler(TABLE_SHARE *share, MEM_ROOT *alloc,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
handlerton *db_type)
Import changeset
2000-07-31 21:29:14 +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
handler *file;
DBUG_ENTER("get_new_handler");
DBUG_PRINT("enter", ("alloc: 0x%lx", (long) alloc));
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
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 (db_type && db_type->state == SHOW_OPTION_YES && db_type->create)
Print right hostname or IP in SHOW PROCESSLIST Use SESSION TABLE_HANDLER as default table handler if given table handler doesn't exists
2003-03-19 20:43:41 +01:00
{
This removes the passing of global hton to engine instance.
2006-09-30 02:19:02 +02:00
if ((file= db_type->create(db_type, share, alloc)))
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
file->init();
DBUG_RETURN(file);
Print right hostname or IP in SHOW PROCESSLIST Use SESSION TABLE_HANDLER as default table handler if given table handler doesn't exists
2003-03-19 20:43:41 +01:00
}
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
/*
Try the default table type
Here the call to current_thd() is ok as we call this function a lot of
times but we enter this branch very seldom.
*/
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
DBUG_RETURN(get_new_handler(share, alloc, ha_default_handlerton(current_thd)));
Import changeset
2000-07-31 21:29:14 +02:00
}
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
#ifdef WITH_PARTITION_STORAGE_ENGINE
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
handler *get_ha_partition(partition_info *part_info)
{
ha_partition *partition;
DBUG_ENTER("get_ha_partition");
This removes the passing of global hton to engine instance.
2006-09-30 02:19:02 +02:00
if ((partition= new ha_partition(partition_hton, part_info)))
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
{
Bug#39084: Getting intermittent errors with statement-based binary logging Problem was that partitioning cached the table flags. These flags could change due to TRANSACTION LEVEL changes. Solution was to remove the cache and always return the table flags from the first partition (if the handler was initialized).
2008-10-29 21:20:04 +01:00
if (partition->initialize_partition(current_thd->mem_root))
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
{
delete partition;
partition= 0;
}
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
else
partition->init();
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
}
else
{
my_error(ER_OUTOFMEMORY, MYF(0), sizeof(ha_partition));
}
DBUG_RETURN(((handler*) partition));
}
#endif
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
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
Register handler error messages for use with my_error().
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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
0 OK
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
!=0 Error
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
*/
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
int ha_init_errors(void)
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
{
#define SETMSG(nr, msg) errmsgs[(nr) - HA_ERR_FIRST]= (msg)
const char **errmsgs;
/* Allocate a pointer array for the error message strings. */
/* Zerofill it to avoid uninitialized gaps. */
if (! (errmsgs= (const char**) my_malloc(HA_ERR_ERRORS * sizeof(char*),
MYF(MY_WME | MY_ZEROFILL))))
return 1;
/* Set the dedicated error messages. */
SETMSG(HA_ERR_KEY_NOT_FOUND, ER(ER_KEY_NOT_FOUND));
SETMSG(HA_ERR_FOUND_DUPP_KEY, ER(ER_DUP_KEY));
SETMSG(HA_ERR_RECORD_CHANGED, "Update wich is recoverable");
SETMSG(HA_ERR_WRONG_INDEX, "Wrong index given to function");
SETMSG(HA_ERR_CRASHED, ER(ER_NOT_KEYFILE));
SETMSG(HA_ERR_WRONG_IN_RECORD, ER(ER_CRASHED_ON_USAGE));
SETMSG(HA_ERR_OUT_OF_MEM, "Table handler out of memory");
SETMSG(HA_ERR_NOT_A_TABLE, "Incorrect file format '%.64s'");
SETMSG(HA_ERR_WRONG_COMMAND, "Command not supported");
SETMSG(HA_ERR_OLD_FILE, ER(ER_OLD_KEYFILE));
SETMSG(HA_ERR_NO_ACTIVE_RECORD, "No record read in update");
SETMSG(HA_ERR_RECORD_DELETED, "Intern record deleted");
SETMSG(HA_ERR_RECORD_FILE_FULL, ER(ER_RECORD_FILE_FULL));
SETMSG(HA_ERR_INDEX_FILE_FULL, "No more room in index file '%.64s'");
SETMSG(HA_ERR_END_OF_FILE, "End in next/prev/first/last");
SETMSG(HA_ERR_UNSUPPORTED, ER(ER_ILLEGAL_HA));
SETMSG(HA_ERR_TO_BIG_ROW, "Too big row");
SETMSG(HA_WRONG_CREATE_OPTION, "Wrong create option");
SETMSG(HA_ERR_FOUND_DUPP_UNIQUE, ER(ER_DUP_UNIQUE));
SETMSG(HA_ERR_UNKNOWN_CHARSET, "Can't open charset");
SETMSG(HA_ERR_WRONG_MRG_TABLE_DEF, ER(ER_WRONG_MRG_TABLE));
SETMSG(HA_ERR_CRASHED_ON_REPAIR, ER(ER_CRASHED_ON_REPAIR));
SETMSG(HA_ERR_CRASHED_ON_USAGE, ER(ER_CRASHED_ON_USAGE));
SETMSG(HA_ERR_LOCK_WAIT_TIMEOUT, ER(ER_LOCK_WAIT_TIMEOUT));
SETMSG(HA_ERR_LOCK_TABLE_FULL, ER(ER_LOCK_TABLE_FULL));
SETMSG(HA_ERR_READ_ONLY_TRANSACTION, ER(ER_READ_ONLY_TRANSACTION));
SETMSG(HA_ERR_LOCK_DEADLOCK, ER(ER_LOCK_DEADLOCK));
SETMSG(HA_ERR_CANNOT_ADD_FOREIGN, ER(ER_CANNOT_ADD_FOREIGN));
Fix bug #3443, better foreign key error messsages.
2005-09-23 15:22:27 +02:00
SETMSG(HA_ERR_NO_REFERENCED_ROW, ER(ER_NO_REFERENCED_ROW_2));
SETMSG(HA_ERR_ROW_IS_REFERENCED, ER(ER_ROW_IS_REFERENCED_2));
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
SETMSG(HA_ERR_NO_SAVEPOINT, "No savepoint with that name");
SETMSG(HA_ERR_NON_UNIQUE_BLOCK_SIZE, "Non unique key block size");
SETMSG(HA_ERR_NO_SUCH_TABLE, "No such table: '%.64s'");
SETMSG(HA_ERR_TABLE_EXIST, ER(ER_TABLE_EXISTS_ERROR));
SETMSG(HA_ERR_NO_CONNECTION, "Could not connect to storage engine");
BUG#6762 ALTER TABLE gives weird results and error message - Removed hardcoded error message from 4.1
2005-04-07 20:17:37 +02:00
SETMSG(HA_ERR_TABLE_DEF_CHANGED, ER(ER_TABLE_DEF_CHANGED));
Fixed BUG#9680: Wrong error from cascading update Applied 9680.patch (by Osku Salerma)
2006-02-09 18:43:10 +01:00
SETMSG(HA_ERR_FOREIGN_DUPLICATE_KEY, "FK constraint would lead to duplicate key");
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
SETMSG(HA_ERR_TABLE_NEEDS_UPGRADE, ER(ER_TABLE_NEEDS_UPGRADE));
Bug #17766: The server accepts to create MERGE tables which cannot work Changed the error reporting (and a crash) when inserting data into a MERGE table that has no underlying tables or no INSERT_METHOD specified by reporting that it is read-only.
2006-07-12 02:28:50 +02:00
SETMSG(HA_ERR_TABLE_READONLY, ER(ER_OPEN_AS_READONLY));
Bug#20573 "strict mode: inserts autogenerated auto_increment value bigger than max" Strict mode should fail if autoincrement value is out of range
2006-08-30 22:20:39 +02:00
SETMSG(HA_ERR_AUTOINC_READ_FAILED, ER(ER_AUTOINC_READ_FAILED));
SETMSG(HA_ERR_AUTOINC_ERANGE, ER(ER_WARN_DATA_OUT_OF_RANGE));
Fix for BUG#18828 - If InnoDB runs out of undo slots, it returns misleading 'table is full' Innodb returns a misleading error message "table is full" when the number of active concurrent transactions is greater than 1024. Fixed by adding errorcode "ER_TOO_MANY_CONCURRENT_TRXS" to the error codes. Innodb should return HA_TOO_MANY_CONCURRENT_TRXS to mysql which is then mapped to ER_TOO_MANY_CONCURRENT_TRXS Note: testcase is not written as this was reproducible only by changing innodb code.
2009-07-17 10:43:53 +02:00
SETMSG(HA_ERR_TOO_MANY_CONCURRENT_TRXS, ER(ER_TOO_MANY_CONCURRENT_TRXS));
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
/* Register the error messages for use with my_error(). */
return my_error_register(errmsgs, HA_ERR_FIRST, HA_ERR_LAST);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
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
Unregister handler error messages.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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
0 OK
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
!=0 Error
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
*/
static int ha_finish_errors(void)
{
const char **errmsgs;
/* Allocate a pointer array for the error message strings. */
if (! (errmsgs= my_error_unregister(HA_ERR_FIRST, HA_ERR_LAST)))
return 1;
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*) errmsgs, MYF(0));
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
return 0;
}
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
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
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
int ha_finalize_handlerton(st_plugin_int *plugin)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
after review fixes
2006-05-31 18:07:32 +02:00
handlerton *hton= (handlerton *)plugin->data;
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
DBUG_ENTER("ha_finalize_handlerton");
Bug#37958 - test main.plugin crash on Mac OS X when selecting from EXAMPLE engine. This patch contains fixes for two problems: 1. As originally reported, the server crashed on Mac OS X when trying to access an EXAMPLE table after the EXAMPLE plugin was installed. It turned out that the dynamically loaded EXAMPLE plugin called the function hash_earch() from a Mac OS X system library, instead of hash_earch() from MySQL's mysys library. Makefile.am in storage/example does not include libmysys. So the Mac OS X linker arranged the hash_search() function to be linked to the system library when the shared object is loaded. One possible solution would be to include libmysys into the linkage of dynamic plugins. But then we must have a libmysys.so, which must be used by the server too. This could have a minimal performance impact, but foremost the change seems to bee too risky at the current state of MySQL 5.1. The selected solution is to rename MySQL's hash_search() to my_hash_search() like it has been done before with hash_insert() and hash_reset(). Since this is the third time, we need to rename a hash_*() function, I did renamed all hash_*() functions to my_hash_*(). To avoid changing a zillion calls to these functions, and announcing this to hundreds of developers, I added defines that map the old names to the new names. This change is in hash.h and hash.c. 2. The other problem was improper implementation of the handlerton-to-plugin mapping. We use a fixed-size array to hold a plugin reference for each handlerton. On every install of a handler plugin, we allocated a new slot of the array. On uninstall we did not free it. After some uninstall/install cycles the array overflowed. We did not check for overflow. One fix is to check for overflow to stop the crashes. Another fix is to free the array slot at uninstall and search for a free slot at plugin install. This change is in handler.cc.
2008-10-01 12:21:15 +02:00
/* hton can be NULL here, if ha_initialize_handlerton() failed. */
if (!hton)
goto end;
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
switch (hton->state)
{
case SHOW_OPTION_NO:
case SHOW_OPTION_DISABLED:
break;
case SHOW_OPTION_YES:
if (installed_htons[hton->db_type] == hton)
installed_htons[hton->db_type]= NULL;
break;
};
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
Panic was being called twice! Both on its own and in the plugin shutdown.... not so good. The code is a bit simpler, and we could now technically remove the panic all entirely if we wanted to.
2006-10-12 03:02:12 +02:00
if (hton->panic)
hton->panic(hton, HA_PANIC_CLOSE);
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
if (plugin->plugin->deinit)
{
/*
Today we have no defined/special behavior for uninstalling
engine plugins.
*/
DBUG_PRINT("info", ("Deinitializing plugin: '%s'", plugin->name.str));
if (plugin->plugin->deinit(NULL))
{
DBUG_PRINT("warning", ("Plugin '%s' deinit function returned error.",
plugin->name.str));
}
}
Bug#37958 - test main.plugin crash on Mac OS X when selecting from EXAMPLE engine. This patch contains fixes for two problems: 1. As originally reported, the server crashed on Mac OS X when trying to access an EXAMPLE table after the EXAMPLE plugin was installed. It turned out that the dynamically loaded EXAMPLE plugin called the function hash_earch() from a Mac OS X system library, instead of hash_earch() from MySQL's mysys library. Makefile.am in storage/example does not include libmysys. So the Mac OS X linker arranged the hash_search() function to be linked to the system library when the shared object is loaded. One possible solution would be to include libmysys into the linkage of dynamic plugins. But then we must have a libmysys.so, which must be used by the server too. This could have a minimal performance impact, but foremost the change seems to bee too risky at the current state of MySQL 5.1. The selected solution is to rename MySQL's hash_search() to my_hash_search() like it has been done before with hash_insert() and hash_reset(). Since this is the third time, we need to rename a hash_*() function, I did renamed all hash_*() functions to my_hash_*(). To avoid changing a zillion calls to these functions, and announcing this to hundreds of developers, I added defines that map the old names to the new names. This change is in hash.h and hash.c. 2. The other problem was improper implementation of the handlerton-to-plugin mapping. We use a fixed-size array to hold a plugin reference for each handlerton. On every install of a handler plugin, we allocated a new slot of the array. On uninstall we did not free it. After some uninstall/install cycles the array overflowed. We did not check for overflow. One fix is to check for overflow to stop the crashes. Another fix is to free the array slot at uninstall and search for a free slot at plugin install. This change is in handler.cc.
2008-10-01 12:21:15 +02:00
/*
In case a plugin is uninstalled and re-installed later, it should
reuse an array slot. Otherwise the number of uninstall/install
cycles would be limited.
*/
hton2plugin[hton->slot]= NULL;
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*)hton, MYF(0));
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
Bug#37958 - test main.plugin crash on Mac OS X when selecting from EXAMPLE engine. This patch contains fixes for two problems: 1. As originally reported, the server crashed on Mac OS X when trying to access an EXAMPLE table after the EXAMPLE plugin was installed. It turned out that the dynamically loaded EXAMPLE plugin called the function hash_earch() from a Mac OS X system library, instead of hash_earch() from MySQL's mysys library. Makefile.am in storage/example does not include libmysys. So the Mac OS X linker arranged the hash_search() function to be linked to the system library when the shared object is loaded. One possible solution would be to include libmysys into the linkage of dynamic plugins. But then we must have a libmysys.so, which must be used by the server too. This could have a minimal performance impact, but foremost the change seems to bee too risky at the current state of MySQL 5.1. The selected solution is to rename MySQL's hash_search() to my_hash_search() like it has been done before with hash_insert() and hash_reset(). Since this is the third time, we need to rename a hash_*() function, I did renamed all hash_*() functions to my_hash_*(). To avoid changing a zillion calls to these functions, and announcing this to hundreds of developers, I added defines that map the old names to the new names. This change is in hash.h and hash.c. 2. The other problem was improper implementation of the handlerton-to-plugin mapping. We use a fixed-size array to hold a plugin reference for each handlerton. On every install of a handler plugin, we allocated a new slot of the array. On uninstall we did not free it. After some uninstall/install cycles the array overflowed. We did not check for overflow. One fix is to check for overflow to stop the crashes. Another fix is to free the array slot at uninstall and search for a free slot at plugin install. This change is in handler.cc.
2008-10-01 12:21:15 +02:00
end:
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
DBUG_RETURN(0);
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
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
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
int ha_initialize_handlerton(st_plugin_int *plugin)
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
{
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
handlerton *hton;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
DBUG_ENTER("ha_initialize_handlerton");
WL#2936 - Falcon & MySQL plugin interface: server variables Fixed a compiler dependent expression to be more portable. Initialized 'dummy' in construct_help_options() so that loaded plugin show themselves as TRUE in the help output.
2007-04-18 14:37:57 +02:00
DBUG_PRINT("plugin", ("initialize plugin: '%s'", plugin->name.str));
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
hton= (handlerton *)my_malloc(sizeof(handlerton),
MYF(MY_WME | MY_ZEROFILL));
/* Historical Requirement */
after review fixes
2006-05-31 18:07:32 +02:00
plugin->data= hton; // shortcut for the future
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
if (plugin->plugin->init && plugin->plugin->init(hton))
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
{
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
sql_print_error("Plugin '%s' init function returned error.",
plugin->name.str);
goto err;
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
}
after review fixes
2006-05-31 18:07:32 +02:00
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
/*
the switch below and hton->state should be removed when
command-line options for plugins will be implemented
*/
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
switch (hton->state) {
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
case SHOW_OPTION_NO:
break;
case SHOW_OPTION_YES:
{
after review fixes
2006-05-31 18:07:32 +02:00
uint tmp;
Bug#37958 - test main.plugin crash on Mac OS X when selecting from EXAMPLE engine. This patch contains fixes for two problems: 1. As originally reported, the server crashed on Mac OS X when trying to access an EXAMPLE table after the EXAMPLE plugin was installed. It turned out that the dynamically loaded EXAMPLE plugin called the function hash_earch() from a Mac OS X system library, instead of hash_earch() from MySQL's mysys library. Makefile.am in storage/example does not include libmysys. So the Mac OS X linker arranged the hash_search() function to be linked to the system library when the shared object is loaded. One possible solution would be to include libmysys into the linkage of dynamic plugins. But then we must have a libmysys.so, which must be used by the server too. This could have a minimal performance impact, but foremost the change seems to bee too risky at the current state of MySQL 5.1. The selected solution is to rename MySQL's hash_search() to my_hash_search() like it has been done before with hash_insert() and hash_reset(). Since this is the third time, we need to rename a hash_*() function, I did renamed all hash_*() functions to my_hash_*(). To avoid changing a zillion calls to these functions, and announcing this to hundreds of developers, I added defines that map the old names to the new names. This change is in hash.h and hash.c. 2. The other problem was improper implementation of the handlerton-to-plugin mapping. We use a fixed-size array to hold a plugin reference for each handlerton. On every install of a handler plugin, we allocated a new slot of the array. On uninstall we did not free it. After some uninstall/install cycles the array overflowed. We did not check for overflow. One fix is to check for overflow to stop the crashes. Another fix is to free the array slot at uninstall and search for a free slot at plugin install. This change is in handler.cc.
2008-10-01 12:21:15 +02:00
ulong fslot;
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
/* now check the db_type for conflict */
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
if (hton->db_type <= DB_TYPE_UNKNOWN ||
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
hton->db_type >= DB_TYPE_DEFAULT ||
installed_htons[hton->db_type])
{
int idx= (int) DB_TYPE_FIRST_DYNAMIC;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
while (idx < (int) DB_TYPE_DEFAULT && installed_htons[idx])
idx++;
if (idx == (int) DB_TYPE_DEFAULT)
{
sql_print_warning("Too many storage engines!");
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
goto err_deinit;
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
}
if (hton->db_type != DB_TYPE_UNKNOWN)
sql_print_warning("Storage engine '%s' has conflicting typecode. "
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
"Assigning value %d.", plugin->plugin->name, idx);
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
hton->db_type= (enum legacy_db_type) idx;
}
Bug#37958 - test main.plugin crash on Mac OS X when selecting from EXAMPLE engine. This patch contains fixes for two problems: 1. As originally reported, the server crashed on Mac OS X when trying to access an EXAMPLE table after the EXAMPLE plugin was installed. It turned out that the dynamically loaded EXAMPLE plugin called the function hash_earch() from a Mac OS X system library, instead of hash_earch() from MySQL's mysys library. Makefile.am in storage/example does not include libmysys. So the Mac OS X linker arranged the hash_search() function to be linked to the system library when the shared object is loaded. One possible solution would be to include libmysys into the linkage of dynamic plugins. But then we must have a libmysys.so, which must be used by the server too. This could have a minimal performance impact, but foremost the change seems to bee too risky at the current state of MySQL 5.1. The selected solution is to rename MySQL's hash_search() to my_hash_search() like it has been done before with hash_insert() and hash_reset(). Since this is the third time, we need to rename a hash_*() function, I did renamed all hash_*() functions to my_hash_*(). To avoid changing a zillion calls to these functions, and announcing this to hundreds of developers, I added defines that map the old names to the new names. This change is in hash.h and hash.c. 2. The other problem was improper implementation of the handlerton-to-plugin mapping. We use a fixed-size array to hold a plugin reference for each handlerton. On every install of a handler plugin, we allocated a new slot of the array. On uninstall we did not free it. After some uninstall/install cycles the array overflowed. We did not check for overflow. One fix is to check for overflow to stop the crashes. Another fix is to free the array slot at uninstall and search for a free slot at plugin install. This change is in handler.cc.
2008-10-01 12:21:15 +02:00
/*
In case a plugin is uninstalled and re-installed later, it should
reuse an array slot. Otherwise the number of uninstall/install
cycles would be limited. So look for a free slot.
*/
DBUG_PRINT("plugin", ("total_ha: %lu", total_ha));
for (fslot= 0; fslot < total_ha; fslot++)
{
if (!hton2plugin[fslot])
break;
}
if (fslot < total_ha)
hton->slot= fslot;
else
{
if (total_ha >= MAX_HA)
{
sql_print_error("Too many plugins loaded. Limit is %lu. "
"Failed on '%s'", (ulong) MAX_HA, plugin->name.str);
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
goto err_deinit;
Bug#37958 - test main.plugin crash on Mac OS X when selecting from EXAMPLE engine. This patch contains fixes for two problems: 1. As originally reported, the server crashed on Mac OS X when trying to access an EXAMPLE table after the EXAMPLE plugin was installed. It turned out that the dynamically loaded EXAMPLE plugin called the function hash_earch() from a Mac OS X system library, instead of hash_earch() from MySQL's mysys library. Makefile.am in storage/example does not include libmysys. So the Mac OS X linker arranged the hash_search() function to be linked to the system library when the shared object is loaded. One possible solution would be to include libmysys into the linkage of dynamic plugins. But then we must have a libmysys.so, which must be used by the server too. This could have a minimal performance impact, but foremost the change seems to bee too risky at the current state of MySQL 5.1. The selected solution is to rename MySQL's hash_search() to my_hash_search() like it has been done before with hash_insert() and hash_reset(). Since this is the third time, we need to rename a hash_*() function, I did renamed all hash_*() functions to my_hash_*(). To avoid changing a zillion calls to these functions, and announcing this to hundreds of developers, I added defines that map the old names to the new names. This change is in hash.h and hash.c. 2. The other problem was improper implementation of the handlerton-to-plugin mapping. We use a fixed-size array to hold a plugin reference for each handlerton. On every install of a handler plugin, we allocated a new slot of the array. On uninstall we did not free it. After some uninstall/install cycles the array overflowed. We did not check for overflow. One fix is to check for overflow to stop the crashes. Another fix is to free the array slot at uninstall and search for a free slot at plugin install. This change is in handler.cc.
2008-10-01 12:21:15 +02:00
}
hton->slot= total_ha++;
}
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
installed_htons[hton->db_type]= hton;
tmp= hton->savepoint_offset;
hton->savepoint_offset= savepoint_alloc_size;
savepoint_alloc_size+= tmp;
after review fixes
2006-05-31 18:07:32 +02:00
hton2plugin[hton->slot]=plugin;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
if (hton->prepare)
total_ha_2pc++;
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
break;
}
/* fall through */
default:
hton->state= SHOW_OPTION_DISABLED;
break;
}
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
/*
This patch updates to remove most global hton needs. Cleans up wrong message in partition.
2006-10-01 03:31:13 +02:00
This is entirely for legacy. We will create a new "disk based" hton and a
"memory" hton which will be configurable longterm. We should be able to
remove partition and myisammrg.
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
*/
switch (hton->db_type) {
case DB_TYPE_HEAP:
heap_hton= hton;
break;
case DB_TYPE_MYISAM:
myisam_hton= hton;
break;
case DB_TYPE_PARTITION_DB:
partition_hton= hton;
break;
default:
break;
};
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
DBUG_RETURN(0);
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
err_deinit:
/*
Let plugin do its inner deinitialization as plugin->init()
was successfully called before.
*/
if (plugin->plugin->deinit)
(void) plugin->plugin->deinit(NULL);
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
err:
Fix for bug #40757: Starting server on Windows with innodb_flush_method=wrong_value causes crash Problem: after a failed plugin initialization, incompletely initialized data remained in the plugin and handlerton data structures. These were used later and caused the crash. Fix: clean-up plugin related data if initialization failed. Note: no test case added, hand tested.
2009-02-10 09:37:27 +01:00
my_free((uchar*) hton, MYF(0));
plugin->data= NULL;
This changes the order of the universe, black is now the new white. In practice this means that handlerton is now created by the server and is passed to the engine. Plugin startups can now also control how plugins are inited (and can optionally pass values). Bit more flexibility to those who want to write plugin interfaces to the database.
2006-09-15 19:28:00 +02:00
DBUG_RETURN(1);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
int ha_init()
{
Cleanup ha_checktype() and redundant variables.
2004-03-02 11:08:50 +01:00
int error= 0;
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
DBUG_ENTER("ha_init");
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ASSERT(total_ha < MAX_HA);
Eliminate warnings noticed by VC7. This includes fixing my_mmap() on Windows to call CreateFileMapping() with correct arguments, and propogating the introduction of query_id_t to everywhere query ids are passed around. (Bug #8826)
2005-03-19 01:12:25 +01:00
/*
Check if there is a transaction-capable storage engine besides the
binary log (which is considered a transaction-capable storage engine in
counting total_ha)
*/
opt_using_transactions= total_ha>(ulong)opt_bin_log;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
savepoint_alloc_size+= sizeof(SAVEPOINT);
WL#3201 " Configure support for server plugins "
2006-04-13 22:49:29 +02:00
DBUG_RETURN(error);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
}
Adding in an ha_end() call. This is the first part of the patch I and Monty have agreed on. In the future engines will have the option of shutting down more quickly via the panic call, or just shutting down during the unloading of a plugin.
2006-11-01 05:10:32 +01:00
int ha_end()
Import changeset
2000-07-31 21:29:14 +02:00
{
Adding in an ha_end() call. This is the first part of the patch I and Monty have agreed on. In the future engines will have the option of shutting down more quickly via the panic call, or just shutting down during the unloading of a plugin.
2006-11-01 05:10:32 +01:00
int error= 0;
DBUG_ENTER("ha_end");
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
Adding in an ha_end() call. This is the first part of the patch I and Monty have agreed on. In the future engines will have the option of shutting down more quickly via the panic call, or just shutting down during the unloading of a plugin.
2006-11-01 05:10:32 +01:00
/*
This should be eventualy based on the graceful shutdown flag.
So if flag is equal to HA_PANIC_CLOSE, the deallocate
the errors.
*/
if (ha_finish_errors())
error= 1;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
Adding in an ha_end() call. This is the first part of the patch I and Monty have agreed on. In the future engines will have the option of shutting down more quickly via the panic call, or just shutting down during the unloading of a plugin.
2006-11-01 05:10:32 +01:00
DBUG_RETURN(error);
}
Import changeset
2000-07-31 21:29:14 +02:00
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
static my_bool dropdb_handlerton(THD *unused1, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *path)
{
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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (hton->state == SHOW_OPTION_YES && hton->drop_database)
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->drop_database(hton, (char *)path);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return FALSE;
}
ut0mem.c Merge changes in InnoDB-3.23.43b ut0ut.c Merge changes in InnoDB-3.23.43b trx0purge.c Merge changes in InnoDB-3.23.43b trx0rec.c Merge changes in InnoDB-3.23.43b trx0trx.c Merge changes in InnoDB-3.23.43b trx0undo.c Merge changes in InnoDB-3.23.43b thr0loc.c Merge changes in InnoDB-3.23.43b sync0arr.c Merge changes in InnoDB-3.23.43b sync0rw.c Merge changes in InnoDB-3.23.43b sync0sync.c Merge changes in InnoDB-3.23.43b srv0srv.c Merge changes in InnoDB-3.23.43b srv0start.c Merge changes in InnoDB-3.23.43b row0ins.c Merge changes in InnoDB-3.23.43b row0mysql.c Merge changes in InnoDB-3.23.43b row0purge.c Merge changes in InnoDB-3.23.43b row0sel.c Merge changes in InnoDB-3.23.43b row0umod.c Merge changes in InnoDB-3.23.43b row0upd.c Merge changes in InnoDB-3.23.43b row0vers.c Merge changes in InnoDB-3.23.43b rem0cmp.c Merge changes in InnoDB-3.23.43b que0que.c Merge changes in InnoDB-3.23.43b pars0opt.c Merge changes in InnoDB-3.23.43b pars0pars.c Merge changes in InnoDB-3.23.43b lexyy.c Merge changes in InnoDB-3.23.43b pars0grm.c Merge changes in InnoDB-3.23.43b page0page.c Merge changes in InnoDB-3.23.43b os0file.c Merge changes in InnoDB-3.23.43b mtr0log.c Merge changes in InnoDB-3.23.43b mem0pool.c Merge changes in InnoDB-3.23.43b log0log.c Merge changes in InnoDB-3.23.43b log0recv.c Merge changes in InnoDB-3.23.43b lock0lock.c Merge changes in InnoDB-3.23.43b ibuf0ibuf.c Merge changes in InnoDB-3.23.43b fil0fil.c Merge changes in InnoDB-3.23.43b dict0crea.c Merge changes in InnoDB-3.23.43b dict0dict.c Merge changes in InnoDB-3.23.43b dict0load.c Merge changes in InnoDB-3.23.43b dict0mem.c Merge changes in InnoDB-3.23.43b data0data.c Merge changes in InnoDB-3.23.43b data0type.c Merge changes in InnoDB-3.23.43b buf0buf.c Merge changes in InnoDB-3.23.43b buf0lru.c Merge changes in InnoDB-3.23.43b btr0btr.c Merge changes in InnoDB-3.23.43b btr0cur.c Merge changes in InnoDB-3.23.43b btr0pcur.c Merge changes in InnoDB-3.23.43b btr0sea.c Merge changes in InnoDB-3.23.43b data0type.ic Merge changes in InnoDB-3.23.43b dict0dict.ic Merge changes in InnoDB-3.23.43b mtr0mtr.ic Merge changes in InnoDB-3.23.43b row0upd.ic Merge changes in InnoDB-3.23.43b sync0ipm.ic Merge changes in InnoDB-3.23.43b sync0rw.ic Merge changes in InnoDB-3.23.43b sync0sync.ic Merge changes in InnoDB-3.23.43b trx0rseg.ic Merge changes in InnoDB-3.23.43b btr0pcur.ic Merge changes in InnoDB-3.23.43b buf0buf.ic Merge changes in InnoDB-3.23.43b data0data.ic Merge changes in InnoDB-3.23.43b row0upd.h Merge changes in InnoDB-3.23.43b srv0srv.h Merge changes in InnoDB-3.23.43b sync0arr.h Merge changes in InnoDB-3.23.43b sync0rw.h Merge changes in InnoDB-3.23.43b sync0sync.h Merge changes in InnoDB-3.23.43b trx0trx.h Merge changes in InnoDB-3.23.43b ut0mem.h Merge changes in InnoDB-3.23.43b data0data.h Merge changes in InnoDB-3.23.43b data0type.h Merge changes in InnoDB-3.23.43b db0err.h Merge changes in InnoDB-3.23.43b dict0crea.h Merge changes in InnoDB-3.23.43b dict0dict.h Merge changes in InnoDB-3.23.43b dict0load.h Merge changes in InnoDB-3.23.43b dict0mem.h Merge changes in InnoDB-3.23.43b dict0types.h Merge changes in InnoDB-3.23.43b fil0fil.h Merge changes in InnoDB-3.23.43b ibuf0ibuf.h Merge changes in InnoDB-3.23.43b lock0lock.h Merge changes in InnoDB-3.23.43b log0log.h Merge changes in InnoDB-3.23.43b mtr0mtr.h Merge changes in InnoDB-3.23.43b rem0cmp.h Merge changes in InnoDB-3.23.43b row0ins.h Merge changes in InnoDB-3.23.43b row0mysql.h Merge changes in InnoDB-3.23.43b btr0cur.h Merge changes in InnoDB-3.23.43b btr0pcur.h Merge changes in InnoDB-3.23.43b btr0sea.h Merge changes in InnoDB-3.23.43b buf0buf.h Merge changes in InnoDB-3.23.43b sql_table.cc Merge changes in InnoDB-3.23.43b sql_db.cc Merge changes in InnoDB-3.23.43b ha_innobase.cc Merge changes in InnoDB-3.23.43b handler.cc Merge changes in InnoDB-3.23.43b ha_innobase.h Merge changes in InnoDB-3.23.43b handler.h Merge changes in InnoDB-3.23.43b
2001-10-10 21:47:08 +02:00
void ha_drop_database(char* path)
{
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
plugin_foreach(NULL, dropdb_handlerton, MYSQL_STORAGE_ENGINE_PLUGIN, path);
}
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +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
static my_bool closecon_handlerton(THD *thd, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *unused)
{
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
handlerton *hton= plugin_data(plugin, handlerton *);
after review fixes
2006-05-31 18:07:32 +02:00
/*
there's no need to rollback here as all transactions must
be rolled back already
*/
BUG#39053 - UNISTALL PLUGIN does not allow the storage engine to cleanup open connections It was possible to UNINSTALL storage engine plugin when binding between THD object and storage engine is still active (e.g. in the middle of transaction). To avoid unclean deactivation (uninstall) of storage engine plugin in the middle of transaction, additional storage engine plugin lock is acquired by thd_set_ha_data(). If ha_data is not null and storage engine plugin was not locked by thd_set_ha_data() in this connection before, storage engine plugin gets locked. If ha_data is null and storage engine plugin was locked by thd_set_ha_data() in this connection before, storage engine plugin lock gets released. If handlerton::close_connection() didn't reset ha_data, server does it immediately after calling handlerton::close_connection(). Note that this is just a framework fix, storage engines must switch to thd_set_ha_data() from thd_ha_data() if they want to see fit.
2010-04-14 11:53:59 +02:00
if (hton->state == SHOW_OPTION_YES && thd_get_ha_data(thd, hton))
{
if (hton->close_connection)
hton->close_connection(hton, thd);
/* make sure ha_data is reset and ha_data_lock is released */
thd_set_ha_data(thd, hton, NULL);
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return FALSE;
ut0mem.c Merge changes in InnoDB-3.23.43b ut0ut.c Merge changes in InnoDB-3.23.43b trx0purge.c Merge changes in InnoDB-3.23.43b trx0rec.c Merge changes in InnoDB-3.23.43b trx0trx.c Merge changes in InnoDB-3.23.43b trx0undo.c Merge changes in InnoDB-3.23.43b thr0loc.c Merge changes in InnoDB-3.23.43b sync0arr.c Merge changes in InnoDB-3.23.43b sync0rw.c Merge changes in InnoDB-3.23.43b sync0sync.c Merge changes in InnoDB-3.23.43b srv0srv.c Merge changes in InnoDB-3.23.43b srv0start.c Merge changes in InnoDB-3.23.43b row0ins.c Merge changes in InnoDB-3.23.43b row0mysql.c Merge changes in InnoDB-3.23.43b row0purge.c Merge changes in InnoDB-3.23.43b row0sel.c Merge changes in InnoDB-3.23.43b row0umod.c Merge changes in InnoDB-3.23.43b row0upd.c Merge changes in InnoDB-3.23.43b row0vers.c Merge changes in InnoDB-3.23.43b rem0cmp.c Merge changes in InnoDB-3.23.43b que0que.c Merge changes in InnoDB-3.23.43b pars0opt.c Merge changes in InnoDB-3.23.43b pars0pars.c Merge changes in InnoDB-3.23.43b lexyy.c Merge changes in InnoDB-3.23.43b pars0grm.c Merge changes in InnoDB-3.23.43b page0page.c Merge changes in InnoDB-3.23.43b os0file.c Merge changes in InnoDB-3.23.43b mtr0log.c Merge changes in InnoDB-3.23.43b mem0pool.c Merge changes in InnoDB-3.23.43b log0log.c Merge changes in InnoDB-3.23.43b log0recv.c Merge changes in InnoDB-3.23.43b lock0lock.c Merge changes in InnoDB-3.23.43b ibuf0ibuf.c Merge changes in InnoDB-3.23.43b fil0fil.c Merge changes in InnoDB-3.23.43b dict0crea.c Merge changes in InnoDB-3.23.43b dict0dict.c Merge changes in InnoDB-3.23.43b dict0load.c Merge changes in InnoDB-3.23.43b dict0mem.c Merge changes in InnoDB-3.23.43b data0data.c Merge changes in InnoDB-3.23.43b data0type.c Merge changes in InnoDB-3.23.43b buf0buf.c Merge changes in InnoDB-3.23.43b buf0lru.c Merge changes in InnoDB-3.23.43b btr0btr.c Merge changes in InnoDB-3.23.43b btr0cur.c Merge changes in InnoDB-3.23.43b btr0pcur.c Merge changes in InnoDB-3.23.43b btr0sea.c Merge changes in InnoDB-3.23.43b data0type.ic Merge changes in InnoDB-3.23.43b dict0dict.ic Merge changes in InnoDB-3.23.43b mtr0mtr.ic Merge changes in InnoDB-3.23.43b row0upd.ic Merge changes in InnoDB-3.23.43b sync0ipm.ic Merge changes in InnoDB-3.23.43b sync0rw.ic Merge changes in InnoDB-3.23.43b sync0sync.ic Merge changes in InnoDB-3.23.43b trx0rseg.ic Merge changes in InnoDB-3.23.43b btr0pcur.ic Merge changes in InnoDB-3.23.43b buf0buf.ic Merge changes in InnoDB-3.23.43b data0data.ic Merge changes in InnoDB-3.23.43b row0upd.h Merge changes in InnoDB-3.23.43b srv0srv.h Merge changes in InnoDB-3.23.43b sync0arr.h Merge changes in InnoDB-3.23.43b sync0rw.h Merge changes in InnoDB-3.23.43b sync0sync.h Merge changes in InnoDB-3.23.43b trx0trx.h Merge changes in InnoDB-3.23.43b ut0mem.h Merge changes in InnoDB-3.23.43b data0data.h Merge changes in InnoDB-3.23.43b data0type.h Merge changes in InnoDB-3.23.43b db0err.h Merge changes in InnoDB-3.23.43b dict0crea.h Merge changes in InnoDB-3.23.43b dict0dict.h Merge changes in InnoDB-3.23.43b dict0load.h Merge changes in InnoDB-3.23.43b dict0mem.h Merge changes in InnoDB-3.23.43b dict0types.h Merge changes in InnoDB-3.23.43b fil0fil.h Merge changes in InnoDB-3.23.43b ibuf0ibuf.h Merge changes in InnoDB-3.23.43b lock0lock.h Merge changes in InnoDB-3.23.43b log0log.h Merge changes in InnoDB-3.23.43b mtr0mtr.h Merge changes in InnoDB-3.23.43b rem0cmp.h Merge changes in InnoDB-3.23.43b row0ins.h Merge changes in InnoDB-3.23.43b row0mysql.h Merge changes in InnoDB-3.23.43b btr0cur.h Merge changes in InnoDB-3.23.43b btr0pcur.h Merge changes in InnoDB-3.23.43b btr0sea.h Merge changes in InnoDB-3.23.43b buf0buf.h Merge changes in InnoDB-3.23.43b sql_table.cc Merge changes in InnoDB-3.23.43b sql_db.cc Merge changes in InnoDB-3.23.43b ha_innobase.cc Merge changes in InnoDB-3.23.43b handler.cc Merge changes in InnoDB-3.23.43b ha_innobase.h Merge changes in InnoDB-3.23.43b handler.h Merge changes in InnoDB-3.23.43b
2001-10-10 21:47:08 +02:00
}
Import changeset
2000-07-31 21:29:14 +02:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@note
don't bother to rollback here, it's done already
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Added innobase handler files + other updates for innobase
2000-12-06 00:54:17 +01:00
void ha_close_connection(THD* thd)
{
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
plugin_foreach(thd, closecon_handlerton, MYSQL_STORAGE_ENGINE_PLUGIN, 0);
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
/* ========================================================================
======================= TRANSACTIONS ===================================*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/**
Transaction handling in the server
==================================
In each client connection, MySQL maintains two transactional
states:
- a statement transaction,
- a standard, also called normal transaction.
Historical note
---------------
"Statement transaction" is a non-standard term that comes
from the times when MySQL supported BerkeleyDB storage engine.
First of all, it should be said that in BerkeleyDB auto-commit
mode auto-commits operations that are atomic to the storage
engine itself, such as a write of a record, and are too
high-granular to be atomic from the application perspective
(MySQL). One SQL statement could involve many BerkeleyDB
auto-committed operations and thus BerkeleyDB auto-commit was of
little use to MySQL.
Secondly, instead of SQL standard savepoints, BerkeleyDB
provided the concept of "nested transactions". In a nutshell,
transactions could be arbitrarily nested, but when the parent
transaction was committed or aborted, all its child (nested)
transactions were handled committed or aborted as well.
Commit of a nested transaction, in turn, made its changes
visible, but not durable: it destroyed the nested transaction,
all its changes would become available to the parent and
currently active nested transactions of this parent.
So the mechanism of nested transactions was employed to
provide "all or nothing" guarantee of SQL statements
required by the standard.
A nested transaction would be created at start of each SQL
statement, and destroyed (committed or aborted) at statement
end. Such nested transaction was internally referred to as
a "statement transaction" and gave birth to the term.
<Historical note ends>
Since then a statement transaction is started for each statement
that accesses transactional tables or uses the binary log. If
the statement succeeds, the statement transaction is committed.
If the statement fails, the transaction is rolled back. Commits
of statement transactions are not durable -- each such
transaction is nested in the normal transaction, and if the
normal transaction is rolled back, the effects of all enclosed
statement transactions are undone as well. Technically,
a statement transaction can be viewed as a savepoint which is
maintained automatically in order to make effects of one
statement atomic.
The normal transaction is started by the user and is ended
usually upon a user request as well. The normal transaction
encloses transactions of all statements issued between
its beginning and its end.
In autocommit mode, the normal transaction is equivalent
to the statement transaction.
Since MySQL supports PSEA (pluggable storage engine
architecture), more than one transactional engine can be
active at a time. Hence transactions, from the server
point of view, are always distributed. In particular,
transactional state is maintained independently for each
engine. In order to commit a transaction the two phase
commit protocol is employed.
Not all statements are executed in context of a transaction.
Administrative and status information statements do not modify
engine data, and thus do not start a statement transaction and
also have no effect on the normal transaction. Examples of such
statements are SHOW STATUS and RESET SLAVE.
Similarly DDL statements are not transactional,
and therefore a transaction is [almost] never started for a DDL
statement. The difference between a DDL statement and a purely
administrative statement though is that a DDL statement always
commits the current transaction before proceeding, if there is
any.
At last, SQL statements that work with non-transactional
engines also have no effect on the transaction state of the
connection. Even though they are written to the binary log,
and the binary log is, overall, transactional, the writes
are done in "write-through" mode, directly to the binlog
file, followed with a OS cache sync, in other words,
bypassing the binlog undo log (translog).
They do not commit the current normal transaction.
A failure of a statement that uses non-transactional tables
would cause a rollback of the statement transaction, but
in case there no non-transactional tables are used,
no statement transaction is started.
Data layout
-----------
The server stores its transaction-related data in
thd->transaction. This structure has two members of type
THD_TRANS. These members correspond to the statement and
normal transactions respectively:
- thd->transaction.stmt contains a list of engines
that are participating in the given statement
- thd->transaction.all contains a list of engines that
have participated in any of the statement transactions started
within the context of the normal transaction.
Each element of the list contains a pointer to the storage
engine, engine-specific transactional data, and engine-specific
transaction flags.
In autocommit mode thd->transaction.all is empty.
Instead, data of thd->transaction.stmt is
used to commit/rollback the normal transaction.
The list of registered engines has a few important properties:
- no engine is registered in the list twice
- engines are present in the list a reverse temporal order --
new participants are always added to the beginning of the list.
Transaction life cycle
----------------------
When a new connection is established, thd->transaction
members are initialized to an empty state.
If a statement uses any tables, all affected engines
are registered in the statement engine list. In
non-autocommit mode, the same engines are registered in
the normal transaction list.
At the end of the statement, the server issues a commit
or a roll back for all engines in the statement list.
At this point transaction flags of an engine, if any, are
propagated from the statement list to the list of the normal
transaction.
When commit/rollback is finished, the statement list is
cleared. It will be filled in again by the next statement,
and emptied again at the next statement's end.
The normal transaction is committed in a similar way
(by going over all engines in thd->transaction.all list)
but at different times:
- upon COMMIT SQL statement is issued by the user
- implicitly, by the server, at the beginning of a DDL statement
or SET AUTOCOMMIT={0|1} statement.
The normal transaction can be rolled back as well:
- if the user has requested so, by issuing ROLLBACK SQL
statement
- if one of the storage engines requested a rollback
by setting thd->transaction_rollback_request. This may
happen in case, e.g., when the transaction in the engine was
chosen a victim of the internal deadlock resolution algorithm
and rolled back internally. When such a situation happens, there
is little the server can do and the only option is to rollback
transactions in all other participating engines. In this case
the rollback is accompanied by an error sent to the user.
As follows from the use cases above, the normal transaction
is never committed when there is an outstanding statement
transaction. In most cases there is no conflict, since
commits of the normal transaction are issued by a stand-alone
administrative or DDL statement, thus no outstanding statement
transaction of the previous statement exists. Besides,
all statements that manipulate with the normal transaction
are prohibited in stored functions and triggers, therefore
no conflicting situation can occur in a sub-statement either.
The remaining rare cases when the server explicitly has
to commit the statement transaction prior to committing the normal
one cover error-handling scenarios (see for example
SQLCOM_LOCK_TABLES).
When committing a statement or a normal transaction, the server
either uses the two-phase commit protocol, or issues a commit
in each engine independently. The two-phase commit protocol
is used only if:
- all participating engines support two-phase commit (provide
handlerton::prepare PSEA API call) and
- transactions in at least two engines modify data (i.e. are
not read-only).
Note that the two phase commit is used for
statement transactions, even though they are not durable anyway.
This is done to ensure logical consistency of data in a multiple-
engine transaction.
For example, imagine that some day MySQL supports unique
constraint checks deferred till the end of statement. In such
case a commit in one of the engines may yield ER_DUP_KEY,
and MySQL should be able to gracefully abort statement
transactions of other participants.
After the normal transaction has been committed,
thd->transaction.all list is cleared.
When a connection is closed, the current normal transaction, if
any, is rolled back.
Roles and responsibilities
--------------------------
The server has no way to know that an engine participates in
the statement and a transaction has been started
in it unless the engine says so. Thus, in order to be
a part of a transaction, the engine must "register" itself.
This is done by invoking trans_register_ha() server call.
Normally the engine registers itself whenever handler::external_lock()
is called. trans_register_ha() can be invoked many times: if
an engine is already registered, the call does nothing.
In case autocommit is not set, the engine must register itself
twice -- both in the statement list and in the normal transaction
list.
In which list to register is a parameter of trans_register_ha().
Note, that although the registration interface in itself is
fairly clear, the current usage practice often leads to undesired
effects. E.g. since a call to trans_register_ha() in most engines
is embedded into implementation of handler::external_lock(), some
DDL statements start a transaction (at least from the server
point of view) even though they are not expected to. E.g.
CREATE TABLE does not start a transaction, since
handler::external_lock() is never called during CREATE TABLE. But
CREATE TABLE ... SELECT does, since handler::external_lock() is
called for the table that is being selected from. This has no
practical effects currently, but must be kept in mind
nevertheless.
Once an engine is registered, the server will do the rest
of the work.
During statement execution, whenever any of data-modifying
PSEA API methods is used, e.g. handler::write_row() or
handler::update_row(), the read-write flag is raised in the
statement transaction for the involved engine.
Currently All PSEA calls are "traced", and the data can not be
changed in a way other than issuing a PSEA call. Important:
unless this invariant is preserved the server will not know that
a transaction in a given engine is read-write and will not
involve the two-phase commit protocol!
At the end of a statement, server call
ha_autocommit_or_rollback() is invoked. This call in turn
invokes handlerton::prepare() for every involved engine.
Prepare is followed by a call to handlerton::commit_one_phase()
If a one-phase commit will suffice, handlerton::prepare() is not
invoked and the server only calls handlerton::commit_one_phase().
At statement commit, the statement-related read-write engine
flag is propagated to the corresponding flag in the normal
transaction. When the commit is complete, the list of registered
engines is cleared.
Rollback is handled in a similar fashion.
Additional notes on DDL and the normal transaction.
---------------------------------------------------
DDLs and operations with non-transactional engines
do not "register" in thd->transaction lists, and thus do not
modify the transaction state. Besides, each DDL in
MySQL is prefixed with an implicit normal transaction commit
(a call to end_active_trans()), and thus leaves nothing
to modify.
However, as it has been pointed out with CREATE TABLE .. SELECT,
some DDL statements can start a *new* transaction.
Behaviour of the server in this case is currently badly
defined.
DDL statements use a form of "semantic" logging
to maintain atomicity: if CREATE TABLE .. SELECT failed,
the newly created table is deleted.
In addition, some DDL statements issue interim transaction
commits: e.g. ALTER TABLE issues a commit after data is copied
from the original table to the internal temporary table. Other
statements, e.g. CREATE TABLE ... SELECT do not always commit
after itself.
And finally there is a group of DDL statements such as
RENAME/DROP TABLE that doesn't start a new transaction
and doesn't commit.
This diversity makes it hard to say what will happen if
by chance a stored function is invoked during a DDL --
whether any modifications it makes will be committed or not
is not clear. Fortunately, SQL grammar of few DDLs allows
invocation of a stored function.
A consistent behaviour is perhaps to always commit the normal
transaction after all DDLs, just like the statement transaction
is always committed at the end of all statements.
*/
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Register a storage engine for a transaction.
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
Every storage engine MUST call this function when it starts
a transaction or a statement (that is it must be called both for the
"beginning of transaction" and "beginning of statement").
Only storage engines registered for the transaction/statement
will know when to commit/rollback it.
Per Heikki's request make trans_register_ha tolerant to multiple registrations
2005-04-05 18:42:34 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
Per Heikki's request make trans_register_ha tolerant to multiple registrations
2005-04-05 18:42:34 +02:00
trans_register_ha is idempotent - storage engine may register many
times per transaction.
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
*/
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
void trans_register_ha(THD *thd, bool all, handlerton *ht_arg)
{
THD_TRANS *trans;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Ha_trx_info *ha_info;
xa assert fixes
2005-02-21 11:51:50 +01:00
DBUG_ENTER("trans_register_ha");
DBUG_PRINT("enter",("%s", all ? "all" : "stmt"));
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
if (all)
{
trans= &thd->transaction.all;
thd->server_status|= SERVER_STATUS_IN_TRANS;
}
else
trans= &thd->transaction.stmt;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
ha_info= thd->ha_data[ht_arg->slot].ha_info + static_cast<unsigned>(all);
if (ha_info->is_started())
DBUG_VOID_RETURN; /* already registered, return */
ha_info->register_ha(trans, ht_arg);
handler.cc, ha_innodb.cc: Fix Bug #8650 : InnoDB does not rollback SQL statement at an error
2005-04-05 16:56:40 +02:00
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
trans->no_2pc|=(ht_arg->prepare==0);
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
if (thd->transaction.xid_state.xid.is_null())
thd->transaction.xid_state.xid.set(thd->query_id);
xa assert fixes
2005-02-21 11:51:50 +01:00
DBUG_VOID_RETURN;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@retval
0 ok
@retval
1 error, transaction was rolled back
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
*/
int ha_prepare(THD *thd)
{
int error=0, all=1;
THD_TRANS *trans=all ? &thd->transaction.all : &thd->transaction.stmt;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Ha_trx_info *ha_info= trans->ha_list;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_prepare");
#ifdef USING_TRANSACTIONS
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (ha_info)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (; ha_info; ha_info= ha_info->next())
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_prepare_count);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (ht->prepare)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if ((err= ht->prepare(ht, thd, all)))
don't crash when in XA PREPARE not all storage engines are 2pc-capable
2005-09-07 18:03:07 +02:00
{
my_error(ER_ERROR_DURING_COMMIT, MYF(0), err);
ha_rollback_trans(thd, all);
error=1;
break;
}
}
else
{
push_warning_printf(thd, MYSQL_ERROR::WARN_LEVEL_WARN,
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
ER_ILLEGAL_HA, ER(ER_ILLEGAL_HA),
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
ha_resolve_storage_engine_name(ht));
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
}
}
#endif /* USING_TRANSACTIONS */
DBUG_RETURN(error);
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/**
Check if we can skip the two-phase commit.
A helper function to evaluate if two-phase commit is mandatory.
As a side effect, propagates the read-only/read-write flags
of the statement transaction to its enclosing normal transaction.
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
If we have at least two engines with read-write changes we must
run a two-phase commit. Otherwise we can run several independent
commits as the only transactional engine has read-write changes
and others are read-only.
@retval 0 All engines are read-only.
@retval 1 We have the only engine with read-write changes.
@retval >1 More than one engine have read-write changes.
Note: return value might NOT be the exact number of
engines with read-write changes.
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
*/
static
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
uint
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
ha_check_and_coalesce_trx_read_only(THD *thd, Ha_trx_info *ha_list,
bool all)
{
/* The number of storage engines that have actual changes. */
unsigned rw_ha_count= 0;
Ha_trx_info *ha_info;
for (ha_info= ha_list; ha_info; ha_info= ha_info->next())
{
if (ha_info->is_trx_read_write())
++rw_ha_count;
if (! all)
{
Ha_trx_info *ha_info_all= &thd->ha_data[ha_info->ht()->slot].ha_info[1];
DBUG_ASSERT(ha_info != ha_info_all);
/*
Merge read-only/read-write information about statement
transaction to its enclosing normal transaction. Do this
only if in a real transaction -- that is, if we know
that ha_info_all is registered in thd->transaction.all.
Since otherwise we only clutter the normal transaction flags.
*/
if (ha_info_all->is_started()) /* FALSE if autocommit. */
ha_info_all->coalesce_trx_with(ha_info);
}
else if (rw_ha_count > 1)
{
/*
It is a normal transaction, so we don't need to merge read/write
information up, and the need for two-phase commit has been
already established. Break the loop prematurely.
*/
break;
}
}
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
return rw_ha_count;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@retval
0 ok
@retval
1 transaction was rolled back
@retval
2 error during commit, data may be inconsistent
@todo
Since we don't support nested statement transactions in 5.0,
we can't commit or rollback stmt transactions while we are inside
stored functions or triggers. So we simply do nothing now.
TODO: This should be fixed in later ( >= 5.1) releases.
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
*/
int ha_commit_trans(THD *thd, bool all)
{
int error= 0, cookie= 0;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/*
'all' means that this is either an explicit commit issued by
user, or an implicit commit issued by a DDL.
*/
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
THD_TRANS *trans= all ? &thd->transaction.all : &thd->transaction.stmt;
Bug#44672: Assertion failed: thd->transaction.xid_state.xid.is_null() The problem is that when a optimization of read-only transactions (bypass 2-phase commit) was implemented, it removed the code that reseted the XID once a transaction wasn't active anymore: sql/sql_parse.cc: - bzero(&thd->transaction.stmt, sizeof(thd->transaction.stmt)); - if (!thd->active_transaction()) - thd->transaction.xid_state.xid.null(); + thd->transaction.stmt.reset(); This mostly worked fine as the transaction commit and rollback functions (in handler.cc) reset the XID once the transaction is ended. But those functions wouldn't reset the XID in case of a empty transaction, leading to a assertion when a new starting a new XA transaction. The solution is to ensure that the XID state is reset when empty transactions are ended (by either commit or rollback). This is achieved by reorganizing the code so that the transaction cleanup routine is invoked whenever a transaction is ended.
2009-06-06 00:16:54 +02:00
/*
"real" is a nick name for a transaction for which a commit will
make persistent changes. E.g. a 'stmt' transaction inside a 'all'
transation is not 'real': even though it's possible to commit it,
the changes are not durable as they might be rolled back if the
enclosing 'all' transaction is rolled back.
*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
bool is_real_trans= all || thd->transaction.all.ha_list == 0;
Ha_trx_info *ha_info= trans->ha_list;
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
my_xid xid= thd->transaction.xid_state.xid.get_my_xid();
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_commit_trans");
Fix for bug #10015 "Crash in InnoDB if stored routines are used". We should not allow explicit or implicit transaction commits inside of stored functions or triggers (so in autocommit mode we should not do commits after execution of sub-statement). Also since we don't support nested statement transactions in 5.0, we shouldn't commit or rollback stmt transactions while we are inside stored functions or triggers. This should be fixed in later (>=5.1) releases.
2005-06-07 12:53:08 +02:00
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/*
We must not commit the normal transaction if a statement
transaction is pending. Otherwise statement transaction
flags will not get propagated to its normal transaction's
counterpart.
*/
DBUG_ASSERT(thd->transaction.stmt.ha_list == NULL ||
trans == &thd->transaction.stmt);
Added Non-prelocked SP execution: Now a PROCEDURE doesn't enter/leave prelocked mode for its body, but lets each statement to get/release its own locks. This allows a broader set of statements to be executed inside PROCEDUREs (but breaks replication) This patch should fix BUG#8072, BUG#8766, BUG#9563, BUG#11126
2005-07-30 10:19:57 +02:00
if (thd->in_sub_stmt)
Fix for bug #10015 "Crash in InnoDB if stored routines are used". We should not allow explicit or implicit transaction commits inside of stored functions or triggers (so in autocommit mode we should not do commits after execution of sub-statement). Also since we don't support nested statement transactions in 5.0, we shouldn't commit or rollback stmt transactions while we are inside stored functions or triggers. This should be fixed in later (>=5.1) releases.
2005-06-07 12:53:08 +02:00
{
/*
Since we don't support nested statement transactions in 5.0,
we can't commit or rollback stmt transactions while we are inside
stored functions or triggers. So we simply do nothing now.
TODO: This should be fixed in later ( >= 5.1) releases.
*/
if (!all)
DBUG_RETURN(0);
/*
We assume that all statements which commit or rollback main transaction
are prohibited inside of stored functions or triggers. So they should
bail out with error even before ha_commit_trans() call. To be 100% safe
let us throw error in non-debug builds.
*/
DBUG_ASSERT(0);
my_error(ER_COMMIT_NOT_ALLOWED_IN_SF_OR_TRG, MYF(0));
DBUG_RETURN(2);
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
#ifdef USING_TRANSACTIONS
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (ha_info)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
uint rw_ha_count;
bool rw_trans;
Bug#52172 test binlog.binlog_index needs --skip-core-file to avoid leaving core files For crash testing: kill the server without generating core file. include/my_dbug.h Use kill(getpid(), SIGKILL) which cannot be caught by signal handlers. All DBUG_XXX macros should be no-ops in optimized mode, do that for DBUG_ABORT as well. sql/handler.cc Kill server without generating core. sql/log.cc Kill server without generating core.
2010-10-18 13:24:34 +02:00
DBUG_EXECUTE_IF("crash_commit_before", DBUG_SUICIDE(););
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
/* Close all cursors that can not survive COMMIT */
if (is_real_trans) /* not a statement commit */
thd->stmt_map.close_transient_cursors();
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
rw_ha_count= ha_check_and_coalesce_trx_read_only(thd, ha_info, all);
/* rw_trans is TRUE when we in a transaction changing data */
rw_trans= is_real_trans && (rw_ha_count > 0);
if (rw_trans &&
wait_if_global_read_lock(thd, 0, 0))
FLUSH TABLES WITH READ LOCK should block writes to binlog too
2005-04-12 17:15:54 +02:00
{
ha_rollback_trans(thd, all);
DBUG_RETURN(1);
}
WL#3602 (SET GLOBAL READONLY) Bug#11733 (COMMITs should not happen if read-only is set) Bug#22009 (Can write to a read-only server under some circumstances) See the work log for details The change consist of a) acquiring the global read lock in SET GLOBAL READONLY b) honoring opt_readonly in ha_commit_trans(), c) honoring opt_readonly in mysql_lock_tables(). a) takes care of the server stability, b) makes the transactional tables safe (Bug 11733) c) makes the non transactional tables safe (Bug 22009)
2006-11-21 04:40:35 +01:00
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
if (rw_trans &&
opt_readonly &&
!(thd->security_ctx->master_access & SUPER_ACL) &&
!thd->slave_thread)
WL#3602 (SET GLOBAL READONLY) Bug#11733 (COMMITs should not happen if read-only is set) Bug#22009 (Can write to a read-only server under some circumstances) See the work log for details The change consist of a) acquiring the global read lock in SET GLOBAL READONLY b) honoring opt_readonly in ha_commit_trans(), c) honoring opt_readonly in mysql_lock_tables(). a) takes care of the server stability, b) makes the transactional tables safe (Bug 11733) c) makes the non transactional tables safe (Bug 22009)
2006-11-21 04:40:35 +01:00
{
my_error(ER_OPTION_PREVENTS_STATEMENT, MYF(0), "--read-only");
ha_rollback_trans(thd, all);
error= 1;
goto end;
}
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
if (!trans->no_2pc && (rw_ha_count > 1))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (; ha_info && !error; ha_info= ha_info->next())
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
/*
Do not call two-phase commit if this particular
transaction is read-only. This allows for simpler
implementation in engines that are always read-only.
*/
if (! ha_info->is_trx_read_write())
continue;
/*
Sic: we know that prepare() is not NULL since otherwise
trans->no_2pc would have been set.
*/
if ((err= ht->prepare(ht, thd, all)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
my_error(ER_ERROR_DURING_COMMIT, MYF(0), err);
FLUSH TABLES WITH READ LOCK should block writes to binlog too
2005-04-12 17:15:54 +02:00
error= 1;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_prepare_count);
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Bug#52172 test binlog.binlog_index needs --skip-core-file to avoid leaving core files For crash testing: kill the server without generating core file. include/my_dbug.h Use kill(getpid(), SIGKILL) which cannot be caught by signal handlers. All DBUG_XXX macros should be no-ops in optimized mode, do that for DBUG_ABORT as well. sql/handler.cc Kill server without generating core. sql/log.cc Kill server without generating core.
2010-10-18 13:24:34 +02:00
DBUG_EXECUTE_IF("crash_commit_after_prepare", DBUG_SUICIDE(););
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +01:00
if (error || (is_real_trans && xid &&
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
(error= !(cookie= tc_log->log_xid(thd, xid)))))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
ha_rollback_trans(thd, all);
FLUSH TABLES WITH READ LOCK should block writes to binlog too
2005-04-12 17:15:54 +02:00
error= 1;
goto end;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Bug#52172 test binlog.binlog_index needs --skip-core-file to avoid leaving core files For crash testing: kill the server without generating core file. include/my_dbug.h Use kill(getpid(), SIGKILL) which cannot be caught by signal handlers. All DBUG_XXX macros should be no-ops in optimized mode, do that for DBUG_ABORT as well. sql/handler.cc Kill server without generating core. sql/log.cc Kill server without generating core.
2010-10-18 13:24:34 +02:00
DBUG_EXECUTE_IF("crash_commit_after_log", DBUG_SUICIDE(););
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
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
error=ha_commit_one_phase(thd, all) ? (cookie ? 2 : 1) : 0;
Bug#52172 test binlog.binlog_index needs --skip-core-file to avoid leaving core files For crash testing: kill the server without generating core file. include/my_dbug.h Use kill(getpid(), SIGKILL) which cannot be caught by signal handlers. All DBUG_XXX macros should be no-ops in optimized mode, do that for DBUG_ABORT as well. sql/handler.cc Kill server without generating core. sql/log.cc Kill server without generating core.
2010-10-18 13:24:34 +02:00
DBUG_EXECUTE_IF("crash_commit_before_unlog", DBUG_SUICIDE(););
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
if (cookie)
query_id and my_xid -> ulonglong fix for binlog+autocommit+tclog comments, style fixes
2005-01-27 22:38:56 +01:00
tc_log->unlog(cookie, xid);
Bug#52172 test binlog.binlog_index needs --skip-core-file to avoid leaving core files For crash testing: kill the server without generating core file. include/my_dbug.h Use kill(getpid(), SIGKILL) which cannot be caught by signal handlers. All DBUG_XXX macros should be no-ops in optimized mode, do that for DBUG_ABORT as well. sql/handler.cc Kill server without generating core. sql/log.cc Kill server without generating core.
2010-10-18 13:24:34 +02:00
DBUG_EXECUTE_IF("crash_commit_after", DBUG_SUICIDE(););
FLUSH TABLES WITH READ LOCK should block writes to binlog too
2005-04-12 17:15:54 +02:00
end:
Fix for bug #35732: read-only blocks SELECT statements in InnoDB Problem: SELECTs prohibited for a transactional SE in autocommit mode if read_only is set. Fix: allow them.
2008-04-08 07:20:58 +02:00
if (rw_trans)
FLUSH TABLES WITH READ LOCK should block writes to binlog too
2005-04-12 17:15:54 +02:00
start_waiting_global_read_lock(thd);
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Bug#44672: Assertion failed: thd->transaction.xid_state.xid.is_null() The problem is that when a optimization of read-only transactions (bypass 2-phase commit) was implemented, it removed the code that reseted the XID once a transaction wasn't active anymore: sql/sql_parse.cc: - bzero(&thd->transaction.stmt, sizeof(thd->transaction.stmt)); - if (!thd->active_transaction()) - thd->transaction.xid_state.xid.null(); + thd->transaction.stmt.reset(); This mostly worked fine as the transaction commit and rollback functions (in handler.cc) reset the XID once the transaction is ended. But those functions wouldn't reset the XID in case of a empty transaction, leading to a assertion when a new starting a new XA transaction. The solution is to ensure that the XID state is reset when empty transactions are ended (by either commit or rollback). This is achieved by reorganizing the code so that the transaction cleanup routine is invoked whenever a transaction is ended.
2009-06-06 00:16:54 +02:00
/* Free resources and perform other cleanup even for 'empty' transactions. */
else if (is_real_trans)
Fix for bug #26288: savepoint not deleted, comit on empty transaction Problem: commit doesn't delete savepoints if there are no changes in the transaction. Fix: delete them in such cases.
2009-03-27 07:24:32 +01:00
thd->transaction.cleanup();
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
#endif /* USING_TRANSACTIONS */
DBUG_RETURN(error);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@note
This function does not care about global read lock. A caller should.
FLUSH TABLES WITH READ LOCK should block writes to binlog too
2005-04-12 17:15:54 +02:00
*/
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int ha_commit_one_phase(THD *thd, bool all)
{
int error=0;
THD_TRANS *trans=all ? &thd->transaction.all : &thd->transaction.stmt;
Bug#44672: Assertion failed: thd->transaction.xid_state.xid.is_null() The problem is that when a optimization of read-only transactions (bypass 2-phase commit) was implemented, it removed the code that reseted the XID once a transaction wasn't active anymore: sql/sql_parse.cc: - bzero(&thd->transaction.stmt, sizeof(thd->transaction.stmt)); - if (!thd->active_transaction()) - thd->transaction.xid_state.xid.null(); + thd->transaction.stmt.reset(); This mostly worked fine as the transaction commit and rollback functions (in handler.cc) reset the XID once the transaction is ended. But those functions wouldn't reset the XID in case of a empty transaction, leading to a assertion when a new starting a new XA transaction. The solution is to ensure that the XID state is reset when empty transactions are ended (by either commit or rollback). This is achieved by reorganizing the code so that the transaction cleanup routine is invoked whenever a transaction is ended.
2009-06-06 00:16:54 +02:00
/*
"real" is a nick name for a transaction for which a commit will
make persistent changes. E.g. a 'stmt' transaction inside a 'all'
transation is not 'real': even though it's possible to commit it,
the changes are not durable as they might be rolled back if the
enclosing 'all' transaction is rolled back.
*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
bool is_real_trans=all || thd->transaction.all.ha_list == 0;
Ha_trx_info *ha_info= trans->ha_list, *ha_info_next;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_commit_one_phase");
#ifdef USING_TRANSACTIONS
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (ha_info)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (; ha_info; ha_info= ha_info_next)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
if ((err= ht->commit(ht, thd, all)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
my_error(ER_ERROR_DURING_COMMIT, MYF(0), err);
error=1;
}
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_commit_count);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
ha_info_next= ha_info->next();
ha_info->reset(); /* keep it conveniently zero-filled */
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
trans->ha_list= 0;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
trans->no_2pc=0;
if (all)
{
#ifdef HAVE_QUERY_CACHE
if (thd->transaction.changed_tables)
query_cache.invalidate(thd->transaction.changed_tables);
Added NDB storage engine
2004-04-15 09:14:14 +02:00
#endif
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
thd->variables.tx_isolation=thd->session_tx_isolation;
}
}
Bug#45548: XA transaction without access to InnoDB tables crashes the server The problem is that the one phase commit function failed to properly end a empty transaction. The solution is to ensure that the transaction cleanup procedure is invoked even for empty transactions.
2009-06-25 17:25:23 +02:00
/* Free resources and perform other cleanup even for 'empty' transactions. */
if (is_real_trans)
thd->transaction.cleanup();
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
#endif /* USING_TRANSACTIONS */
DBUG_RETURN(error);
}
int ha_rollback_trans(THD *thd, bool all)
{
int error=0;
THD_TRANS *trans=all ? &thd->transaction.all : &thd->transaction.stmt;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Ha_trx_info *ha_info= trans->ha_list, *ha_info_next;
Bug#44672: Assertion failed: thd->transaction.xid_state.xid.is_null() The problem is that when a optimization of read-only transactions (bypass 2-phase commit) was implemented, it removed the code that reseted the XID once a transaction wasn't active anymore: sql/sql_parse.cc: - bzero(&thd->transaction.stmt, sizeof(thd->transaction.stmt)); - if (!thd->active_transaction()) - thd->transaction.xid_state.xid.null(); + thd->transaction.stmt.reset(); This mostly worked fine as the transaction commit and rollback functions (in handler.cc) reset the XID once the transaction is ended. But those functions wouldn't reset the XID in case of a empty transaction, leading to a assertion when a new starting a new XA transaction. The solution is to ensure that the XID state is reset when empty transactions are ended (by either commit or rollback). This is achieved by reorganizing the code so that the transaction cleanup routine is invoked whenever a transaction is ended.
2009-06-06 00:16:54 +02:00
/*
"real" is a nick name for a transaction for which a commit will
make persistent changes. E.g. a 'stmt' transaction inside a 'all'
transation is not 'real': even though it's possible to commit it,
the changes are not durable as they might be rolled back if the
enclosing 'all' transaction is rolled back.
*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
bool is_real_trans=all || thd->transaction.all.ha_list == 0;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_rollback_trans");
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/*
We must not rollback the normal transaction if a statement
transaction is pending.
*/
DBUG_ASSERT(thd->transaction.stmt.ha_list == NULL ||
trans == &thd->transaction.stmt);
Added Non-prelocked SP execution: Now a PROCEDURE doesn't enter/leave prelocked mode for its body, but lets each statement to get/release its own locks. This allows a broader set of statements to be executed inside PROCEDUREs (but breaks replication) This patch should fix BUG#8072, BUG#8766, BUG#9563, BUG#11126
2005-07-30 10:19:57 +02:00
if (thd->in_sub_stmt)
Fix for bug #10015 "Crash in InnoDB if stored routines are used". We should not allow explicit or implicit transaction commits inside of stored functions or triggers (so in autocommit mode we should not do commits after execution of sub-statement). Also since we don't support nested statement transactions in 5.0, we shouldn't commit or rollback stmt transactions while we are inside stored functions or triggers. This should be fixed in later (>=5.1) releases.
2005-06-07 12:53:08 +02:00
{
/*
If we are inside stored function or trigger we should not commit or
rollback current statement transaction. See comment in ha_commit_trans()
call for more information.
*/
if (!all)
DBUG_RETURN(0);
DBUG_ASSERT(0);
my_error(ER_COMMIT_NOT_ALLOWED_IN_SF_OR_TRG, MYF(0));
DBUG_RETURN(1);
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
#ifdef USING_TRANSACTIONS
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (ha_info)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
A fix and a test case for Bug#10760 and complementary cleanups. The idea of the patch is that every cursor gets its own lock id for table level locking. Thus cursors are protected from updates performed within the same connection. Additionally a list of transient (must be closed at commit) cursors is maintained and all transient cursors are closed when necessary. Lastly, this patch adds support for deadlock timeouts to TLL locking when using cursors. + post-review fixes.
2005-07-19 20:21:12 +02:00
/* Close all cursors that can not survive ROLLBACK */
if (is_real_trans) /* not a statement commit */
thd->stmt_map.close_transient_cursors();
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (; ha_info; ha_info= ha_info_next)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
if ((err= ht->rollback(ht, thd, all)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{ // cannot happen
my_error(ER_ERROR_DURING_ROLLBACK, MYF(0), err);
error=1;
}
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_rollback_count);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
ha_info_next= ha_info->next();
ha_info->reset(); /* keep it conveniently zero-filled */
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
trans->ha_list= 0;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
trans->no_2pc=0;
BUG#43171 - Assertion failed: thd->transaction.xid_state.xid.is_null() XA START may cause assertion failure/server crash when it is called after unilateral roll back issued by the Resource Manager (both in regular transaction and after XA transaction). The problem was that rm_error variable wasn't set/reset properly.
2009-10-28 16:39:08 +01:00
if (is_real_trans && thd->transaction_rollback_request &&
thd->transaction.xid_state.xa_state != XA_NOTR)
Bug#44672: Assertion failed: thd->transaction.xid_state.xid.is_null() The problem is that when a optimization of read-only transactions (bypass 2-phase commit) was implemented, it removed the code that reseted the XID once a transaction wasn't active anymore: sql/sql_parse.cc: - bzero(&thd->transaction.stmt, sizeof(thd->transaction.stmt)); - if (!thd->active_transaction()) - thd->transaction.xid_state.xid.null(); + thd->transaction.stmt.reset(); This mostly worked fine as the transaction commit and rollback functions (in handler.cc) reset the XID once the transaction is ended. But those functions wouldn't reset the XID in case of a empty transaction, leading to a assertion when a new starting a new XA transaction. The solution is to ensure that the XID state is reset when empty transactions are ended (by either commit or rollback). This is achieved by reorganizing the code so that the transaction cleanup routine is invoked whenever a transaction is ended.
2009-06-06 00:16:54 +02:00
thd->transaction.xid_state.rm_error= thd->main_da.sql_errno();
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
if (all)
thd->variables.tx_isolation=thd->session_tx_isolation;
}
Fix for bug #26288: savepoint not deleted, comit on empty transaction Problem: commit doesn't delete savepoints if there are no changes in the transaction. Fix: delete them in such cases.
2009-03-27 07:24:32 +01:00
/* Always cleanup. Even if there nht==0. There may be savepoints. */
Bug#44672: Assertion failed: thd->transaction.xid_state.xid.is_null() The problem is that when a optimization of read-only transactions (bypass 2-phase commit) was implemented, it removed the code that reseted the XID once a transaction wasn't active anymore: sql/sql_parse.cc: - bzero(&thd->transaction.stmt, sizeof(thd->transaction.stmt)); - if (!thd->active_transaction()) - thd->transaction.xid_state.xid.null(); + thd->transaction.stmt.reset(); This mostly worked fine as the transaction commit and rollback functions (in handler.cc) reset the XID once the transaction is ended. But those functions wouldn't reset the XID in case of a empty transaction, leading to a assertion when a new starting a new XA transaction. The solution is to ensure that the XID state is reset when empty transactions are ended (by either commit or rollback). This is achieved by reorganizing the code so that the transaction cleanup routine is invoked whenever a transaction is ended.
2009-06-06 00:16:54 +02:00
if (is_real_trans)
Fix for bug #26288: savepoint not deleted, comit on empty transaction Problem: commit doesn't delete savepoints if there are no changes in the transaction. Fix: delete them in such cases.
2009-03-27 07:24:32 +01:00
thd->transaction.cleanup();
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
#endif /* USING_TRANSACTIONS */
Bug#24989: The DEADLOCK error is improperly handled by InnoDB. When innodb detects a deadlock it calls ha_rollback_trans() to rollback the main transaction. But such action isn't allowed from inside of triggers and functions. When it happen the 'Explicit or implicit commit' error is thrown even if there is no commit/rollback statements in the trigger/function. This leads to the user confusion. Now the convert_error_code_to_mysql() function doesn't call the ha_rollback_trans() function directly but rather calls the mark_transaction_to_rollback function and returns an error. The sp_rcontext::find_handler() now doesn't allow errors to be caught by the trigger/function error handlers when the thd->is_fatal_sub_stmt_error flag is set. Procedures are still allowed to catch such errors. The sp_rcontext::find_handler function now accepts a THD handle as a parameter. The transaction_rollback_request and the is_fatal_sub_stmt_error flags are added to the THD class. The are initialized by the THD class constructor. Now the ha_autocommit_or_rollback function rolls back main transaction when not in a sub statement and the thd->transaction_rollback_request is set. The THD::restore_sub_statement_state function now resets the thd->is_fatal_sub_stmt_error flag on exit from a sub-statement.
2007-07-30 15:14:34 +02:00
if (all)
thd->transaction_rollback_request= FALSE;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
/*
If a non-transactional table was updated, warn; don't warn if this is a
slave thread (because when a slave thread executes a ROLLBACK, it has
been read from the binary log, so it's 100% sure and normal to produce
error ER_WARNING_NOT_COMPLETE_ROLLBACK. If we sent the warning to the
slave SQL thread, it would not stop the thread but just be printed in
the error log; but we don't want users to wonder why they have this
message in the error log, so we don't send it.
*/
(pushing for Andrei) Bug #27417 thd->no_trans_update.stmt lost value inside of SF-exec-stack Once had been set the flag might later got reset inside of a stored routine execution stack. The reason was in that there was no check if a new statement started at time of resetting. The artifact affects most of binlogable DML queries. Notice, that multi-update is wrapped up within bug@27716 fix, multi-delete bug@29136. Fixed with saving parent's statement flag of whether the statement modified non-transactional table, and unioning (merging) the value with that was gained in mysql_execute_command. Resettling thd->no_trans_update members into thd->transaction.`member`; Asserting code; Effectively the following properties are held. 1. At the end of a substatement thd->transaction.stmt.modified_non_trans_table reflects the fact if such a table got modified by the substatement. That also respects THD::really_abort_on_warnin() requirements. 2. Eventually thd->transaction.stmt.modified_non_trans_table will be computed as the union of the values of all invoked sub-statements. That fixes this bug#27417; Computing of thd->transaction.all.modified_non_trans_table is refined to base to the stmt's value for all the case including insert .. select statement which before the patch had an extra issue bug@28960. Minor issues are covered with mysql_load, mysql_delete, and binloggin of insert in to temp_table select. The supplied test verifies limitely, mostly asserts. The ultimate testing is defered for bug@13270, bug@23333.
2007-07-30 17:27:36 +02:00
if (is_real_trans && thd->transaction.all.modified_non_trans_table &&
Fixed compiler warnings Fixed compile-pentium64 scripts Fixed wrong estimate of update_with_key_prefix in sql-bench Merge bk-internal.mysql.com:/home/bk/mysql-5.1 into mysql.com:/home/my/mysql-5.1 Fixed unsafe define of uint4korr() Fixed that --extern works with mysql-test-run.pl Small trivial cleanups This also fixes a bug in counting number of rows that are updated when we have many simultanous queries Move all connection handling and command exectuion main loop from sql_parse.cc to sql_connection.cc Split handle_one_connection() into reusable sub functions. Split create_new_thread() into reusable sub functions. Added thread_scheduler; Preliminary interface code for future thread_handling code. Use 'my_thread_id' for internal thread id's Make thr_alarm_kill() to depend on thread_id instead of thread Make thr_abort_locks_for_thread() depend on thread_id instead of thread In store_globals(), set my_thread_var->id to be thd->thread_id. Use my_thread_var->id as basis for my_thread_name() The above changes makes the connection we have between THD and threads more soft. Added a lot of DBUG_PRINT() and DBUG_ASSERT() functions Fixed compiler warnings Fixed core dumps when running with --debug Removed setting of signal masks (was never used) Made event code call pthread_exit() (portability fix) Fixed that event code doesn't call DBUG_xxx functions before my_thread_init() is called. Made handling of thread_id and thd->variables.pseudo_thread_id uniform. Removed one common 'not freed memory' warning from mysqltest Fixed a couple of usage of not initialized warnings (unlikely cases) Suppress compiler warnings from bdb and (for the moment) warnings from ndb
2007-02-23 12:13:55 +01:00
!thd->slave_thread && thd->killed != THD::KILL_CONNECTION)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_WARN,
ER_WARNING_NOT_COMPLETE_ROLLBACK,
ER(ER_WARNING_NOT_COMPLETE_ROLLBACK));
DBUG_RETURN(error);
Added innobase handler files + other updates for innobase
2000-12-06 00:54:17 +01:00
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
This is used to commit or rollback a single statement depending on
the value of error.
@note
Note that if the autocommit is on, then the following call inside
InnoDB will commit or rollback the whole transaction (= the statement). The
autocommit mechanism built into InnoDB is based on counting locks, but if
the user has used LOCK TABLES then that mechanism does not know to do the
commit.
All statements are now run in a sub transaction Give warning if we are using non transactional tables and do ROLLBACK Porting of crash-me and benchmarks to FrontBase
2000-11-24 00:51:18 +01:00
*/
Import changeset
2000-07-31 21:29:14 +02:00
int ha_autocommit_or_rollback(THD *thd, int error)
{
DBUG_ENTER("ha_autocommit_or_rollback");
All statements are now run in a sub transaction Give warning if we are using non transactional tables and do ROLLBACK Porting of crash-me and benchmarks to FrontBase
2000-11-24 00:51:18 +01:00
#ifdef USING_TRANSACTIONS
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (thd->transaction.stmt.ha_list)
Import changeset
2000-07-31 21:29:14 +02:00
{
Merged some functions and removed some unused client functions. Remember UNION for ALTER TABLE Added test for if we are supporting transactions. Don't allow REPLACE to replace a row when we have generated an auto_increment key Fixed bug when using BLOB keys Fixed bug in SET @variable=user.
2001-03-06 14:24:08 +01:00
if (!error)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (ha_commit_trans(thd, 0))
Merged some functions and removed some unused client functions. Remember UNION for ALTER TABLE Added test for if we are supporting transactions. Don't allow REPLACE to replace a row when we have generated an auto_increment key Fixed bug when using BLOB keys Fixed bug in SET @variable=user.
2001-03-06 14:24:08 +01:00
error=1;
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
else
{
(void) ha_rollback_trans(thd, 0);
if (thd->transaction_rollback_request && !thd->in_sub_stmt)
(void) ha_rollback(thd);
}
Revert of the push of bug #20837 due to failing regression tests.
2009-12-01 10:19:51 +01:00
thd->variables.tx_isolation=thd->session_tx_isolation;
Import changeset
2000-07-31 21:29:14 +02:00
}
#endif
DBUG_RETURN(error);
}
Reviewing new pushed code - CHAR() now returns binary string as default - CHAR(X*65536+Y*256+Z) is now equal to CHAR(X,Y,Z) independent of the character set for CHAR() - Test for both ETIMEDOUT and ETIME from pthread_cond_timedwait() (Some old systems returns ETIME and it's safer to test for both values than to try to write a wrapper for each old system) - Fixed new introduced bug in NOT BETWEEN X and X - Ensure we call commit_by_xid or rollback_by_xid for all engines, even if one engine has failed - Use octet2hex() for all conversion of string to hex - Simplify and optimize code
2005-10-11 23:58:22 +02:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
struct xahton_st {
XID *xid;
int result;
};
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
static my_bool xacommit_handlerton(THD *unused1, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
XA (not completely polished out yet)
2005-01-16 13:16:23 +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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (hton->state == SHOW_OPTION_YES && hton->recover)
{
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->commit_by_xid(hton, ((struct xahton_st *)arg)->xid);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
((struct xahton_st *)arg)->result= 0;
}
return FALSE;
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +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
static my_bool xarollback_handlerton(THD *unused1, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
{
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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (hton->state == SHOW_OPTION_YES && hton->recover)
Reviewing new pushed code - CHAR() now returns binary string as default - CHAR(X*65536+Y*256+Z) is now equal to CHAR(X,Y,Z) independent of the character set for CHAR() - Test for both ETIMEDOUT and ETIME from pthread_cond_timedwait() (Some old systems returns ETIME and it's safer to test for both values than to try to write a wrapper for each old system) - Fixed new introduced bug in NOT BETWEEN X and X - Ensure we call commit_by_xid or rollback_by_xid for all engines, even if one engine has failed - Use octet2hex() for all conversion of string to hex - Simplify and optimize code
2005-10-11 23:58:22 +02:00
{
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->rollback_by_xid(hton, ((struct xahton_st *)arg)->xid);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
((struct xahton_st *)arg)->result= 0;
Reviewing new pushed code - CHAR() now returns binary string as default - CHAR(X*65536+Y*256+Z) is now equal to CHAR(X,Y,Z) independent of the character set for CHAR() - Test for both ETIMEDOUT and ETIME from pthread_cond_timedwait() (Some old systems returns ETIME and it's safer to test for both values than to try to write a wrapper for each old system) - Fixed new introduced bug in NOT BETWEEN X and X - Ensure we call commit_by_xid or rollback_by_xid for all engines, even if one engine has failed - Use octet2hex() for all conversion of string to hex - Simplify and optimize code
2005-10-11 23:58:22 +02:00
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return FALSE;
}
int ha_commit_or_rollback_by_xid(XID *xid, bool commit)
{
struct xahton_st xaop;
xaop.xid= xid;
xaop.result= 1;
after review fixes
2006-05-31 18:07:32 +02:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
plugin_foreach(NULL, commit ? xacommit_handlerton : xarollback_handlerton,
MYSQL_STORAGE_ENGINE_PLUGIN, &xaop);
return xaop.result;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Removed compiler warnings
2002-02-09 03:00:24 +01:00
Reviewing new pushed code - CHAR() now returns binary string as default - CHAR(X*65536+Y*256+Z) is now equal to CHAR(X,Y,Z) independent of the character set for CHAR() - Test for both ETIMEDOUT and ETIME from pthread_cond_timedwait() (Some old systems returns ETIME and it's safer to test for both values than to try to write a wrapper for each old system) - Fixed new introduced bug in NOT BETWEEN X and X - Ensure we call commit_by_xid or rollback_by_xid for all engines, even if one engine has failed - Use octet2hex() for all conversion of string to hex - Simplify and optimize code
2005-10-11 23:58:22 +02:00
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
#ifndef DBUG_OFF
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@note
This does not need to be multi-byte safe or anything
*/
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
static char* xid_to_str(char *buf, XID *xid)
{
int i;
char *s=buf;
*s++='\'';
for (i=0; i < xid->gtrid_length+xid->bqual_length; i++)
{
uchar c=(uchar)xid->data[i];
Cleanups during review of code Fixed newly introduced bug in rollup
2005-03-21 22:41:28 +01:00
/* is_next_dig is set if next character is a number */
bool is_next_dig= FALSE;
sql/handler.cc smarter xid-to-str routiine fixed assert crash in XA RECOVER sql/sql_parse.cc XA COMMIT/ROLLBACK did not send_ok in some cases
2005-03-16 08:42:06 +01:00
if (i < XIDDATASIZE)
{
Cleanups during review of code Fixed newly introduced bug in rollup
2005-03-21 22:41:28 +01:00
char ch= xid->data[i+1];
is_next_dig= (ch >= '0' && ch <='9');
sql/handler.cc smarter xid-to-str routiine fixed assert crash in XA RECOVER sql/sql_parse.cc XA COMMIT/ROLLBACK did not send_ok in some cases
2005-03-16 08:42:06 +01:00
}
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
if (i == xid->gtrid_length)
{
*s++='\'';
if (xid->bqual_length)
{
*s++='.';
*s++='\'';
}
}
if (c < 32 || c > 126)
{
*s++='\\';
Cleanups during review of code Fixed newly introduced bug in rollup
2005-03-21 22:41:28 +01:00
/*
If next character is a number, write current character with
3 octal numbers to ensure that the next number is not seen
as part of the octal number
*/
sql/handler.cc smarter xid-to-str routiine fixed assert crash in XA RECOVER sql/sql_parse.cc XA COMMIT/ROLLBACK did not send_ok in some cases
2005-03-16 08:42:06 +01:00
if (c > 077 || is_next_dig)
*s++=_dig_vec_lower[c >> 6];
if (c > 007 || is_next_dig)
*s++=_dig_vec_lower[(c >> 3) & 7];
*s++=_dig_vec_lower[c & 7];
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
}
else
{
if (c == '\'' || c == '\\')
*s++='\\';
*s++=c;
}
}
*s++='\'';
*s=0;
return buf;
}
#endif
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
recover() step of xa.
@note
there are three modes of operation:
- automatic recover after a crash
in this case commit_list != 0, tc_heuristic_recover==0
all xids from commit_list are committed, others are rolled back
- manual (heuristic) recover
in this case commit_list==0, tc_heuristic_recover != 0
DBA has explicitly specified that all prepared transactions should
be committed (or rolled back).
- no recovery (MySQL did not detect a crash)
in this case commit_list==0, tc_heuristic_recover == 0
there should be no prepared transactions in this case.
Removed compiler warnings
2002-02-09 03:00:24 +01:00
*/
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
struct xarecover_st
{
int len, found_foreign_xids, found_my_xids;
XID *list;
HASH *commit_list;
bool dry_run;
};
XA (not completely polished out yet)
2005-01-16 13:16:23 +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
static my_bool xarecover_handlerton(THD *unused, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
{
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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
struct xarecover_st *info= (struct xarecover_st *) arg;
int got;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (hton->state == SHOW_OPTION_YES && hton->recover)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
while ((got= hton->recover(hton, info->list, info->len)) > 0 )
ha_innobase.cc, ha_innobase.h, handler.h, handler.cc, sql_class.h, log.cc: Tell table handlers where a binlog segment for a trx ends
2002-01-22 21:57:56 +01:00
{
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
sql_print_information("Found %d prepared transaction(s) in %s",
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
got, ha_resolve_storage_engine_name(hton));
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
for (int i=0; i < got; i ++)
{
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
my_xid x=info->list[i].get_my_xid();
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
if (!x) // not "mine" - that is generated by external TM
if no xa recovery (or after it): warning on startup if prepared foreign xids error on startup if prepared our xids temporarily: always rollback prepared our xids instead of an error
2005-03-12 20:09:54 +01:00
{
sql/handler.cc smarter xid-to-str routiine fixed assert crash in XA RECOVER sql/sql_parse.cc XA COMMIT/ROLLBACK did not send_ok in some cases
2005-03-16 08:42:06 +01:00
#ifndef DBUG_OFF
char buf[XIDDATASIZE*4+6]; // see xid_to_str
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
sql_print_information("ignore xid %s", xid_to_str(buf, info->list+i));
sql/handler.cc smarter xid-to-str routiine fixed assert crash in XA RECOVER sql/sql_parse.cc XA COMMIT/ROLLBACK did not send_ok in some cases
2005-03-16 08:42:06 +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
xid_cache_insert(info->list+i, XA_PREPARED);
info->found_foreign_xids++;
if no xa recovery (or after it): warning on startup if prepared foreign xids error on startup if prepared our xids temporarily: always rollback prepared our xids instead of an error
2005-03-12 20:09:54 +01:00
continue;
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (info->dry_run)
if no xa recovery (or after it): warning on startup if prepared foreign xids error on startup if prepared our xids temporarily: always rollback prepared our xids instead of an error
2005-03-12 20:09:54 +01:00
{
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
info->found_my_xids++;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
continue;
if no xa recovery (or after it): warning on startup if prepared foreign xids error on startup if prepared our xids temporarily: always rollback prepared our xids instead of an error
2005-03-12 20:09:54 +01:00
}
// recovery mode
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (info->commit_list ?
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
hash_search(info->commit_list, (uchar *)&x, sizeof(x)) != 0 :
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
tc_heuristic_recover == TC_HEURISTIC_RECOVER_COMMIT)
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
{
#ifndef DBUG_OFF
char buf[XIDDATASIZE*4+6]; // see xid_to_str
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
sql_print_information("commit xid %s", xid_to_str(buf, info->list+i));
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
#endif
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->commit_by_xid(hton, info->list+i);
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
else
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
{
#ifndef DBUG_OFF
char buf[XIDDATASIZE*4+6]; // see xid_to_str
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
sql_print_information("rollback xid %s",
xid_to_str(buf, info->list+i));
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
#endif
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->rollback_by_xid(hton, info->list+i);
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (got < info->len)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
break;
ha_innobase.cc, ha_innobase.h, handler.h, handler.cc, sql_class.h, log.cc: Tell table handlers where a binlog segment for a trx ends
2002-01-22 21:57:56 +01:00
}
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return FALSE;
}
int ha_recover(HASH *commit_list)
{
struct xarecover_st info;
DBUG_ENTER("ha_recover");
info.found_foreign_xids= info.found_my_xids= 0;
info.commit_list= commit_list;
info.dry_run= (info.commit_list==0 && tc_heuristic_recover==0);
info.list= NULL;
/* commit_list and tc_heuristic_recover cannot be set both */
DBUG_ASSERT(info.commit_list==0 || tc_heuristic_recover==0);
/* if either is set, total_ha_2pc must be set too */
DBUG_ASSERT(info.dry_run || total_ha_2pc>(ulong)opt_bin_log);
if (total_ha_2pc <= (ulong)opt_bin_log)
DBUG_RETURN(0);
if (info.commit_list)
sql_print_information("Starting crash recovery...");
#ifndef WILL_BE_DELETED_LATER
/*
for now, only InnoDB supports 2pc. It means we can always safely
rollback all pending transactions, without risking inconsistent data
*/
DBUG_ASSERT(total_ha_2pc == (ulong) opt_bin_log+1); // only InnoDB and binlog
tc_heuristic_recover= TC_HEURISTIC_RECOVER_ROLLBACK; // forcing ROLLBACK
info.dry_run=FALSE;
#endif
for (info.len= MAX_XID_LIST_SIZE ;
info.list==0 && info.len > MIN_XID_LIST_SIZE; info.len/=2)
{
info.list=(XID *)my_malloc(info.len*sizeof(XID), MYF(0));
}
if (!info.list)
{
sql_print_error(ER(ER_OUTOFMEMORY), info.len*sizeof(XID));
DBUG_RETURN(1);
}
plugin_foreach(NULL, xarecover_handlerton,
MYSQL_STORAGE_ENGINE_PLUGIN, &info);
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*)info.list, MYF(0));
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (info.found_foreign_xids)
sql_print_warning("Found %d prepared XA transactions",
info.found_foreign_xids);
if (info.dry_run && info.found_my_xids)
if no xa recovery (or after it): warning on startup if prepared foreign xids error on startup if prepared our xids temporarily: always rollback prepared our xids instead of an error
2005-03-12 20:09:54 +01:00
{
sql_print_error("Found %d prepared transactions! It means that mysqld was "
"not shut down properly last time and critical recovery "
"information (last binlog or %s file) was manually deleted "
"after a crash. You have to start mysqld with "
"--tc-heuristic-recover switch to commit or rollback "
"pending transactions.",
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
info.found_my_xids, opt_tc_log_file);
if no xa recovery (or after it): warning on startup if prepared foreign xids error on startup if prepared our xids temporarily: always rollback prepared our xids instead of an error
2005-03-12 20:09:54 +01:00
DBUG_RETURN(1);
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (info.commit_list)
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
sql_print_information("Crash recovery finished.");
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_RETURN(0);
ha_innobase.cc, ha_innobase.h, handler.h, handler.cc, sql_class.h, log.cc: Tell table handlers where a binlog segment for a trx ends
2002-01-22 21:57:56 +01:00
}
All statements are now run in a sub transaction Give warning if we are using non transactional tables and do ROLLBACK Porting of crash-me and benchmarks to FrontBase
2000-11-24 00:51:18 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
return the list of XID's to a client, the same way SHOW commands do.
Many files: Eliminate the LOCK_log bottleneck in group commit in binlogging: flush InnoDB log files only after it has been released
2003-05-03 01:29:40 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
I didn't find in XA specs that an RM cannot return the same XID twice,
so mysql_xa_recover does not filter XID's to ensure uniqueness.
It can be easily fixed later, if necessary.
Many files: Eliminate the LOCK_log bottleneck in group commit in binlogging: flush InnoDB log files only after it has been released
2003-05-03 01:29:40 +02:00
*/
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
bool mysql_xa_recover(THD *thd)
Many files: Eliminate the LOCK_log bottleneck in group commit in binlogging: flush InnoDB log files only after it has been released
2003-05-03 01:29:40 +02:00
{
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
List<Item> field_list;
Protocol *protocol= thd->protocol;
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
int i=0;
XID_STATE *xs;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("mysql_xa_recover");
Polishing: use constants instead of magic numbers.
2007-03-09 06:05:08 +01:00
field_list.push_back(new Item_int("formatID", 0, MY_INT32_NUM_DECIMAL_DIGITS));
field_list.push_back(new Item_int("gtrid_length", 0, MY_INT32_NUM_DECIMAL_DIGITS));
field_list.push_back(new Item_int("bqual_length", 0, MY_INT32_NUM_DECIMAL_DIGITS));
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
field_list.push_back(new Item_empty_string("data",XIDDATASIZE));
if (protocol->send_fields(&field_list,
Protocol::SEND_NUM_ROWS | Protocol::SEND_EOF))
DBUG_RETURN(1);
Many files: Eliminate the LOCK_log bottleneck in group commit in binlogging: flush InnoDB log files only after it has been released
2003-05-03 01:29:40 +02:00
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
pthread_mutex_lock(&LOCK_xid_cache);
After merge fixes
2005-08-25 08:38:06 +02:00
while ((xs= (XID_STATE*)hash_element(&xid_cache, i++)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
if (xs->xa_state==XA_PREPARED)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
protocol->prepare_for_resend();
protocol->store_longlong((longlong)xs->xid.formatID, FALSE);
protocol->store_longlong((longlong)xs->xid.gtrid_length, FALSE);
protocol->store_longlong((longlong)xs->xid.bqual_length, FALSE);
protocol->store(xs->xid.data, xs->xid.gtrid_length+xs->xid.bqual_length,
&my_charset_bin);
if (protocol->write())
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
pthread_mutex_unlock(&LOCK_xid_cache);
DBUG_RETURN(1);
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
}
}
Many files: Eliminate the LOCK_log bottleneck in group commit in binlogging: flush InnoDB log files only after it has been released
2003-05-03 01:29:40 +02:00
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
pthread_mutex_unlock(&LOCK_xid_cache);
Rename send_eof() to my_eof() for consistency with my_ok() and my_error()
2008-02-19 13:58:08 +01:00
my_eof(thd);
BUG#12162 - one can start two transactions with the same XID. Now we keep all active XID's in a hash
2005-08-12 21:15:01 +02:00
DBUG_RETURN(0);
Many files: Eliminate the LOCK_log bottleneck in group commit in binlogging: flush InnoDB log files only after it has been released
2003-05-03 01:29:40 +02:00
}
All statements are now run in a sub transaction Give warning if we are using non transactional tables and do ROLLBACK Porting of crash-me and benchmarks to FrontBase
2000-11-24 00:51:18 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@details
ha_innodb.h, ha_innodb.cc, handler.h, handler.cc, sql_class.cc: Fix a hang on the adaptive hash S-latch if an application program uses mysql_use_result() and performs queries on two connections at the same time
2003-03-03 18:31:01 +01:00
This function should be called when MySQL sends rows of a SELECT result set
or the EOF mark to the client. It releases a possible adaptive hash index
S-latch held by thd in InnoDB and also releases a possible InnoDB query
FIFO ticket to enter InnoDB. To save CPU time, InnoDB allows a thd to
keep them over several calls of the InnoDB handler interface when a join
is executed. But when we let the control to pass to the client they have
to be released because if the application program uses mysql_use_result(),
it may deadlock on the S-latch if the application on another connection
performs another SQL query. In MySQL-4.1 this is even more important because
there a connection can have several SELECT queries open at the same time.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param thd the thread handle of the current connection
@return
always 0
ha_innodb.h, ha_innodb.cc, handler.h, handler.cc, sql_class.cc: Fix a hang on the adaptive hash S-latch if an application program uses mysql_use_result() and performs queries on two connections at the same time
2003-03-03 18:31:01 +01:00
*/
Fix to remove engine dependency from handler.cc file, the last one for functions.
2006-03-26 11:32:52 +02:00
ha_innodb.h, ha_innodb.cc, handler.h, handler.cc, sql_class.cc: Fix a hang on the adaptive hash S-latch if an application program uses mysql_use_result() and performs queries on two connections at the same time
2003-03-03 18:31:01 +01:00
int ha_release_temporary_latches(THD *thd)
{
Bug#34409 LOCK_plugin contention via ha_release_temporary_latches/plugin_foreach use thread-local data structures in ha_release_temporary_latches()
2008-07-21 12:01:22 +02:00
Ha_trx_info *info;
Fix to remove engine dependency from handler.cc file, the last one for functions.
2006-03-26 11:32:52 +02:00
Bug#34409 LOCK_plugin contention via ha_release_temporary_latches/plugin_foreach use thread-local data structures in ha_release_temporary_latches()
2008-07-21 12:01:22 +02:00
/*
Note that below we assume that only transactional storage engines
may need release_temporary_latches(). If this will ever become false,
we could iterate on thd->open_tables instead (and remove duplicates
as if (!seen[hton->slot]) { seen[hton->slot]=1; ... }).
*/
for (info= thd->transaction.stmt.ha_list; info; info= info->next())
{
handlerton *hton= info->ht();
if (hton && hton->release_temporary_latches)
hton->release_temporary_latches(hton, thd);
}
Compile fix: An "int" function had accidentally lost its "return 0;", re-add it.
2006-01-06 12:21:15 +01:00
return 0;
ha_innodb.h, ha_innodb.cc, handler.h, handler.cc, sql_class.cc: Fix a hang on the adaptive hash S-latch if an application program uses mysql_use_result() and performs queries on two connections at the same time
2003-03-03 18:31:01 +01:00
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int ha_rollback_to_savepoint(THD *thd, SAVEPOINT *sv)
Import changeset
2000-07-31 21:29:14 +02:00
{
int error=0;
Fix for bug #13825 "Triggers: crash if release savepoint" and for general handling of savepoints in stored routines. Fixed ha_rollback_to_savepoint()/ha_savepoint()/ha_release_savepoint() functions to properly handle savepoints inside of stored functions and triggers. Also now when we invoke stored function or trigger we create new savepoint level. We destroy it at the end of function/trigger execution and return back to old savepoint level.
2005-11-19 13:09:23 +01:00
THD_TRANS *trans= (thd->in_sub_stmt ? &thd->transaction.stmt :
&thd->transaction.all);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Ha_trx_info *ha_info, *ha_info_next;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_rollback_to_savepoint");
Lots of code fixes to the replication code (especially the binary logging and index log file handling) Fixed bugs in my last changeset that made MySQL hard to compile. Added mutex around some data that could cause table cache corruptions when using OPTIMIZE TABLE / REPAIR TABLE or automatic repair of MyISAM tables. Added mutex around some data in the slave start/stop code that could cause THD linked list corruptions Extended my_chsize() to allow one to specify a filler character. Extend vio_blocking to return the old state (This made some usage of this function much simpler) Added testing for some functions that they caller have got the required mutexes before calling the function. Use setrlimit() to ensure that we can write core file if one specifies --core-file. Added --slave-compressed-protocol Made 2 the minimum length for ft_min_word_len Added variables foreign_key_checks & unique_checks. Less logging from replication code (if not started with --log-warnings) Changed that SHOW INNODB STATUS requre the SUPER privilege More DBUG statements and a lot of new code comments
2002-08-08 02:12:02 +02:00
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
trans->no_2pc=0;
/*
rolling back to savepoint in all storage engines that were part of the
transaction when the savepoint was set
*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (ha_info= sv->ha_list; ha_info; ha_info= ha_info->next())
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
DBUG_ASSERT(ht);
DBUG_ASSERT(ht->savepoint_set != 0);
if ((err= ht->savepoint_rollback(ht, thd,
(uchar *)(sv+1)+ht->savepoint_offset)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{ // cannot happen
my_error(ER_ERROR_DURING_ROLLBACK, MYF(0), err);
error=1;
removed unneeded table name pointer transaction-live memory cleanup moved to ha_commit/ha_rollback fixed query cache validator to work correctly on systems that allocate unaligned dat
2002-06-08 23:58:05 +02:00
}
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_savepoint_rollback_count);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
trans->no_2pc|= ht->prepare == 0;
Merged some functions and removed some unused client functions. Remember UNION for ALTER TABLE Added test for if we are supporting transactions. Don't allow REPLACE to replace a row when we have generated an auto_increment key Fixed bug when using BLOB keys Fixed bug in SET @variable=user.
2001-03-06 14:24:08 +01:00
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
/*
rolling back the transaction in all storage engines that were not part of
the transaction when the savepoint was set
*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (ha_info= trans->ha_list; ha_info != sv->ha_list;
ha_info= ha_info_next)
Import changeset
2000-07-31 21:29:14 +02:00
{
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
if ((err= ht->rollback(ht, thd, !thd->in_sub_stmt)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{ // cannot happen
my_error(ER_ERROR_DURING_ROLLBACK, MYF(0), err);
error=1;
2 minor edits, plus fix for BUG#1113 "INSERT into non-trans table SELECT ; ROLLBACK" does not send warning" and fix for BUG#873 "In transaction, INSERT to non-trans table is written too early to binlog". Now we don't always write the non-trans update immediately to the binlog; if there is something in the binlog cache we write it to the binlog cache (because the non-trans update could depend on a trans table which was modified earlier in the transaction); then in case of ROLLBACK, we write the binlog cache to the binlog, wrapped with BEGIN/ROLLBACK. This guarantees that the slave does the same updates. For ROLLBACK TO SAVEPOINT: when we execute a SAVEPOINT command we write it to the binlog cache. At ROLLBACK TO SAVEPOINT, if some non-trans table was updated, we write ROLLBACK TO SAVEPOINT to the binlog cache; when the transaction terminates (COMMIT/ROLLBACK), the binlog cache will be flushed to the binlog (because of the non-trans update) so we'll have SAVEPOINT and ROLLBACK TO SAVEPOINT in the binlog. Apart from this rare case of updates of mixed table types in transaction, the usual way is still clear the binlog cache at ROLLBACK, or chop it at ROLLBACK TO SAVEPOINT (meaning the SAVEPOINT command is also chopped, which is fine). Note that BUG#873 encompasses subbugs 1) and 2) of BUG#333 "3 binlogging bugs when doing INSERT with mixed InnoDB/MyISAM".
2003-08-22 15:39:24 +02:00
}
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_rollback_count);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
ha_info_next= ha_info->next();
ha_info->reset(); /* keep it conveniently zero-filled */
Merged some functions and removed some unused client functions. Remember UNION for ALTER TABLE Added test for if we are supporting transactions. Don't allow REPLACE to replace a row when we have generated an auto_increment key Fixed bug when using BLOB keys Fixed bug in SET @variable=user.
2001-03-06 14:24:08 +01:00
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
trans->ha_list= sv->ha_list;
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(error);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@note
according to the sql standard (ISO/IEC 9075-2:2003)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
section "4.33.4 SQL-statements and transaction states",
SAVEPOINT is *not* transaction-initiating SQL-statement
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
*/
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int ha_savepoint(THD *thd, SAVEPOINT *sv)
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
{
int error=0;
Fix for bug #13825 "Triggers: crash if release savepoint" and for general handling of savepoints in stored routines. Fixed ha_rollback_to_savepoint()/ha_savepoint()/ha_release_savepoint() functions to properly handle savepoints inside of stored functions and triggers. Also now when we invoke stored function or trigger we create new savepoint level. We destroy it at the end of function/trigger execution and return back to old savepoint level.
2005-11-19 13:09:23 +01:00
THD_TRANS *trans= (thd->in_sub_stmt ? &thd->transaction.stmt :
&thd->transaction.all);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Ha_trx_info *ha_info= trans->ha_list;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_savepoint");
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
#ifdef USING_TRANSACTIONS
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (; ha_info; ha_info= ha_info->next())
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
{
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
DBUG_ASSERT(ht);
if (! ht->savepoint_set)
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
{
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
my_error(ER_CHECK_NOT_IMPLEMENTED, MYF(0), "SAVEPOINT");
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
error=1;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
break;
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if ((err= ht->savepoint_set(ht, thd, (uchar *)(sv+1)+ht->savepoint_offset)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{ // cannot happen
my_error(ER_GET_ERRNO, MYF(0), err);
WL#1967 Support for COMMIT/ROLLBACK optional arguments
2005-02-01 20:48:05 +01:00
error=1;
}
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_savepoint_count);
WL#1967 Support for COMMIT/ROLLBACK optional arguments
2005-02-01 20:48:05 +01:00
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/*
Remember the list of registered storage engines. All new
engines are prepended to the beginning of the list.
*/
sv->ha_list= trans->ha_list;
WL#1967 Support for COMMIT/ROLLBACK optional arguments
2005-02-01 20:48:05 +01:00
#endif /* USING_TRANSACTIONS */
DBUG_RETURN(error);
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int ha_release_savepoint(THD *thd, SAVEPOINT *sv)
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
{
int error=0;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
Ha_trx_info *ha_info= sv->ha_list;
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
DBUG_ENTER("ha_release_savepoint");
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
for (; ha_info; ha_info= ha_info->next())
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
{
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
int err;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
handlerton *ht= ha_info->ht();
/* Savepoint life time is enclosed into transaction life time. */
DBUG_ASSERT(ht);
if (!ht->savepoint_release)
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
continue;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if ((err= ht->savepoint_release(ht, thd,
(uchar *)(sv+1) + ht->savepoint_offset)))
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
{ // cannot happen
my_error(ER_GET_ERRNO, MYF(0), err);
error=1;
2 minor edits, plus fix for BUG#1113 "INSERT into non-trans table SELECT ; ROLLBACK" does not send warning" and fix for BUG#873 "In transaction, INSERT to non-trans table is written too early to binlog". Now we don't always write the non-trans update immediately to the binlog; if there is something in the binlog cache we write it to the binlog cache (because the non-trans update could depend on a trans table which was modified earlier in the transaction); then in case of ROLLBACK, we write the binlog cache to the binlog, wrapped with BEGIN/ROLLBACK. This guarantees that the slave does the same updates. For ROLLBACK TO SAVEPOINT: when we execute a SAVEPOINT command we write it to the binlog cache. At ROLLBACK TO SAVEPOINT, if some non-trans table was updated, we write ROLLBACK TO SAVEPOINT to the binlog cache; when the transaction terminates (COMMIT/ROLLBACK), the binlog cache will be flushed to the binlog (because of the non-trans update) so we'll have SAVEPOINT and ROLLBACK TO SAVEPOINT in the binlog. Apart from this rare case of updates of mixed table types in transaction, the usual way is still clear the binlog cache at ROLLBACK, or chop it at ROLLBACK TO SAVEPOINT (meaning the SAVEPOINT command is also chopped, which is fine). Note that BUG#873 encompasses subbugs 1) and 2) of BUG#333 "3 binlogging bugs when doing INSERT with mixed InnoDB/MyISAM".
2003-08-22 15:39:24 +02:00
}
Many files: Merge InnoDB-4.0.14: SAVEPOINT now implemented; InnoDB now accepts also column prefix keys; crashing bug in ON UPDATE CASCADE fixed; page checksum formula fixed
2003-06-15 00:04:28 +02:00
}
DBUG_RETURN(error);
}
WL#1596 "make mysqldump --master-data --single-transaction able to do online dump of InnoDB AND report reliable binlog coordinates corresponding to the dump". The good news is that now mysqldump can be used to get an online backup of InnoDB *which works for point-in-time recovery and replication slave creation*. Formerly, mysqldump --master-data --single-transaction used to call in fact mysqldump --master-data, so the dump was not an online dump (took big lock all time of dump). The only lock which is now taken in this patch is at the beginning of the dump: mysqldump does: FLUSH TABLES WITH READ LOCK; START TRANSACTION WITH CONSISTENT SNAPSHOT; SHOW MASTER STATUS; UNLOCK TABLES; so the lock time is in fact the time FLUSH TABLES WITH READ LOCK takes to return (can be 0 or very long, if a table is undergoing a huge update). I have done some more minor changes listed in the paragraph of mysqldump.c. WL#2237 "WITH CONSISTENT SNAPSHOT clause for START TRANSACTION": it's a START TRANSACTION which additionally starts a consistent read on all capable storage engine (i.e. InnoDB). So, can serve as a replacement for BEGIN; SELECT * FROM some_innodb_table LIMIT 1; which starts a consistent read too.
2004-11-10 17:56:45 +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
static my_bool snapshot_handlerton(THD *thd, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
{
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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (hton->state == SHOW_OPTION_YES &&
hton->start_consistent_snapshot)
{
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->start_consistent_snapshot(hton, thd);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
*((bool *)arg)= false;
}
return FALSE;
}
WL#1596 "make mysqldump --master-data --single-transaction able to do online dump of InnoDB AND report reliable binlog coordinates corresponding to the dump". The good news is that now mysqldump can be used to get an online backup of InnoDB *which works for point-in-time recovery and replication slave creation*. Formerly, mysqldump --master-data --single-transaction used to call in fact mysqldump --master-data, so the dump was not an online dump (took big lock all time of dump). The only lock which is now taken in this patch is at the beginning of the dump: mysqldump does: FLUSH TABLES WITH READ LOCK; START TRANSACTION WITH CONSISTENT SNAPSHOT; SHOW MASTER STATUS; UNLOCK TABLES; so the lock time is in fact the time FLUSH TABLES WITH READ LOCK takes to return (can be 0 or very long, if a table is undergoing a huge update). I have done some more minor changes listed in the paragraph of mysqldump.c. WL#2237 "WITH CONSISTENT SNAPSHOT clause for START TRANSACTION": it's a START TRANSACTION which additionally starts a consistent read on all capable storage engine (i.e. InnoDB). So, can serve as a replacement for BEGIN; SELECT * FROM some_innodb_table LIMIT 1; which starts a consistent read too.
2004-11-10 17:56:45 +01:00
int ha_start_consistent_snapshot(THD *thd)
{
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
bool warn= true;
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
plugin_foreach(thd, snapshot_handlerton, MYSQL_STORAGE_ENGINE_PLUGIN, &warn);
WL#1596 "make mysqldump --master-data --single-transaction able to do online dump of InnoDB AND report reliable binlog coordinates corresponding to the dump". The good news is that now mysqldump can be used to get an online backup of InnoDB *which works for point-in-time recovery and replication slave creation*. Formerly, mysqldump --master-data --single-transaction used to call in fact mysqldump --master-data, so the dump was not an online dump (took big lock all time of dump). The only lock which is now taken in this patch is at the beginning of the dump: mysqldump does: FLUSH TABLES WITH READ LOCK; START TRANSACTION WITH CONSISTENT SNAPSHOT; SHOW MASTER STATUS; UNLOCK TABLES; so the lock time is in fact the time FLUSH TABLES WITH READ LOCK takes to return (can be 0 or very long, if a table is undergoing a huge update). I have done some more minor changes listed in the paragraph of mysqldump.c. WL#2237 "WITH CONSISTENT SNAPSHOT clause for START TRANSACTION": it's a START TRANSACTION which additionally starts a consistent read on all capable storage engine (i.e. InnoDB). So, can serve as a replacement for BEGIN; SELECT * FROM some_innodb_table LIMIT 1; which starts a consistent read too.
2004-11-10 17:56:45 +01:00
/*
Same idea as when one wants to CREATE TABLE in one engine which does not
exist:
*/
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
if (warn)
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_WARN, ER_UNKNOWN_ERROR,
"This MySQL server does not support any "
"consistent-read capable storage engine");
WL#1596 "make mysqldump --master-data --single-transaction able to do online dump of InnoDB AND report reliable binlog coordinates corresponding to the dump". The good news is that now mysqldump can be used to get an online backup of InnoDB *which works for point-in-time recovery and replication slave creation*. Formerly, mysqldump --master-data --single-transaction used to call in fact mysqldump --master-data, so the dump was not an online dump (took big lock all time of dump). The only lock which is now taken in this patch is at the beginning of the dump: mysqldump does: FLUSH TABLES WITH READ LOCK; START TRANSACTION WITH CONSISTENT SNAPSHOT; SHOW MASTER STATUS; UNLOCK TABLES; so the lock time is in fact the time FLUSH TABLES WITH READ LOCK takes to return (can be 0 or very long, if a table is undergoing a huge update). I have done some more minor changes listed in the paragraph of mysqldump.c. WL#2237 "WITH CONSISTENT SNAPSHOT clause for START TRANSACTION": it's a START TRANSACTION which additionally starts a consistent read on all capable storage engine (i.e. InnoDB). So, can serve as a replacement for BEGIN; SELECT * FROM some_innodb_table LIMIT 1; which starts a consistent read too.
2004-11-10 17:56:45 +01:00
return 0;
}
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
static my_bool flush_handlerton(THD *thd, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
Import changeset
2000-07-31 21:29:14 +02:00
{
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
handlerton *hton= plugin_data(plugin, handlerton *);
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
if (hton->state == SHOW_OPTION_YES && hton->flush_logs &&
hton->flush_logs(hton))
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return TRUE;
return FALSE;
}
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
bool ha_flush_logs(handlerton *db_type)
{
if (db_type == NULL)
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
{
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (plugin_foreach(NULL, flush_handlerton,
MYSQL_STORAGE_ENGINE_PLUGIN, 0))
return TRUE;
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
else
{
if (db_type->state != SHOW_OPTION_YES ||
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
(db_type->flush_logs && db_type->flush_logs(db_type)))
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return TRUE;
}
return FALSE;
Import changeset
2000-07-31 21:29:14 +02:00
}
BUG#45638 - Create temporary table with engine innodb fails Create temporary InnoDB table fails on case insensitive filesystems, when lower_case_table_names is 2 (e.g. OS X) and temporary directory path contains upper case letters. The problem was that tmpdir prefix was converted to lower case when table was created, but was passed as is when table was opened. Fixed by leaving tmpdir prefix part intact.
2009-09-09 11:38:50 +02:00
/**
@brief make canonical filename
@param[in] file table handler
@param[in] path original path
@param[out] tmp_path buffer for canonized path
@details Lower case db name and table name path parts for
non file based tables when lower_case_table_names
is 2 (store as is, compare in lower case).
Filesystem path prefix (mysql_data_home or tmpdir)
is left intact.
@note tmp_path may be left intact if no conversion was
performed.
@retval canonized path
@todo This may be done more efficiently when table path
gets built. Convert this function to something like
ASSERT_CANONICAL_FILENAME.
*/
Bug#37402: Mysql cant read partitioned table with capital letter in the name Problem was that ha_partition had HA_FILE_BASED flag set (since it uses a .par file), but after open it uses the first partitions flags, which results in different case handling for create and for open. Solution was to change the underlying partition name so it was consistent. (Only happens when lower_case_table_names = 2, i.e. Mac OS X and storage engines without HA_FILE_BASED, like InnoDB and Memory.) (Recommit after adding rename of check_lowercase_names to get_canonical_filename, and moved it from handler.h to mysql_priv.h) NOTE: if a mixed case name for a partitioned table was created when lower_case_table_name = 2 it should be renamed or dropped before using the updated version (See bug#37402 for more info)
2008-07-11 01:14:13 +02:00
const char *get_canonical_filename(handler *file, const char *path,
char *tmp_path)
bug #28309 First insert violates UNIQUE constraint - was "memory" table empty? If we have lower_case_table_names == 2 (usually on case insensitive file systems) we sometimes make 'homedir' part of the path sent to the handler into lowercase. So in this case HEAP engine couldn't properly find and remove HP_SHARE, what caused the bug.
2007-06-07 11:03:43 +02:00
{
BUG#45638 - Create temporary table with engine innodb fails Create temporary InnoDB table fails on case insensitive filesystems, when lower_case_table_names is 2 (e.g. OS X) and temporary directory path contains upper case letters. The problem was that tmpdir prefix was converted to lower case when table was created, but was passed as is when table was opened. Fixed by leaving tmpdir prefix part intact.
2009-09-09 11:38:50 +02:00
uint i;
bug #28309 First insert violates UNIQUE constraint - was "memory" table empty? If we have lower_case_table_names == 2 (usually on case insensitive file systems) we sometimes make 'homedir' part of the path sent to the handler into lowercase. So in this case HEAP engine couldn't properly find and remove HP_SHARE, what caused the bug.
2007-06-07 11:03:43 +02:00
if (lower_case_table_names != 2 || (file->ha_table_flags() & HA_FILE_BASED))
return path;
BUG#45638 - Create temporary table with engine innodb fails Create temporary InnoDB table fails on case insensitive filesystems, when lower_case_table_names is 2 (e.g. OS X) and temporary directory path contains upper case letters. The problem was that tmpdir prefix was converted to lower case when table was created, but was passed as is when table was opened. Fixed by leaving tmpdir prefix part intact.
2009-09-09 11:38:50 +02:00
for (i= 0; i <= mysql_tmpdir_list.max; i++)
{
if (is_prefix(path, mysql_tmpdir_list.list[i]))
return path;
}
bug #28309 First insert violates UNIQUE constraint - was "memory" table empty? If we have lower_case_table_names == 2 (usually on case insensitive file systems) we sometimes make 'homedir' part of the path sent to the handler into lowercase. So in this case HEAP engine couldn't properly find and remove HP_SHARE, what caused the bug.
2007-06-07 11:03:43 +02:00
/* Ensure that table handler get path in lower case */
if (tmp_path != path)
strmov(tmp_path, path);
/*
we only should turn into lowercase database/table part
so start the process after homedirectory
*/
my_casedn_str(files_charset_info, tmp_path + mysql_data_home_len);
return tmp_path;
}
In ha_delete_table, use a standard mechanism to intercept the error message and convert it to a warning instead of direct manipulation with the thread error stack. Fix a bug in handler::print_erorr when a garbled message was printed for HA_ERR_NO_SUCH_TABLE. This is a pre-requisite patch for the fix for Bug#12713 Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem
2007-10-30 20:35:14 +01:00
/**
An interceptor to hijack the text of the error message without
setting an error in the thread. We need the text to present it
in the form of a warning to the user.
*/
struct Ha_delete_table_error_handler: public Internal_error_handler
{
public:
virtual bool handle_error(uint sql_errno,
const char *message,
MYSQL_ERROR::enum_warning_level level,
THD *thd);
char buff[MYSQL_ERRMSG_SIZE];
};
bool
Ha_delete_table_error_handler::
handle_error(uint sql_errno,
const char *message,
MYSQL_ERROR::enum_warning_level level,
THD *thd)
{
/* Grab the error message */
strmake(buff, message, sizeof(buff)-1);
return TRUE;
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
Lots of fixes for BDB tables Change DROP TABLE to first drop the data, then the .frm file
2000-12-08 16:04:57 +01:00
This should return ENOENT if the file doesn't exists.
The .frm file will be deleted only if we return 0 or ENOENT
*/
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
int ha_delete_table(THD *thd, handlerton *table_type, const char *path,
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
const char *db, const char *alias, bool generate_warning)
Import changeset
2000-07-31 21:29:14 +02:00
{
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
handler *file;
Some small portability fixes. Added support for lower_case_table_names=2, which is to be used on case insensitive file systems. This tells MySQL to preserve the used case of filenames and database names to make it esier to move files between cases sensitive can case insensitive file systems (like Windows and Linux)
2003-12-30 12:14:21 +01:00
char tmp_path[FN_REFLEN];
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
int error;
TABLE dummy_table;
TABLE_SHARE dummy_share;
DBUG_ENTER("ha_delete_table");
bzero((char*) &dummy_table, sizeof(dummy_table));
bzero((char*) &dummy_share, sizeof(dummy_share));
dummy_table.s= &dummy_share;
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
/* DB_TYPE_UNKNOWN is used in ALTER TABLE when renaming only .frm files */
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (table_type == NULL ||
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
! (file=get_new_handler((TABLE_SHARE*)0, thd->mem_root, table_type)))
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
DBUG_RETURN(ENOENT);
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
Bug#37402: Mysql cant read partitioned table with capital letter in the name Problem was that ha_partition had HA_FILE_BASED flag set (since it uses a .par file), but after open it uses the first partitions flags, which results in different case handling for create and for open. Solution was to change the underlying partition name so it was consistent. (Only happens when lower_case_table_names = 2, i.e. Mac OS X and storage engines without HA_FILE_BASED, like InnoDB and Memory.) (Recommit after adding rename of check_lowercase_names to get_canonical_filename, and moved it from handler.h to mysql_priv.h) NOTE: if a mixed case name for a partitioned table was created when lower_case_table_name = 2 it should be renamed or dropped before using the updated version (See bug#37402 for more info)
2008-07-11 01:14:13 +02:00
path= get_canonical_filename(file, path, tmp_path);
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
if ((error= file->ha_delete_table(path)) && generate_warning)
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
{
/*
Because file->print_error() use my_error() to generate the error message
In ha_delete_table, use a standard mechanism to intercept the error message and convert it to a warning instead of direct manipulation with the thread error stack. Fix a bug in handler::print_erorr when a garbled message was printed for HA_ERR_NO_SUCH_TABLE. This is a pre-requisite patch for the fix for Bug#12713 Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem
2007-10-30 20:35:14 +01:00
we use an internal error handler to intercept it and store the text
in a temporary buffer. Later the message will be presented to user
as a warning.
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
*/
In ha_delete_table, use a standard mechanism to intercept the error message and convert it to a warning instead of direct manipulation with the thread error stack. Fix a bug in handler::print_erorr when a garbled message was printed for HA_ERR_NO_SUCH_TABLE. This is a pre-requisite patch for the fix for Bug#12713 Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem
2007-10-30 20:35:14 +01:00
Ha_delete_table_error_handler ha_delete_table_error_handler;
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
/* Fill up strucutures that print_error may need */
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
dummy_share.path.str= (char*) path;
dummy_share.path.length= strlen(path);
dummy_share.db.str= (char*) db;
dummy_share.db.length= strlen(db);
dummy_share.table_name.str= (char*) alias;
dummy_share.table_name.length= strlen(alias);
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
dummy_table.alias= alias;
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
file->change_table_ptr(&dummy_table, &dummy_share);
In ha_delete_table, use a standard mechanism to intercept the error message and convert it to a warning instead of direct manipulation with the thread error stack. Fix a bug in handler::print_erorr when a garbled message was printed for HA_ERR_NO_SUCH_TABLE. This is a pre-requisite patch for the fix for Bug#12713 Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem
2007-10-30 20:35:14 +01:00
thd->push_internal_handler(&ha_delete_table_error_handler);
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
file->print_error(error, 0);
In ha_delete_table, use a standard mechanism to intercept the error message and convert it to a warning instead of direct manipulation with the thread error stack. Fix a bug in handler::print_erorr when a garbled message was printed for HA_ERR_NO_SUCH_TABLE. This is a pre-requisite patch for the fix for Bug#12713 Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem
2007-10-30 20:35:14 +01:00
thd->pop_internal_handler();
/*
XXX: should we convert *all* errors to warnings here?
What if the error is fatal?
*/
push_warning(thd, MYSQL_ERROR::WARN_LEVEL_ERROR, error,
ha_delete_table_error_handler.buff);
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
delete file;
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
DBUG_RETURN(error);
Import changeset
2000-07-31 21:29:14 +02:00
}
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
/****************************************************************************
** General handler functions
****************************************************************************/
BUG#20256 - LOCK WRITE - MyISAM Only MyISAM tables locked with LOCK TABLES ... WRITE were affected. A query that is optimized with index_merge doesn't reflect rows inserted within LOCK TABLES. MyISAM doesn't flush a state within LOCK TABLES. index_merge optimization creates a copy of the handler, which thus gets outdated MyISAM state. New handler->clone() method is introduced to fix this problem. For non-MyISAM storage engines it allocates a handler and opens it with ha_open(). For MyISAM it additionally copies MyISAM state pointer to cloned handler.
2006-09-12 15:25:35 +02:00
handler *handler::clone(MEM_ROOT *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
handler *new_handler= get_new_handler(table->s, mem_root, table->s->db_type());
BUG#27732 "Possible memory leak with index_merge" The bug was that handler::clone/handler::ha_open() call caused allocation of cloned_copy->ref on the handler->table->mem_root. The allocated memory could not be reclaimed until the table is flushed, so it was possible to exhaust memory by repeatedly running index_merge queries without doing table flushes. The fix: - make handler::clone() allocate new_handler->ref on the passed mem_root - make handler::ha_open() not allocate this->ref if it has already been allocated There is no testcase as it is not possible to check small leaks from testsuite.
2008-02-07 03:10:04 +01:00
/*
Allocate handler->ref here because otherwise ha_open will allocate it
on this->table->mem_root and we will not be able to reclaim that memory
when the clone handler object is destroyed.
*/
Merge kaamos.(none):/data/src/opt/mysql-5.0-opt into kaamos.(none):/data/src/opt/mysql-5.1-opt
2008-02-13 10:12:00 +01:00
if (!(new_handler->ref= (uchar*) alloc_root(mem_root, ALIGN_SIZE(ref_length)*2)))
BUG#27732 "Possible memory leak with index_merge" The bug was that handler::clone/handler::ha_open() call caused allocation of cloned_copy->ref on the handler->table->mem_root. The allocated memory could not be reclaimed until the table is flushed, so it was possible to exhaust memory by repeatedly running index_merge queries without doing table flushes. The fix: - make handler::clone() allocate new_handler->ref on the passed mem_root - make handler::ha_open() not allocate this->ref if it has already been allocated There is no testcase as it is not possible to check small leaks from testsuite.
2008-02-07 03:10:04 +01:00
return NULL;
After merge fix.
2006-09-12 18:38:57 +02:00
if (new_handler && !new_handler->ha_open(table,
table->s->normalized_path.str,
table->db_stat,
BUG#20256 - LOCK WRITE - MyISAM Only MyISAM tables locked with LOCK TABLES ... WRITE were affected. A query that is optimized with index_merge doesn't reflect rows inserted within LOCK TABLES. MyISAM doesn't flush a state within LOCK TABLES. index_merge optimization creates a copy of the handler, which thus gets outdated MyISAM state. New handler->clone() method is introduced to fix this problem. For non-MyISAM storage engines it allocates a handler and opens it with ha_open(). For MyISAM it additionally copies MyISAM state pointer to cloned handler.
2006-09-12 15:25:35 +02:00
HA_OPEN_IGNORE_IF_LOCKED))
return new_handler;
return NULL;
}
Import changeset
2000-07-31 21:29:14 +02:00
Fix the tree: plugin-related changes, necessary for CSV storage engine to be statically compiled. Derived from antony's patch.
2006-01-19 22:40:56 +01:00
void handler::ha_statistic_increment(ulong SSV::*offset) const
{
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(table->in_use->status_var.*offset);
Fix the tree: plugin-related changes, necessary for CSV storage engine to be statically compiled. Derived from antony's patch.
2006-01-19 22:40:56 +01:00
}
more wl#2936 fixes: removed implicit ha_thd() calls (too error-prone), fixed an assert crash
2007-04-15 15:47:27 +02:00
void **handler::ha_data(THD *thd) const
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
{
Never access thd->ha_data directly, use getters/setters from the plugin API instead. This is a pre-requisite of the fix for Bug 12713, which changes the data type of thd->ha_data from void * to struct Ha_data.
2007-08-31 08:19:52 +02:00
return thd_ha_data(thd, ht);
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
}
THD *handler::ha_thd(void) const
{
wl#2936 - fixing problems
2007-04-13 19:23:02 +02:00
DBUG_ASSERT(!table || !table->in_use || table->in_use == current_thd);
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
return (table && table->in_use) ? table->in_use : current_thd;
}
Fix Bug #18559 "log tables cannot change engine, and gets deadlocked when dropping w/ log on" Log tables rely on concurrent insert machinery to add data. This means that log tables are always opened and locked by special (artificial) logger threads. Because of this, the thread which tries to drop a log table starts to wait for the table to be unlocked. Which will happen only if the log table is disabled. Alike situation happens if one tries to alter a log table. However in addition to the problem above, alter table calls check_if_locking_is_allowed() routine for the engine. The routine does not allow alter for the log tables. So, alter doesn't start waiting forever for logs to be disabled, but returns with an error. Another problem is that not all engines could be used for the log tables. That's because they need concurrent insert. In this patch we: (1) Explicitly disallow to drop/alter a log table if it is currently used by the logger. (2) Update MyISAM to support log tables (3) Allow to drop log tables/alter log tables if log is disabled At the same time we (4) Disallow to alter log tables to unsupported engine (after this patch CSV and MyISAM are alowed) Recommit with review fixes.
2006-08-03 19:28:15 +02:00
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
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 database-handler.
IMPLEMENTATION
Try O_RDONLY if cannot open as O_RDWR
Don't wait for locks if not HA_OPEN_WAIT_IF_LOCKED is set
*/
int handler::ha_open(TABLE *table_arg, const char *name, int mode,
int test_if_locked)
Import changeset
2000-07-31 21:29:14 +02:00
{
int error;
check for mysql_bin_log.is_open() before my_b_tell(&thd->transaction.trans_log in ha_commit_trans - why it didn't crash earlier ?
2004-08-22 14:23:52 +02:00
DBUG_ENTER("handler::ha_open");
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("enter",
("name: %s db_type: %d db_stat: %d mode: %d lock_test: %d",
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
name, ht->db_type, table_arg->db_stat, mode,
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
test_if_locked));
table= table_arg;
DBUG_ASSERT(table->s == table_share);
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(alloc_root_inited(&table->mem_root));
Import changeset
2000-07-31 21:29:14 +02:00
if ((error=open(name,mode,test_if_locked)))
{
if ((error == EACCES || error == EROFS) && mode == O_RDWR &&
(table->db_stat & HA_TRY_READ_ONLY))
{
table->db_stat|=HA_READ_ONLY;
error=open(name,O_RDONLY,test_if_locked);
}
}
if (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
my_errno= error; /* Safeguard */
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_PRINT("error",("error: %d errno: %d",error,errno));
}
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
if (table->s->db_options_in_use & HA_OPTION_READ_ONLY_DATA)
Import changeset
2000-07-31 21:29:14 +02:00
table->db_stat|=HA_READ_ONLY;
Move HA_EXTRA_NO_READCHECK to ha_open Fixed bug in multi-table-delete
2002-01-16 22:02:26 +01:00
(void) extra(HA_EXTRA_NO_READCHECK); // Not needed in SQL
BUG#27732 "Possible memory leak with index_merge" The bug was that handler::clone/handler::ha_open() call caused allocation of cloned_copy->ref on the handler->table->mem_root. The allocated memory could not be reclaimed until the table is flushed, so it was possible to exhaust memory by repeatedly running index_merge queries without doing table flushes. The fix: - make handler::clone() allocate new_handler->ref on the passed mem_root - make handler::ha_open() not allocate this->ref if it has already been allocated There is no testcase as it is not possible to check small leaks from testsuite.
2008-02-07 03:10:04 +01:00
/* ref is already allocated for us if we're called from handler::clone() */
Merge kaamos.(none):/data/src/opt/mysql-5.0-opt into kaamos.(none):/data/src/opt/mysql-5.1-opt
2008-02-13 10:12:00 +01:00
if (!ref && !(ref= (uchar*) alloc_root(&table->mem_root,
BUG#27732 "Possible memory leak with index_merge" The bug was that handler::clone/handler::ha_open() call caused allocation of cloned_copy->ref on the handler->table->mem_root. The allocated memory could not be reclaimed until the table is flushed, so it was possible to exhaust memory by repeatedly running index_merge queries without doing table flushes. The fix: - make handler::clone() allocate new_handler->ref on the passed mem_root - make handler::ha_open() not allocate this->ref if it has already been allocated There is no testcase as it is not possible to check small leaks from testsuite.
2008-02-07 03:10:04 +01:00
ALIGN_SIZE(ref_length)*2)))
Import changeset
2000-07-31 21:29:14 +02:00
{
close();
error=HA_ERR_OUT_OF_MEM;
}
else
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
dup_ref=ref+ALIGN_SIZE(ref_length);
cached_table_flags= table_flags();
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
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Read first row (only) from a table.
This is the second patch for bdb removeal. This takes care of all options and variables that rely on bdb.
2006-08-10 10:50:37 +02:00
This is never called for InnoDB tables, as these table types
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
has the HA_STATS_RECORDS_IS_EXACT set.
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +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
int handler::read_first_row(uchar * buf, uint primary_key)
Import changeset
2000-07-31 21:29:14 +02:00
{
register int error;
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
DBUG_ENTER("handler::read_first_row");
Import changeset
2000-07-31 21:29:14 +02:00
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
ha_statistic_increment(&SSV::ha_read_first_count);
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
/*
If there is very few deleted rows in the table, find the first row by
scanning the table.
Bug #6748 heap_rfirst() doesn't work (and never did!) range for BETWEEN typo fixed
2004-11-22 14:53:18 +01:00
TODO remove the test for HA_READ_ORDER
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
*/
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 (stats.deleted < 10 || primary_key >= MAX_KEY ||
Bug #6748 heap_rfirst() doesn't work (and never did!) range for BETWEEN typo fixed
2004-11-22 14:53:18 +01:00
!(index_flags(primary_key, 0, 0) & HA_READ_ORDER))
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
{
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
(void) ha_rnd_init(1);
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
while ((error= rnd_next(buf)) == HA_ERR_RECORD_DELETED) ;
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
(void) ha_rnd_end();
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
}
else
{
/* Find the first row through the primary key */
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
(void) ha_index_init(primary_key, 0);
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
error=index_first(buf);
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
(void) ha_index_end();
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(error);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Generate the next auto-increment number based on increment and offset.
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
computes the lowest number
- strictly greater than "nr"
- of the form: auto_increment_offset + N * auto_increment_increment
manually merged
2005-02-14 21:50:09 +01:00
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
In most cases increment= offset= 1, in which case we get:
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@verbatim 1,2,3,4,5,... @endverbatim
If increment=10 and offset=5 and previous number is 1, we get:
@verbatim 1,5,15,25,35,... @endverbatim
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
*/
inline ulonglong
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
compute_next_insert_id(ulonglong nr,struct system_variables *variables)
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
{
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
if (variables->auto_increment_increment == 1)
return (nr+1); // optimization of the formula below
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
nr= (((nr+ variables->auto_increment_increment -
variables->auto_increment_offset)) /
(ulonglong) variables->auto_increment_increment);
return (nr* (ulonglong) variables->auto_increment_increment +
variables->auto_increment_offset);
}
fixes after merge. Updates to test's results. We now reset the THD members related to auto_increment+binlog in MYSQL_LOG::write(). This is better than in THD::cleanup_after_query(), which was not able to distinguish between SELECT myfunc1(),myfunc2() and INSERT INTO t SELECT myfunc1(),myfunc2() from a binlogging point of view. Rows_log_event::exec_event() now calls lex_start() instead of mysql_init_query() because the latter now does too much (it resets the binlog format).
2006-07-10 18:41:03 +02:00
void handler::adjust_next_insert_id_after_explicit_value(ulonglong nr)
{
/*
If we have set THD::next_insert_id previously and plan to insert an
explicitely-specified value larger than this, we need to increase
THD::next_insert_id to be greater than the explicit value.
*/
if ((next_insert_id > 0) && (nr >= next_insert_id))
set_next_insert_id(compute_next_insert_id(nr, &table->in_use->variables));
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
fixes after merge. Updates to test's results. We now reset the THD members related to auto_increment+binlog in MYSQL_LOG::write(). This is better than in THD::cleanup_after_query(), which was not able to distinguish between SELECT myfunc1(),myfunc2() and INSERT INTO t SELECT myfunc1(),myfunc2() from a binlogging point of view. Rows_log_event::exec_event() now calls lex_start() instead of mysql_init_query() because the latter now does too much (it resets the binlog format).
2006-07-10 18:41:03 +02:00
Computes the largest number X:
- smaller than or equal to "nr"
- of the form: auto_increment_offset + N * auto_increment_increment
where N>=0.
SYNOPSIS
prev_insert_id
nr Number to "round down"
variables variables struct containing auto_increment_increment and
auto_increment_offset
RETURN
The number X if it exists, "nr" otherwise.
*/
inline ulonglong
prev_insert_id(ulonglong nr, struct system_variables *variables)
{
if (unlikely(nr < variables->auto_increment_offset))
{
/*
There's nothing good we can do here. That is a pathological case, where
the offset is larger than the column's max possible value, i.e. not even
the first sequence value may be inserted. User will receive warning.
*/
DBUG_PRINT("info",("auto_increment: nr: %lu cannot honour "
"auto_increment_offset: %lu",
Remove compiler warnings (Mostly in DBUG_PRINT() and unused arguments) Fixed bug in query cache when used with traceing (--with-debug) Fixed memory leak in mysqldump Removed warnings from mysqltest scripts (replaced -- with #)
2006-11-20 21:42:06 +01:00
(ulong) nr, variables->auto_increment_offset));
fixes after merge. Updates to test's results. We now reset the THD members related to auto_increment+binlog in MYSQL_LOG::write(). This is better than in THD::cleanup_after_query(), which was not able to distinguish between SELECT myfunc1(),myfunc2() and INSERT INTO t SELECT myfunc1(),myfunc2() from a binlogging point of view. Rows_log_event::exec_event() now calls lex_start() instead of mysql_init_query() because the latter now does too much (it resets the binlog format).
2006-07-10 18:41:03 +02:00
return nr;
}
if (variables->auto_increment_increment == 1)
return nr; // optimization of the formula below
nr= (((nr - variables->auto_increment_offset)) /
(ulonglong) variables->auto_increment_increment);
return (nr * (ulonglong) variables->auto_increment_increment +
variables->auto_increment_offset);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Update the auto_increment field if necessary.
Change update_auto_increment to return 1 if get_auto_increment() returned ~(ulonglong) This makes it easier to give an error in the handler if there was a problem generating an auto-increment value
2005-05-18 09:41:35 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
Updates columns with type NEXT_NUMBER if:
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
- If column value is set to NULL (in which case
auto_increment_field_not_null is 0)
- If column is set to 0 and (sql_mode & MODE_NO_AUTO_VALUE_ON_ZERO) is not
set. In the future we will only set NEXT_NUMBER fields if one sets them
to NULL (or they are not included in the insert list).
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
In those cases, we check if the currently reserved interval still has
values we have not used. If yes, we pick the smallest one and use it.
Otherwise:
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
- If a list of intervals has been provided to the statement via SET
INSERT_ID or via an Intvar_log_event (in a replication slave), we pick the
first unused interval from this list, consider it as reserved.
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
- Otherwise we set the column for the first row to the value
next_insert_id(get_auto_increment(column))) which is usually
max-used-column-value+1.
We call get_auto_increment() for the first row in a multi-row
statement. get_auto_increment() will tell us the interval of values it
reserved for us.
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
- In both cases, for the following rows we use those reserved values without
calling the handler again (we just progress in the interval, computing
each new value from the previous one). Until we have exhausted them, then
we either take the next provided interval or call get_auto_increment()
again to reserve a new interval.
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
- In both cases, the reserved intervals are remembered in
thd->auto_inc_intervals_in_cur_stmt_for_binlog if statement-based
binlogging; the last reserved interval is remembered in
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
auto_inc_interval_for_cur_row. The number of reserved intervals is
remembered in auto_inc_intervals_count. It differs from the number of
elements in thd->auto_inc_intervals_in_cur_stmt_for_binlog() because the
latter list is cumulative over all statements forming one binlog event
(when stored functions and triggers are used), and collapses two
contiguous intervals in one (see its append() method).
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
The idea is that generated auto_increment values are predictable and
independent of the column values in the table. This is needed to be
able to replicate into a table that already has rows with a higher
auto-increment value than the one that is inserted.
After we have already generated an auto-increment number and the user
inserts a column with a higher value than the last used one, we will
start counting from the inserted value.
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
This function's "outputs" are: the table's auto_increment field is filled
with a value, thd->next_insert_id is filled with the value to use for the
next row, if a value was autogenerated for the current row it is stored in
thd->insert_id_for_cur_row, if get_auto_increment() was called
thd->auto_inc_interval_for_cur_row is modified, if that interval is not
present in thd->auto_inc_intervals_in_cur_stmt_for_binlog it is added to
this list.
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@todo
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
Replace all references to "next number" or NEXT_NUMBER to
"auto_increment", everywhere (see below: there is
table->auto_increment_field_not_null, and there also exists
table->next_number_field, it's not consistent).
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
0 ok
@retval
HA_ERR_AUTOINC_READ_FAILED get_auto_increment() was called and
returned ~(ulonglong) 0
@retval
HA_ERR_AUTOINC_ERANGE storing value in field caused strict mode
failure.
Increase max package length to 512M for mysql and mysqldump. Faster 'read_first_row' (Fixes slow 'preparing' state) Read constant tables earlier, which provides better optimzations when using tables with <=1 row. This also fixes a complicated bug involving const tables.
2002-01-23 01:52:26 +01:00
*/
Import changeset
2000-07-31 21:29:14 +02:00
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
#define AUTO_INC_DEFAULT_NB_ROWS 1 // Some prefer 1024 here
#define AUTO_INC_DEFAULT_NB_MAX_BITS 16
#define AUTO_INC_DEFAULT_NB_MAX ((1 << AUTO_INC_DEFAULT_NB_MAX_BITS) - 1)
Bug#20573 "strict mode: inserts autogenerated auto_increment value bigger than max" Strict mode should fail if autoincrement value is out of range
2006-08-30 22:20:39 +02:00
int handler::update_auto_increment()
Import changeset
2000-07-31 21:29:14 +02:00
{
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
ulonglong nr, nb_reserved_values;
bool append= FALSE;
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
THD *thd= table->in_use;
struct system_variables *variables= &thd->variables;
check for mysql_bin_log.is_open() before my_b_tell(&thd->transaction.trans_log in ha_commit_trans - why it didn't crash earlier ?
2004-08-22 14:23:52 +02:00
DBUG_ENTER("handler::update_auto_increment");
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
/*
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
next_insert_id is a "cursor" into the reserved interval, it may go greater
than the interval, but not smaller.
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
*/
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
DBUG_ASSERT(next_insert_id >= auto_inc_interval_for_cur_row.minimum());
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
if ((nr= table->next_number_field->val_int()) != 0 ||
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 first patch, fixing a number of the warnings, predominantly "suggest using parentheses around && in ||", and empty for and while bodies.
2009-06-09 14:55:30 +02:00
(table->auto_increment_field_not_null &&
thd->variables.sql_mode & MODE_NO_AUTO_VALUE_ON_ZERO))
Merged some functions and removed some unused client functions. Remember UNION for ALTER TABLE Added test for if we are supporting transactions. Don't allow REPLACE to replace a row when we have generated an auto_increment key Fixed bug when using BLOB keys Fixed bug in SET @variable=user.
2001-03-06 14:24:08 +01:00
{
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
/*
Update next_insert_id if we had already generated a value in this
statement (case of INSERT VALUES(null),(3763),(null):
the last NULL needs to insert 3764, not the value of the first NULL plus
1).
*/
fixes after merge. Updates to test's results. We now reset the THD members related to auto_increment+binlog in MYSQL_LOG::write(). This is better than in THD::cleanup_after_query(), which was not able to distinguish between SELECT myfunc1(),myfunc2() and INSERT INTO t SELECT myfunc1(),myfunc2() from a binlogging point of view. Rows_log_event::exec_event() now calls lex_start() instead of mysql_init_query() because the latter now does too much (it resets the binlog format).
2006-07-10 18:41:03 +02:00
adjust_next_insert_id_after_explicit_value(nr);
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
insert_id_for_cur_row= 0; // didn't generate anything
Change update_auto_increment to return 1 if get_auto_increment() returned ~(ulonglong) This makes it easier to give an error in the handler if there was a problem generating an auto-increment value
2005-05-18 09:41:35 +02:00
DBUG_RETURN(0);
Merged some functions and removed some unused client functions. Remember UNION for ALTER TABLE Added test for if we are supporting transactions. Don't allow REPLACE to replace a row when we have generated an auto_increment key Fixed bug when using BLOB keys Fixed bug in SET @variable=user.
2001-03-06 14:24:08 +01:00
}
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
if ((nr= next_insert_id) >= auto_inc_interval_for_cur_row.maximum())
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
{
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/* next_insert_id is beyond what is reserved, so we reserve more. */
const Discrete_interval *forced=
thd->auto_inc_intervals_forced.get_next();
if (forced != NULL)
{
nr= forced->minimum();
nb_reserved_values= forced->values();
}
else
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
{
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/*
handler::estimation_rows_to_insert was set by
handler::ha_start_bulk_insert(); if 0 it means "unknown".
*/
ulonglong nb_desired_values;
/*
If an estimation was given to the engine:
- use it.
- if we already reserved numbers, it means the estimation was
not accurate, then we'll reserve 2*AUTO_INC_DEFAULT_NB_ROWS the 2nd
time, twice that the 3rd time etc.
If no estimation was given, use those increasing defaults from the
start, starting from AUTO_INC_DEFAULT_NB_ROWS.
Don't go beyond a max to not reserve "way too much" (because
reservation means potentially losing unused values).
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
Note that in prelocked mode no estimation is given.
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
*/
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
if ((auto_inc_intervals_count == 0) && (estimation_rows_to_insert > 0))
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
nb_desired_values= estimation_rows_to_insert;
else /* go with the increasing defaults */
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
{
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/* avoid overflow in formula, with this if() */
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
if (auto_inc_intervals_count <= AUTO_INC_DEFAULT_NB_MAX_BITS)
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
{
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
nb_desired_values= AUTO_INC_DEFAULT_NB_ROWS *
(1 << auto_inc_intervals_count);
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
set_if_smaller(nb_desired_values, AUTO_INC_DEFAULT_NB_MAX);
}
else
nb_desired_values= AUTO_INC_DEFAULT_NB_MAX;
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
}
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/* This call ignores all its parameters but nr, currently */
get_auto_increment(variables->auto_increment_offset,
variables->auto_increment_increment,
nb_desired_values, &nr,
&nb_reserved_values);
if (nr == ~(ulonglong) 0)
Merge acurtis@bk-internal.mysql.com:/home/bk/mysql-5.0-engines into xiphis.org:/home/antony/work2/engines-merge
2006-09-18 21:16:03 +02:00
DBUG_RETURN(HA_ERR_AUTOINC_READ_FAILED); // Mark failure
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/*
That rounding below should not be needed when all engines actually
respect offset and increment in get_auto_increment(). But they don't
so we still do it. Wonder if for the not-first-in-index we should do
it. Hope that this rounding didn't push us out of the interval; even
if it did we cannot do anything about it (calling the engine again
will not help as we inserted no row).
*/
nr= compute_next_insert_id(nr-1, variables);
}
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
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 (table->s->next_number_keypart == 0)
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
{
/* We must defer the appending until "nr" has been possibly truncated */
append= TRUE;
}
else
{
/*
For such auto_increment there is no notion of interval, just a
singleton. The interval is not even stored in
thd->auto_inc_interval_for_cur_row, so we are sure to call the engine
for next row.
*/
DBUG_PRINT("info",("auto_increment: special not-first-in-index"));
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
}
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
}
DBUG_PRINT("info",("auto_increment: %lu", (ulong) nr));
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
if (unlikely(table->next_number_field->store((longlong) nr, TRUE)))
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
{
Fix for BUG#20524 "auto_increment_* not observed when inserting a too large value": the bug was that if MySQL generated a value for an auto_increment column, based on auto_increment_* variables, and this value was bigger than the column's max possible value, then that max possible value was inserted (after issuing a warning). But this didn't honour auto_increment_* variables (and so could cause conflicts in a master-master replication where one master is supposed to generated only even numbers, and the other only odd numbers), so now we "round down" this max possible value to honour auto_increment_* variables, before inserting it.
2006-07-06 14:37:09 +02:00
/*
Merge acurtis@bk-internal.mysql.com:/home/bk/mysql-5.0-engines into xiphis.org:/home/antony/work2/engines-merge
2006-09-18 21:16:03 +02:00
first test if the query was aborted due to strict mode constraints
*/
if (thd->killed == THD::KILL_BAD_DATA)
DBUG_RETURN(HA_ERR_AUTOINC_ERANGE);
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
/*
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
field refused this value (overflow) and truncated it, use the result of
fixes after merge. Updates to test's results. We now reset the THD members related to auto_increment+binlog in MYSQL_LOG::write(). This is better than in THD::cleanup_after_query(), which was not able to distinguish between SELECT myfunc1(),myfunc2() and INSERT INTO t SELECT myfunc1(),myfunc2() from a binlogging point of view. Rows_log_event::exec_event() now calls lex_start() instead of mysql_init_query() because the latter now does too much (it resets the binlog format).
2006-07-10 18:41:03 +02:00
the truncation (which is going to be inserted); however we try to
decrease it to honour auto_increment_* variables.
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
That will shift the left bound of the reserved interval, we don't
bother shifting the right bound (anyway any other value from this
interval will cause a duplicate key).
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
*/
fixes after merge. Updates to test's results. We now reset the THD members related to auto_increment+binlog in MYSQL_LOG::write(). This is better than in THD::cleanup_after_query(), which was not able to distinguish between SELECT myfunc1(),myfunc2() and INSERT INTO t SELECT myfunc1(),myfunc2() from a binlogging point of view. Rows_log_event::exec_event() now calls lex_start() instead of mysql_init_query() because the latter now does too much (it resets the binlog format).
2006-07-10 18:41:03 +02:00
nr= prev_insert_id(table->next_number_field->val_int(), variables);
if (unlikely(table->next_number_field->store((longlong) nr, TRUE)))
nr= table->next_number_field->val_int();
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
}
if (append)
{
auto_inc_interval_for_cur_row.replace(nr, nb_reserved_values,
variables->auto_increment_increment);
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
auto_inc_intervals_count++;
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/* Row-based replication does not need to store intervals in binlog */
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
if (mysql_bin_log.is_open() && !thd->current_stmt_binlog_row_based)
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
thd->auto_inc_intervals_in_cur_stmt_for_binlog.append(auto_inc_interval_for_cur_row.minimum(),
auto_inc_interval_for_cur_row.values(),
variables->auto_increment_increment);
post merge fixes
2006-09-19 00:52:16 +02:00
}
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/*
Record this autogenerated value. If the caller then
succeeds to insert this value, it will call
record_first_successful_insert_id_in_cur_stmt()
which will set first_successful_insert_id_in_cur_stmt if it's not
already set.
*/
insert_id_for_cur_row= nr;
/*
Set next insert id to point to next auto-increment value to be able to
handle multi-row statements.
*/
set_next_insert_id(compute_next_insert_id(nr, variables));
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
post merge fixes
2006-09-19 00:52:16 +02:00
DBUG_RETURN(0);
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
}
Import changeset
2000-07-31 21:29:14 +02:00
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
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
MySQL signal that it changed the column bitmap
USAGE
This is for handlers that needs to setup their own column bitmaps.
Normally the handler should set up their own column bitmaps in
index_init() or rnd_init() and in any column_bitmaps_signal() call after
this.
The handler is allowd to do changes to the bitmap after a index_init or
rnd_init() call is made as after this, MySQL will not use the bitmap
for any program logic checking.
*/
void handler::column_bitmaps_signal()
{
DBUG_ENTER("column_bitmaps_signal");
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
DBUG_PRINT("info", ("read_set: 0x%lx write_set: 0x%lx", (long) table->read_set,
(long) table->write_set));
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_VOID_RETURN;
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
Reserves an interval of auto_increment values from the handler.
SYNOPSIS
get_auto_increment()
offset
increment
nb_desired_values how many values we want
first_value (OUT) the first value reserved by the handler
nb_reserved_values (OUT) how many values the handler reserved
offset and increment means that we want values to be of the form
offset + N * increment, where N>=0 is integer.
If the function sets *first_value to ~(ulonglong)0 it means an error.
If the function sets *nb_reserved_values to ULONGLONG_MAX it means it has
reserved to "positive infinite".
*/
void handler::get_auto_increment(ulonglong offset, ulonglong increment,
ulonglong nb_desired_values,
ulonglong *first_value,
ulonglong *nb_reserved_values)
Import changeset
2000-07-31 21:29:14 +02:00
{
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
ulonglong nr;
Import changeset
2000-07-31 21:29:14 +02:00
int error;
AUTO_INCREMENT support for MERGE HA_AUTO_PART_KEY support for handler parent class
2002-07-21 23:55:32 +02:00
Import changeset
2000-07-31 21:29:14 +02:00
(void) extra(HA_EXTRA_KEYREAD);
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
table->mark_columns_used_by_index_no_reset(table->s->next_number_index,
table->read_set);
column_bitmaps_signal();
Patch for push of wl1354 Partitioning
2005-07-18 13:31:02 +02:00
index_init(table->s->next_number_index, 1);
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 (table->s->next_number_keypart == 0)
AUTO_INCREMENT support for MERGE HA_AUTO_PART_KEY support for handler parent class
2002-07-21 23:55:32 +02:00
{ // Autoincrement at key-start
error=index_last(table->record[1]);
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
/*
MySQL implicitely assumes such method does locking (as MySQL decides to
use nr+increment without checking again with the handler, in
handler::update_auto_increment()), so reserves to infinite.
*/
*nb_reserved_values= ULONGLONG_MAX;
AUTO_INCREMENT support for MERGE HA_AUTO_PART_KEY support for handler parent class
2002-07-21 23:55:32 +02:00
}
else
{
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 key[MAX_KEY_LENGTH];
Mainly resolving Guilhem's 4.1 patch to 5.0
2004-11-12 04:01:46 +01:00
key_copy(key, table->record[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
table->key_info + table->s->next_number_index,
table->s->next_number_key_offset);
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
error= index_read_map(table->record[1], key,
make_prev_keypart_map(table->s->next_number_keypart),
HA_READ_PREFIX_LAST);
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
/*
MySQL needs to call us for next row: assume we are inserting ("a",null)
here, we return 3, and next this statement will want to insert
("b",null): there is no reason why ("b",3+1) would be the good row to
insert: maybe it already exists, maybe 3+1 is too large...
*/
*nb_reserved_values= 1;
AUTO_INCREMENT support for MERGE HA_AUTO_PART_KEY support for handler parent class
2002-07-21 23:55:32 +02:00
}
Import changeset
2000-07-31 21:29:14 +02:00
if (error)
nr=1;
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
nr= ((ulonglong) table->next_number_field->
val_int_offset(table->s->rec_buff_length)+1);
Import changeset
2000-07-31 21:29:14 +02:00
index_end();
AUTO_INCREMENT support for MERGE HA_AUTO_PART_KEY support for handler parent class
2002-07-21 23:55:32 +02:00
(void) extra(HA_EXTRA_NO_KEYREAD);
First push for WL#3146 "less locking in auto_increment". It is a 0-real-change patch. New prototype for get_auto_increment() (but new arguments not yet used), to be able to reserve a finite interval of auto_increment values from cooperating engines. A hint on how many values to reserve is found in handler::estimation_rows_to_insert, filled by ha_start_bulk_insert(), new wrapper around start_bulk_insert(). NOTE: this patch changes nothing, for all engines. But it makes the API ready for those engines which will want to do reservation. More csets will come to complete WL#3146.
2006-06-02 22:21:32 +02:00
*first_value= nr;
Import changeset
2000-07-31 21:29:14 +02:00
}
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
void handler::ha_release_auto_increment()
{
release_auto_increment();
insert_id_for_cur_row= 0;
auto_inc_interval_for_cur_row.replace(0, 0, 0);
Fix for BUG#31612 "Trigger fired multiple times leads to gaps in auto_increment sequence". The bug was that if a trigger fired multiple times inside a top statement (for example top-statement is a multi-row INSERT, and trigger is ON INSERT), and that trigger inserted into an auto_increment column, then gaps could be observed in the auto_increment sequence, even if there were no other users of the database (no concurrency). It was wrong usage of THD::auto_inc_intervals_in_cur_stmt_for_binlog. Note that the fix changes "class handler", I'll tell the Storage Engine API team.
2008-10-06 16:06:59 +02:00
auto_inc_intervals_count= 0;
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
if (next_insert_id > 0)
{
next_insert_id= 0;
/*
this statement used forced auto_increment values if there were some,
wipe them away for other statements.
*/
table->in_use->auto_inc_intervals_forced.empty();
}
}
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
void handler::print_keydup_error(uint key_nr, const char *msg)
bug #14573 (Error on adding auto_increment on to a column with '0' values)
2006-05-12 15:02:42 +02:00
{
/* Write the duplicated key in the error message */
char key[MAX_KEY_LENGTH];
String str(key,sizeof(key),system_charset_info);
Bug #21072 Duplicate key error in NDB references wrong key: Added new printout for duplicate key error when key is unknown
2006-10-16 12:14:38 +02:00
if (key_nr == MAX_KEY)
{
/* Key is unknown */
Bug #21072 Duplicate key error in NDB references wrong key: Re-wrote string usage to avoid valgrind warnings
2006-10-25 15:34:47 +02:00
str.copy("", 0, system_charset_info);
Bug #28842 Different 'duplicate key' error code between 5.0 and 5.1 The patch for WL 1563 added a new duplicate key error message so that the key name could be provided instead of the key number. But the error code for the new message was used even though that did not need to change. This could cause unnecessary problems for applications that used the old ER_DUP_ENTRY error code to detect duplicate key errors.
2007-06-06 19:57:07 +02:00
my_printf_error(ER_DUP_ENTRY, msg, MYF(0), str.c_ptr(), "*UNKNOWN*");
Bug #21072 Duplicate key error in NDB references wrong key: Added new printout for duplicate key error when key is unknown
2006-10-16 12:14:38 +02:00
}
else
bug #14573 (Error on adding auto_increment on to a column with '0' values)
2006-05-12 15:02:42 +02:00
{
Bug #21072 Duplicate key error in NDB references wrong key: Added new printout for duplicate key error when key is unknown
2006-10-16 12:14:38 +02:00
/* Table is opened and defined at this point */
key_unpack(&str,table,(uint) key_nr);
uint max_length=MYSQL_ERRMSG_SIZE-(uint) strlen(msg);
if (str.length() >= max_length)
{
str.length(max_length-4);
str.append(STRING_WITH_LEN("..."));
}
Bug #28842 Different 'duplicate key' error code between 5.0 and 5.1 The patch for WL 1563 added a new duplicate key error message so that the key name could be provided instead of the key number. But the error code for the new message was used even though that did not need to change. This could cause unnecessary problems for applications that used the old ER_DUP_ENTRY error code to detect duplicate key errors.
2007-06-06 19:57:07 +02:00
my_printf_error(ER_DUP_ENTRY, msg,
Bug #38701: Crash in String::append when inserting duplicate empty strings an uft8 SET col When reporting a duplicate key error the server was making incorrect assumptions on what the state of the value string to include in the error is. Fixed by accessing the data in this string in a "safe" way (without relying on it having a terminating 0). Detected by code analysis and fixed a similar problem in reporting the foreign key duplicate errors.
2008-09-05 17:21:59 +02:00
MYF(0), str.c_ptr_safe(), table->key_info[key_nr].name);
bug #14573 (Error on adding auto_increment on to a column with '0' values)
2006-05-12 15:02:42 +02:00
}
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Print error that we got from handler function.
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
In case of delete table it's only safe to use the following parts of
the 'table' structure:
- table->s->path
- table->alias
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
*/
Import changeset
2000-07-31 21:29:14 +02:00
void handler::print_error(int error, myf errflag)
{
check for mysql_bin_log.is_open() before my_b_tell(&thd->transaction.trans_log in ha_commit_trans - why it didn't crash earlier ?
2004-08-22 14:23:52 +02:00
DBUG_ENTER("handler::print_error");
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_PRINT("enter",("error: %d",error));
int textno=ER_GET_ERRNO;
switch (error) {
bug #702 - ALTER TABLE tm UNION=(t1,t2) doesn't work if t1.MYD is compressed
2003-07-15 20:43:57 +02:00
case EACCES:
textno=ER_OPEN_AS_READONLY;
break;
Import changeset
2000-07-31 21:29:14 +02:00
case EAGAIN:
textno=ER_FILE_USED;
break;
case ENOENT:
textno=ER_FILE_NOT_FOUND;
break;
case HA_ERR_KEY_NOT_FOUND:
case HA_ERR_NO_ACTIVE_RECORD:
case HA_ERR_END_OF_FILE:
textno=ER_KEY_NOT_FOUND;
break;
low-level error messages cleanup
2003-01-28 17:42:08 +01:00
case HA_ERR_WRONG_MRG_TABLE_DEF:
Import changeset
2000-07-31 21:29:14 +02:00
textno=ER_WRONG_MRG_TABLE;
break;
case HA_ERR_FOUND_DUPP_KEY:
{
uint key_nr=get_dup_key(error);
if ((int) key_nr >= 0)
{
Give warnings for unused objects Changed error message to be compatible with old error file Added new error message for new DUP_ENTRY syntax
2007-01-22 17:42:52 +01:00
print_keydup_error(key_nr, ER(ER_DUP_ENTRY_WITH_KEY_NAME));
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_VOID_RETURN;
}
textno=ER_DUP_KEY;
break;
}
Fixed BUG#9680: Wrong error from cascading update Applied 9680.patch (by Osku Salerma)
2006-02-09 18:43:10 +01:00
case HA_ERR_FOREIGN_DUPLICATE_KEY:
{
uint key_nr= get_dup_key(error);
if ((int) key_nr >= 0)
{
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 max_length;
Fixed BUG#9680: Wrong error from cascading update Applied 9680.patch (by Osku Salerma)
2006-02-09 18:43:10 +01:00
/* Write the key in the error message */
char key[MAX_KEY_LENGTH];
String str(key,sizeof(key),system_charset_info);
/* Table is opened and defined at this point */
key_unpack(&str,table,(uint) key_nr);
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
max_length= (MYSQL_ERRMSG_SIZE-
(uint) strlen(ER(ER_FOREIGN_DUPLICATE_KEY)));
Fixed BUG#9680: Wrong error from cascading update Applied 9680.patch (by Osku Salerma)
2006-02-09 18:43:10 +01:00
if (str.length() >= max_length)
{
str.length(max_length-4);
str.append(STRING_WITH_LEN("..."));
}
my_error(ER_FOREIGN_DUPLICATE_KEY, MYF(0), table_share->table_name.str,
Bug #38701: Crash in String::append when inserting duplicate empty strings an uft8 SET col When reporting a duplicate key error the server was making incorrect assumptions on what the state of the value string to include in the error is. Fixed by accessing the data in this string in a "safe" way (without relying on it having a terminating 0). Detected by code analysis and fixed a similar problem in reporting the foreign key duplicate errors.
2008-09-05 17:21:59 +02:00
str.c_ptr_safe(), key_nr+1);
Fixed BUG#9680: Wrong error from cascading update Applied 9680.patch (by Osku Salerma)
2006-02-09 18:43:10 +01:00
DBUG_VOID_RETURN;
}
textno= ER_DUP_KEY;
break;
}
Fix for bug #6516 (Server crash loading spatial data) (after discussion with SerG)
2004-12-10 13:06:49 +01:00
case HA_ERR_NULL_IN_SPATIAL:
bug #21790 (UNKNOWN ERROR message in geometry) We issued UNKNOWN ERROR initially in this place and forgot to fix it when we implemented informative error message for this
2006-10-01 13:36:26 +02:00
my_error(ER_CANT_CREATE_GEOMETRY_OBJECT, MYF(0));
DBUG_VOID_RETURN;
Import changeset
2000-07-31 21:29:14 +02:00
case HA_ERR_FOUND_DUPP_UNIQUE:
textno=ER_DUP_UNIQUE;
break;
case HA_ERR_RECORD_CHANGED:
textno=ER_CHECKREAD;
break;
case HA_ERR_CRASHED:
textno=ER_NOT_KEYFILE;
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
case HA_ERR_WRONG_IN_RECORD:
textno= ER_CRASHED_ON_USAGE;
break;
Split error message for crashed table to 3 different error messages. Fixes for Windows
2000-11-22 02:45:02 +01:00
case HA_ERR_CRASHED_ON_USAGE:
textno=ER_CRASHED_ON_USAGE;
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
case HA_ERR_NOT_A_TABLE:
textno= error;
break;
Split error message for crashed table to 3 different error messages. Fixes for Windows
2000-11-22 02:45:02 +01:00
case HA_ERR_CRASHED_ON_REPAIR:
textno=ER_CRASHED_ON_REPAIR;
break;
Import changeset
2000-07-31 21:29:14 +02:00
case HA_ERR_OUT_OF_MEM:
void handler::print_error(): cleanup
2005-08-08 20:35:51 +02:00
textno=ER_OUT_OF_RESOURCES;
break;
Import changeset
2000-07-31 21:29:14 +02:00
case HA_ERR_WRONG_COMMAND:
textno=ER_ILLEGAL_HA;
break;
case HA_ERR_OLD_FILE:
textno=ER_OLD_KEYFILE;
break;
case HA_ERR_UNSUPPORTED:
textno=ER_UNSUPPORTED_EXTENSION;
break;
case HA_ERR_RECORD_FILE_FULL:
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
case HA_ERR_INDEX_FILE_FULL:
Fix for BUG#39886 - Table full for MEMORY table is not written into error log When a MEMORY table is full the error is returned to client but not written to error log. Fixed the handler api to write the error mesage to error log when the table is full. Note: No TestCase included as testing the error log is non-trivial.
2009-01-28 12:52:53 +01:00
{
void handler::print_error(): cleanup
2005-08-08 20:35:51 +02:00
textno=ER_RECORD_FILE_FULL;
Fix for BUG#39886 - Table full for MEMORY table is not written into error log When a MEMORY table is full the error is returned to client but not written to error log. Fixed the handler api to write the error mesage to error log when the table is full. Note: No TestCase included as testing the error log is non-trivial.
2009-01-28 12:52:53 +01:00
/* Write the error message to error log */
errflag|= ME_NOREFRESH;
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
break;
Fix for BUG#39886 - Table full for MEMORY table is not written into error log When a MEMORY table is full the error is returned to client but not written to error log. Fixed the handler api to write the error mesage to error log when the table is full. Note: No TestCase included as testing the error log is non-trivial.
2009-01-28 12:52:53 +01:00
}
Pushing all the Gemini changes above the table handler.
2001-05-29 15:29:08 +02:00
case HA_ERR_LOCK_WAIT_TIMEOUT:
textno=ER_LOCK_WAIT_TIMEOUT;
break;
case HA_ERR_LOCK_TABLE_FULL:
textno=ER_LOCK_TABLE_FULL;
break;
Fix for LOAD DATA FROM MASTER when using tables with checksum Fixed some wrong test cases
2001-09-18 03:58:15 +02:00
case HA_ERR_LOCK_DEADLOCK:
textno=ER_LOCK_DEADLOCK;
break;
Pushing all the Gemini changes above the table handler.
2001-05-29 15:29:08 +02:00
case HA_ERR_READ_ONLY_TRANSACTION:
textno=ER_READ_ONLY_TRANSACTION;
break;
Merge
2001-10-30 16:38:44 +01:00
case HA_ERR_CANNOT_ADD_FOREIGN:
textno=ER_CANNOT_ADD_FOREIGN;
break;
case HA_ERR_ROW_IS_REFERENCED:
Fix bug #3443, better foreign key error messsages.
2005-09-23 15:22:27 +02:00
{
String str;
get_error_message(error, &str);
my_error(ER_ROW_IS_REFERENCED_2, MYF(0), str.c_ptr_safe());
DBUG_VOID_RETURN;
}
Merge
2001-10-30 16:38:44 +01:00
case HA_ERR_NO_REFERENCED_ROW:
Fix bug #3443, better foreign key error messsages.
2005-09-23 15:22:27 +02:00
{
String str;
get_error_message(error, &str);
my_error(ER_NO_REFERENCED_ROW_2, MYF(0), str.c_ptr_safe());
DBUG_VOID_RETURN;
}
BUG#6762 ALTER TABLE gives weird results and error message - Removed hardcoded error message from 4.1
2005-04-07 20:17:37 +02:00
case HA_ERR_TABLE_DEF_CHANGED:
textno=ER_TABLE_DEF_CHANGED;
break;
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
case HA_ERR_NO_SUCH_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
my_error(ER_NO_SUCH_TABLE, MYF(0), table_share->db.str,
table_share->table_name.str);
In ha_delete_table, use a standard mechanism to intercept the error message and convert it to a warning instead of direct manipulation with the thread error stack. Fix a bug in handler::print_erorr when a garbled message was printed for HA_ERR_NO_SUCH_TABLE. This is a pre-requisite patch for the fix for Bug#12713 Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem
2007-10-30 20:35:14 +01:00
DBUG_VOID_RETURN;
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
case HA_ERR_RBR_LOGGING_FAILED:
textno= ER_BINLOG_ROW_LOGGING_FAILED;
break;
WL#1563 - Modify MySQL to support fast CREATE/DROP INDEX NDB cluster is not fully supported. This will be added with WL 1892 (NDB Handler: Add support for CREATE/DROP INDEX). Some preparatory code for this is already present though. A change for the "duplicate key" error message is planned for another changeset.
2006-01-12 10:05:07 +01:00
case HA_ERR_DROP_INDEX_FK:
{
const char *ptr= "???";
uint key_nr= get_dup_key(error);
if ((int) key_nr >= 0)
ptr= table->key_info[key_nr].name;
my_error(ER_DROP_INDEX_FK, MYF(0), ptr);
DBUG_VOID_RETURN;
}
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
case HA_ERR_TABLE_NEEDS_UPGRADE:
textno=ER_TABLE_NEEDS_UPGRADE;
break;
Bug #17766: The server accepts to create MERGE tables which cannot work Changed the error reporting (and a crash) when inserting data into a MERGE table that has no underlying tables or no INSERT_METHOD specified by reporting that it is read-only.
2006-07-12 02:28:50 +02:00
case HA_ERR_TABLE_READONLY:
textno= ER_OPEN_AS_READONLY;
break;
Bug#20573 "strict mode: inserts autogenerated auto_increment value bigger than max" Strict mode should fail if autoincrement value is out of range
2006-08-30 22:20:39 +02:00
case HA_ERR_AUTOINC_READ_FAILED:
textno= ER_AUTOINC_READ_FAILED;
break;
case HA_ERR_AUTOINC_ERANGE:
textno= ER_WARN_DATA_OUT_OF_RANGE;
break;
Fix for BUG#18828 - If InnoDB runs out of undo slots, it returns misleading 'table is full' Innodb returns a misleading error message "table is full" when the number of active concurrent transactions is greater than 1024. Fixed by adding errorcode "ER_TOO_MANY_CONCURRENT_TRXS" to the error codes. Innodb should return HA_TOO_MANY_CONCURRENT_TRXS to mysql which is then mapped to ER_TOO_MANY_CONCURRENT_TRXS Note: testcase is not written as this was reproducible only by changing innodb code.
2009-07-17 10:43:53 +02:00
case HA_ERR_TOO_MANY_CONCURRENT_TRXS:
textno= ER_TOO_MANY_CONCURRENT_TRXS;
break;
Import changeset
2000-07-31 21:29:14 +02:00
default:
{
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
/* The error was "unknown" to this function.
Ask handler if it has got a message for this error */
bool temporary= FALSE;
Fixed prototype of get_error_message to use String to return error message WL#1747 and #1746 allow user to decide if ordered index should be created or not
2004-05-24 12:35:39 +02:00
String str;
temporary= get_error_message(error, &str);
if (!str.is_empty())
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
{
Uses table_type() function to get name of storage engine before printing engine specific error message.
2004-05-25 12:10:54 +02:00
const char* engine= table_type();
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
if (temporary)
now my_printf_error is not better then my_error, but my_error call is shorter used only one implementation of format parser of (printf) fixed multistatement
2004-11-13 18:35:51 +01:00
my_error(ER_GET_TEMPORARY_ERRMSG, MYF(0), error, str.ptr(), engine);
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
else
now my_printf_error is not better then my_error, but my_error call is shorter used only one implementation of format parser of (printf) fixed multistatement
2004-11-13 18:35:51 +01:00
my_error(ER_GET_ERRMSG, MYF(0), error, str.ptr(), engine);
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
}
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
else
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
my_error(ER_GET_ERRNO,errflag,error);
Import changeset
2000-07-31 21:29:14 +02:00
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
my_error(textno, errflag, table_share->table_name.str, error);
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_VOID_RETURN;
}
Fix for MacOSX and symlinks Fix for USE_FRM and crashed index file
2003-05-14 00:27:26 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Return an error message specific to this handler.
XA (not completely polished out yet)
2005-01-16 13:16:23 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param error error code previously returned by handler
@param buf pointer to String where to add error message
print xa recovery progress add names to handlertons trans_need_2pc() macro
2005-03-13 21:58:09 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@return
Returns true if this is a temporary error
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Fixed prototype of get_error_message to use String to return error message WL#1747 and #1746 allow user to decide if ordered index should be created or not
2004-05-24 12:35:39 +02:00
bool handler::get_error_message(int error, String* buf)
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
{
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
return FALSE;
WL #1729 Handler: error text for NDB errors - New solution after discussions with Sergei, no handler specific code or error messages should be in sql layer. next_result, only check for error if check is -1 Improved index_read
2004-05-13 11:56:45 +02:00
}
Fix for bug#39585: innodb and myisam corruption after binary upgrade from <=5.0.46 to >=5.0.48 Problem: 'check table .. for upgrade' doesn't detect incompatible collation changes made in 5.0.48. Fix: check for incompatible collation changes.
2008-10-08 15:52:57 +02:00
/**
Check for incompatible collation changes.
@retval
HA_ADMIN_NEEDS_UPGRADE Table may have data requiring upgrade.
@retval
0 No upgrade required.
*/
int handler::check_collation_compatibility()
{
ulong mysql_version= table->s->mysql_version;
Fix for bug #40053: 'check table .. for upgrade' doesn't detect collation change made in 5.1.24-rc Problem: 'CHECK TABLE ... FOR UPGRADE' did not check for incompatible collation changes made in MySQL 5.1.24-rc. Fix: add the check.
2008-10-17 14:45:17 +02:00
if (mysql_version < 50124)
Fix for bug#39585: innodb and myisam corruption after binary upgrade from <=5.0.46 to >=5.0.48 Problem: 'check table .. for upgrade' doesn't detect incompatible collation changes made in 5.0.48. Fix: check for incompatible collation changes.
2008-10-08 15:52:57 +02:00
{
KEY *key= table->key_info;
KEY *key_end= key + table->s->keys;
for (; key < key_end; key++)
{
KEY_PART_INFO *key_part= key->key_part;
KEY_PART_INFO *key_part_end= key_part + key->key_parts;
for (; key_part < key_part_end; key_part++)
{
if (!key_part->fieldnr)
continue;
Field *field= table->field[key_part->fieldnr - 1];
uint cs_number= field->charset()->number;
Fix for bug #40053: 'check table .. for upgrade' doesn't detect collation change made in 5.1.24-rc Problem: 'CHECK TABLE ... FOR UPGRADE' did not check for incompatible collation changes made in MySQL 5.1.24-rc. Fix: add the check.
2008-10-17 14:45:17 +02:00
if ((mysql_version < 50048 &&
(cs_number == 11 || /* ascii_general_ci - bug #29499, bug #27562 */
cs_number == 41 || /* latin7_general_ci - bug #29461 */
cs_number == 42 || /* latin7_general_cs - bug #29461 */
cs_number == 20 || /* latin7_estonian_cs - bug #29461 */
cs_number == 21 || /* latin2_hungarian_ci - bug #29461 */
cs_number == 22 || /* koi8u_general_ci - bug #29461 */
cs_number == 23 || /* cp1251_ukrainian_ci - bug #29461 */
cs_number == 26)) || /* cp1250_general_ci - bug #29461 */
(mysql_version < 50124 &&
(cs_number == 33 || /* utf8_general_ci - bug #27877 */
cs_number == 35))) /* ucs2_general_ci - bug #27877 */
Fix for bug#39585: innodb and myisam corruption after binary upgrade from <=5.0.46 to >=5.0.48 Problem: 'check table .. for upgrade' doesn't detect incompatible collation changes made in 5.0.48. Fix: check for incompatible collation changes.
2008-10-08 15:52:57 +02:00
return HA_ADMIN_NEEDS_UPGRADE;
}
}
}
return 0;
}
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
int handler::ha_check_for_upgrade(HA_CHECK_OPT *check_opt)
{
Fix for bug#39585: innodb and myisam corruption after binary upgrade from <=5.0.46 to >=5.0.48 Problem: 'check table .. for upgrade' doesn't detect incompatible collation changes made in 5.0.48. Fix: check for incompatible collation changes.
2008-10-08 15:52:57 +02:00
int error;
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
KEY *keyinfo, *keyend;
KEY_PART_INFO *keypart, *keypartend;
if (!table->s->mysql_version)
{
/* check for blob-in-key error */
keyinfo= table->key_info;
keyend= table->key_info + table->s->keys;
for (; keyinfo < keyend; keyinfo++)
{
keypart= keyinfo->key_part;
keypartend= keypart + keyinfo->key_parts;
for (; keypart < keypartend; keypart++)
{
if (!keypart->fieldnr)
continue;
Field *field= table->field[keypart->fieldnr-1];
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)
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
{
if (check_opt->sql_flags & TT_FOR_UPGRADE)
check_opt->flags= T_MEDIUM;
return HA_ADMIN_NEEDS_CHECK;
}
}
}
}
Fixed bug #36055: mysql_upgrade doesn't really 'upgrade' tables The REPAIR TABLE ... USE_FRM query silently corrupts data of tables with old .FRM file version. The mysql_upgrade client program or the REPAIR TABLE query (without the USE_FRM clause) can't prevent this trouble, because in the common case they don't upgrade .FRM file to compatible structure. 1. Evaluation of the REPAIR TABLE ... USE_FRM query has been modified to reject such tables with the message: "Failed repairing incompatible .FRM file". 2. REPAIR TABLE query (without USE_FRM clause) evaluation has been modified to upgrade .FRM files to current version. 3. CHECK TABLE ... FOR UPGRADE query evaluation has been modified to return error status when .FRM file has incompatible version. 4. mysql_upgrade and mysqlcheck client programs call CHECK TABLE FOR UPGRADE and REPAIR TABLE queries, so their behaviors have been changed too to upgrade .FRM files with incompatible version numbers.
2008-05-12 18:01:13 +02:00
if (table->s->frm_version != FRM_VER_TRUE_VARCHAR)
return HA_ADMIN_NEEDS_ALTER;
Fix for bug#39585: innodb and myisam corruption after binary upgrade from <=5.0.46 to >=5.0.48 Problem: 'check table .. for upgrade' doesn't detect incompatible collation changes made in 5.0.48. Fix: check for incompatible collation changes.
2008-10-08 15:52:57 +02:00
if ((error= check_collation_compatibility()))
return error;
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
return check_for_upgrade(check_opt);
}
int handler::check_old_types()
{
Field** field;
if (!table->s->mysql_version)
{
/* check for bad DECIMAL field */
for (field= table->field; (*field); field++)
{
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)
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
{
return HA_ADMIN_NEEDS_ALTER;
}
Bug#19371 VARBINARY() have trailing zeros after upgrade from 4.1 - Detect if a table has field of type MYSQL_TYPE_VAR_STRING while running "CHECK TABLE t FOR UPGRADE" and indicate it need to be fixed with "REPAIR TABLE t". - When running a "REPAIR TABLE t" or "ALTER TABLE t FORCE" on the above table, install a special copy function to trim off the trailing spaces which we safely can say that the pre 5.0 mysqld didn't put there.
2006-11-09 12:00:27 +01:00
if ((*field)->type() == MYSQL_TYPE_VAR_STRING)
{
return HA_ADMIN_NEEDS_ALTER;
}
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
}
}
return 0;
}
Clean up one place that misuses LOCK_open (dead code)
2007-02-22 17:45:06 +01:00
static bool update_frm_version(TABLE *table)
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
{
char path[FN_REFLEN];
File file;
int result= 1;
DBUG_ENTER("update_frm_version");
BUG#26325 - TEMPORARY TABLE "corrupt" after first read, according to CHECK TABLE CHECK/REPAIR TABLE reports "File not found" error when issued against temporary table. Fixed by disabling a brunch of code (in case it gets temporary table) that is responsible for updating frm version as it is not needed for temporary tables.
2007-07-19 10:51:31 +02:00
/*
No need to update frm version in case table was created or checked
by server with the same version. This also ensures that we do not
update frm version for temporary tables as this code doesn't support
temporary tables.
*/
if (table->s->mysql_version == MYSQL_VERSION_ID)
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
DBUG_RETURN(0);
WL#2645 (CHECK FOR UPGRADE) 5.1-related fixes
2006-02-17 10:37:37 +01:00
strxmov(path, table->s->normalized_path.str, reg_ext, NullS);
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
if ((file= my_open(path, O_RDWR|O_BINARY, MYF(MY_WME))) >= 0)
{
uchar version[4];
WL#2645 (CHECK FOR UPGRADE) 5.1-related fixes
2006-02-17 10:37:37 +01:00
char *key= table->s->table_cache_key.str;
uint key_length= table->s->table_cache_key.length;
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
TABLE *entry;
HASH_SEARCH_STATE state;
int4store(version, MYSQL_VERSION_ID);
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 ((result= my_pwrite(file,(uchar*) version,4,51L,MYF_RW)))
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
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
for (entry=(TABLE*) hash_first(&open_cache,(uchar*) key,key_length, &state);
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
entry;
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
entry= (TABLE*) hash_next(&open_cache,(uchar*) key,key_length, &state))
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
entry->s->mysql_version= MYSQL_VERSION_ID;
}
err:
if (file >= 0)
VOID(my_close(file,MYF(MY_WME)));
DBUG_RETURN(result);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
@return
key if error because of duplicated keys
*/
Import changeset
2000-07-31 21:29:14 +02:00
uint handler::get_dup_key(int error)
{
check for mysql_bin_log.is_open() before my_b_tell(&thd->transaction.trans_log in ha_commit_trans - why it didn't crash earlier ?
2004-08-22 14:23:52 +02:00
DBUG_ENTER("handler::get_dup_key");
Import changeset
2000-07-31 21:29:14 +02:00
table->file->errkey = (uint) -1;
Fixed BUG#9680: Wrong error from cascading update Applied 9680.patch (by Osku Salerma)
2006-02-09 18:43:10 +01:00
if (error == HA_ERR_FOUND_DUPP_KEY || error == HA_ERR_FOREIGN_DUPLICATE_KEY ||
error == HA_ERR_FOUND_DUPP_UNIQUE || error == HA_ERR_NULL_IN_SPATIAL ||
error == HA_ERR_DROP_INDEX_FK)
Backport of bug#38719 from 6.0 to 5.1 handler::get_dup_key used the called handler for the info call, but used table->file handler for errkey. Fixed by using table->file->info instead.
2009-02-18 21:29:30 +01:00
table->file->info(HA_STATUS_ERRKEY | HA_STATUS_NO_LOCK);
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(table->file->errkey);
}
Fix for MacOSX and symlinks Fix for USE_FRM and crashed index file
2003-05-14 00:27:26 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Delete all files with extension from bas_ext().
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param name Base name of table
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
We assume that the handler may return more extensions than
was actually used for the file.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
0 If we successfully deleted at least one file from base_ext and
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
didn't get any other errors than ENOENT
@retval
!0 Error
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
*/
Import changeset
2000-07-31 21:29:14 +02:00
int handler::delete_table(const char *name)
{
Bug#30102: Rename table does corrupt tables with partition files on failure Problem was that a failing rename just left the partitions at the state it was at the failure. Solution was to try to revert the started rename if a failure occured.
2009-07-24 12:11:23 +02:00
int saved_error= 0;
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
int error= 0;
int enoent_or_zero= ENOENT; // Error if no file was deleted
a fix (bug #7205: "archive" test fails on UnixWare when using prepared statements)
2005-03-09 15:33:01 +01:00
char buff[FN_REFLEN];
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
Import changeset
2000-07-31 21:29:14 +02:00
for (const char **ext=bas_ext(); *ext ; 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
fn_format(buff, name, "", *ext, MY_UNPACK_FILENAME|MY_APPEND_EXT);
a fix (bug #7205: "archive" test fails on UnixWare when using prepared statements)
2005-03-09 15:33:01 +01:00
if (my_delete_with_symlink(buff, MYF(0)))
Lots of fixes for BDB tables Change DROP TABLE to first drop the data, then the .frm file
2000-12-08 16:04:57 +01:00
{
Bug#30102: Rename table does corrupt tables with partition files on failure Problem was that a failing rename just left the partitions at the state it was at the failure. Solution was to try to revert the started rename if a failure occured.
2009-07-24 12:11:23 +02:00
if (my_errno != ENOENT)
{
/*
If error on the first existing file, return the error.
Otherwise delete as much as possible.
*/
if (enoent_or_zero)
return my_errno;
saved_error= my_errno;
}
Lots of fixes for BDB tables Change DROP TABLE to first drop the data, then the .frm file
2000-12-08 16:04:57 +01:00
}
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
else
Generate warning in ha_delete_table() if files is missing in handler
2005-02-21 19:41:48 +01:00
enoent_or_zero= 0; // No error for ENOENT
Review of new pushed code (XA & other) Portability fixes and cleanups Fixed setting of 'res' in mysql_execute_command()
2005-02-21 13:47:57 +01:00
error= enoent_or_zero;
Import changeset
2000-07-31 21:29:14 +02:00
}
Bug#30102: Rename table does corrupt tables with partition files on failure Problem was that a failing rename just left the partitions at the state it was at the failure. Solution was to try to revert the started rename if a failure occured.
2009-07-24 12:11:23 +02:00
return saved_error ? saved_error : error;
Import changeset
2000-07-31 21:29:14 +02:00
}
int handler::rename_table(const char * from, const char * to)
{
BUG#9911 After review fixes - Remove ha_archive::rename_table and move the fix to handler::rename_table
2005-04-28 10:49:18 +02:00
int error= 0;
Bug#30102: Rename table does corrupt tables with partition files on failure Problem was that a failing rename just left the partitions at the state it was at the failure. Solution was to try to revert the started rename if a failure occured.
2009-07-24 12:11:23 +02:00
const char **ext, **start_ext;
start_ext= bas_ext();
for (ext= start_ext; *ext ; ext++)
Import changeset
2000-07-31 21:29:14 +02:00
{
BUG#9911 After review fixes - Remove ha_archive::rename_table and move the fix to handler::rename_table
2005-04-28 10:49:18 +02:00
if (rename_file_ext(from, to, *ext))
{
if ((error=my_errno) != ENOENT)
break;
error= 0;
}
Import changeset
2000-07-31 21:29:14 +02:00
}
Bug#30102: Rename table does corrupt tables with partition files on failure Problem was that a failing rename just left the partitions at the state it was at the failure. Solution was to try to revert the started rename if a failure occured.
2009-07-24 12:11:23 +02:00
if (error)
{
/* Try to revert the rename. Ignore errors. */
for (; ext >= start_ext; ext--)
rename_file_ext(to, from, *ext);
}
BUG#9911 After review fixes - Remove ha_archive::rename_table and move the fix to handler::rename_table
2005-04-28 10:49:18 +02:00
return error;
Import changeset
2000-07-31 21:29:14 +02:00
}
WL #528: Faster free_tmp_table
2005-11-06 08:29:03 +01:00
void handler::drop_table(const char *name)
{
close();
delete_table(name);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Performs checks upon the table.
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param thd thread doing CHECK TABLE operation
@param check_opt options from the parser
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
HA_ADMIN_OK Successful upgrade
@retval
HA_ADMIN_NEEDS_UPGRADE Table has structures requiring upgrade
@retval
HA_ADMIN_NEEDS_ALTER Table has structures requiring ALTER TABLE
@retval
HA_ADMIN_NOT_IMPLEMENTED
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
*/
int handler::ha_check(THD *thd, HA_CHECK_OPT *check_opt)
{
int error;
if ((table->s->mysql_version >= MYSQL_VERSION_ID) &&
(check_opt->sql_flags & TT_FOR_UPGRADE))
return 0;
if (table->s->mysql_version < MYSQL_VERSION_ID)
{
if ((error= check_old_types()))
return error;
error= ha_check_for_upgrade(check_opt);
if (error && (error != HA_ADMIN_NEEDS_CHECK))
return error;
if (!error && (check_opt->sql_flags & TT_FOR_UPGRADE))
return 0;
}
if ((error= check(thd, check_opt)))
return error;
Clean up one place that misuses LOCK_open (dead code)
2007-02-22 17:45:06 +01:00
return update_frm_version(table);
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
}
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
/**
A helper function to mark a transaction read-write,
if it is started.
*/
inline
void
handler::mark_trx_read_write()
{
Ha_trx_info *ha_info= &ha_thd()->ha_data[ht->slot].ha_info[0];
/*
When a storage engine method is called, the transaction must
have been started, unless it's a DDL call, for which the
storage engine starts the transaction internally, and commits
it internally, without registering in the ha_list.
Unfortunately here we can't know know for sure if the engine
has registered the transaction or not, so we must check.
*/
if (ha_info->is_started())
{
DBUG_ASSERT(has_transactions());
/*
table_share can be NULL in ha_delete_table(). See implementation
of standalone function ha_delete_table() in sql_base.cc.
*/
if (table_share == NULL || table_share->tmp_table == NO_TMP_TABLE)
ha_info->set_trx_read_write();
}
}
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
/**
Repair table: public interface.
@sa handler::repair()
*/
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
int handler::ha_repair(THD* thd, HA_CHECK_OPT* check_opt)
{
int result;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
if ((result= repair(thd, check_opt)))
return result;
Clean up one place that misuses LOCK_open (dead code)
2007-02-22 17:45:06 +01:00
return update_frm_version(table);
WL#2645 (CHECK TABLE FOR UPGRADE) necessary implementation in the server mysql_upgrade script added
2006-02-17 07:52:32 +01:00
}
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
/**
Bulk update row: public interface.
@sa handler::bulk_update_row()
*/
int
handler::ha_bulk_update_row(const uchar *old_data, uchar *new_data,
uint *dup_key_found)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return bulk_update_row(old_data, new_data, dup_key_found);
}
/**
Delete all rows: public interface.
@sa handler::delete_all_rows()
*/
int
handler::ha_delete_all_rows()
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return delete_all_rows();
}
/**
Reset auto increment: public interface.
@sa handler::reset_auto_increment()
*/
int
handler::ha_reset_auto_increment(ulonglong value)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return reset_auto_increment(value);
}
/**
Backup table: public interface.
@sa handler::backup()
*/
int
handler::ha_backup(THD* thd, HA_CHECK_OPT* check_opt)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return backup(thd, check_opt);
}
/**
Restore table: public interface.
@sa handler::restore()
*/
int
handler::ha_restore(THD* thd, HA_CHECK_OPT* check_opt)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return restore(thd, check_opt);
}
/**
Optimize table: public interface.
@sa handler::optimize()
*/
int
handler::ha_optimize(THD* thd, HA_CHECK_OPT* check_opt)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return optimize(thd, check_opt);
}
/**
Analyze table: public interface.
@sa handler::analyze()
*/
int
handler::ha_analyze(THD* thd, HA_CHECK_OPT* check_opt)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return analyze(thd, check_opt);
}
/**
Check and repair table: public interface.
@sa handler::check_and_repair()
*/
bool
handler::ha_check_and_repair(THD *thd)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return check_and_repair(thd);
}
/**
Disable indexes: public interface.
@sa handler::disable_indexes()
*/
int
handler::ha_disable_indexes(uint mode)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return disable_indexes(mode);
}
/**
Enable indexes: public interface.
@sa handler::enable_indexes()
*/
int
handler::ha_enable_indexes(uint mode)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return enable_indexes(mode);
}
/**
Discard or import tablespace: public interface.
@sa handler::discard_or_import_tablespace()
*/
int
handler::ha_discard_or_import_tablespace(my_bool discard)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return discard_or_import_tablespace(discard);
}
/**
Prepare for alter: public interface.
Called to prepare an *online* ALTER.
@sa handler::prepare_for_alter()
*/
void
handler::ha_prepare_for_alter()
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
prepare_for_alter();
}
/**
Rename table: public interface.
@sa handler::rename_table()
*/
int
handler::ha_rename_table(const char *from, const char *to)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return rename_table(from, to);
}
/**
Delete table: public interface.
@sa handler::delete_table()
*/
int
handler::ha_delete_table(const char *name)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return delete_table(name);
}
/**
Drop table in the engine: public interface.
@sa handler::drop_table()
*/
void
handler::ha_drop_table(const char *name)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return drop_table(name);
}
/**
Create a table in the engine: public interface.
@sa handler::create()
*/
int
handler::ha_create(const char *name, TABLE *form, HA_CREATE_INFO *info)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return create(name, form, info);
}
/**
Create handler files for CREATE TABLE: public interface.
@sa handler::create_handler_files()
*/
int
handler::ha_create_handler_files(const char *name, const char *old_name,
int action_flag, HA_CREATE_INFO *info)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return create_handler_files(name, old_name, action_flag, info);
}
/**
Change partitions: public interface.
@sa handler::change_partitions()
*/
int
handler::ha_change_partitions(HA_CREATE_INFO *create_info,
const char *path,
Aligned declaration and defintion of ha_change_partitions(), to avoid link error using SunStudio 12.
2008-10-11 18:35:49 +02:00
ulonglong * const copied,
ulonglong * const deleted,
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
const uchar *pack_frm_data,
size_t pack_frm_len)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return change_partitions(create_info, path, copied, deleted,
pack_frm_data, pack_frm_len);
}
/**
Drop partitions: public interface.
@sa handler::drop_partitions()
*/
int
handler::ha_drop_partitions(const char *path)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return drop_partitions(path);
}
/**
Rename partitions: public interface.
@sa handler::rename_partitions()
*/
int
handler::ha_rename_partitions(const char *path)
{
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
return rename_partitions(path);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
bug#8151 - truncate leaves a transaction open deadlock in MYSQL_LOG::new_file() style fixes
2005-03-02 10:38:25 +01:00
Tell the storage engine that it is allowed to "disable transaction" in the
handler. It is a hint that ACID is not required - it is used in NDB for
ALTER TABLE, for example, when data are copied to temporary table.
A storage engine may treat this hint any way it likes. NDB for example
starts to commit every now and then automatically.
This hint can be safely ignored.
Pushing all the Gemini changes above the table handler.
2001-05-29 15:29:08 +02:00
*/
renamed ha_recovery_logging to ha_enable_transaction added tests to alter table for "large" alter tables and truncates in ndbcluster added debug printout in restart() in ndbcluster added flag THD::transaction.on to enable/disable transaction
2004-09-03 17:11:09 +02:00
int ha_enable_transaction(THD *thd, bool on)
Pushing all the Gemini changes above the table handler.
2001-05-29 15:29:08 +02:00
{
int error=0;
renamed ha_recovery_logging to ha_enable_transaction added tests to alter table for "large" alter tables and truncates in ndbcluster added debug printout in restart() in ndbcluster added flag THD::transaction.on to enable/disable transaction
2004-09-03 17:11:09 +02:00
DBUG_ENTER("ha_enable_transaction");
Removed extra spaces Added extra debug
2007-10-01 14:32:07 +02:00
DBUG_PRINT("enter", ("on: %d", (int) on));
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
if ((thd->transaction.on= on))
bug#8151 - truncate leaves a transaction open deadlock in MYSQL_LOG::new_file() style fixes
2005-03-02 10:38:25 +01:00
{
/*
Now all storage engines should have transaction handling enabled.
But some may have it enabled all the time - "disabling" transactions
is an optimization hint that storage engine is free to ignore.
So, let's commit an open transaction (if any) now.
*/
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
if (!(error= ha_commit_trans(thd, 0)))
Fix crash in BDB from improper cleanup of transactions, and avoid problem in NDB that the fix would cause due to improper registration of a transaction that isn't meant to be seen by the handler layer. (Bug #14212) Thanks to Sergei Golubchik for helping with this.
2005-12-10 00:00:33 +01:00
error= end_trans(thd, COMMIT);
bug#8151 - truncate leaves a transaction open deadlock in MYSQL_LOG::new_file() style fixes
2005-03-02 10:38:25 +01:00
}
Pushing all the Gemini changes above the table handler.
2001-05-29 15:29:08 +02:00
DBUG_RETURN(error);
}
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
int handler::index_next_same(uchar *buf, const uchar *key, uint keylen)
Import changeset
2000-07-31 21:29:14 +02:00
{
int error;
Bug#22351 - handler::index_next_same() call to key_cmp_if_same() uses the wrong buffer handler::index_next_same() did not take into account that the internally called function key_cmp_if_same() uses the fixed buffer table->record[0] for key comparison instead of the buffer provided by the caller of handler::index_next_same(). Added code to temporarily redirect table->record[0] and the fields used for the key to the record buffer provided by the caller of handler::index_next_same(). The test case is in partition.test already.
2007-11-07 09:30:41 +01:00
DBUG_ENTER("index_next_same");
Import changeset
2000-07-31 21:29:14 +02:00
if (!(error=index_next(buf)))
{
Bug#22351 - handler::index_next_same() call to key_cmp_if_same() uses the wrong buffer handler::index_next_same() did not take into account that the internally called function key_cmp_if_same() uses the fixed buffer table->record[0] for key comparison instead of the buffer provided by the caller of handler::index_next_same(). Added code to temporarily redirect table->record[0] and the fields used for the key to the record buffer provided by the caller of handler::index_next_same(). The test case is in partition.test already.
2007-11-07 09:30:41 +01:00
my_ptrdiff_t ptrdiff= buf - table->record[0];
Bug #43414 Parenthesis (and other) warnings compiling MySQL with gcc 4.3.2 Cleaning up warnings not present in 5.0.
2009-09-23 15:21:29 +02:00
uchar *UNINIT_VAR(save_record_0);
KEY *UNINIT_VAR(key_info);
KEY_PART_INFO *UNINIT_VAR(key_part);
KEY_PART_INFO *UNINIT_VAR(key_part_end);
Bug#22351 - handler::index_next_same() call to key_cmp_if_same() uses the wrong buffer handler::index_next_same() did not take into account that the internally called function key_cmp_if_same() uses the fixed buffer table->record[0] for key comparison instead of the buffer provided by the caller of handler::index_next_same(). Added code to temporarily redirect table->record[0] and the fields used for the key to the record buffer provided by the caller of handler::index_next_same(). The test case is in partition.test already.
2007-11-07 09:30:41 +01:00
/*
key_cmp_if_same() compares table->record[0] against 'key'.
In parts it uses table->record[0] directly, in parts it uses
field objects with their local pointers into table->record[0].
If 'buf' is distinct from table->record[0], we need to move
all record references. This is table->record[0] itself and
the field pointers of the fields used in this key.
*/
if (ptrdiff)
{
save_record_0= table->record[0];
table->record[0]= buf;
key_info= table->key_info + active_index;
key_part= key_info->key_part;
key_part_end= key_part + key_info->key_parts;
for (; key_part < key_part_end; key_part++)
{
DBUG_ASSERT(key_part->field);
key_part->field->move_field_offset(ptrdiff);
}
}
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
if (key_cmp_if_same(table, key, active_index, keylen))
Import changeset
2000-07-31 21:29:14 +02:00
{
table->status=STATUS_NOT_FOUND;
error=HA_ERR_END_OF_FILE;
}
Bug#22351 - handler::index_next_same() call to key_cmp_if_same() uses the wrong buffer handler::index_next_same() did not take into account that the internally called function key_cmp_if_same() uses the fixed buffer table->record[0] for key comparison instead of the buffer provided by the caller of handler::index_next_same(). Added code to temporarily redirect table->record[0] and the fields used for the key to the record buffer provided by the caller of handler::index_next_same(). The test case is in partition.test already.
2007-11-07 09:30:41 +01:00
/* Move back if necessary. */
if (ptrdiff)
{
table->record[0]= save_record_0;
for (key_part= key_info->key_part; key_part < key_part_end; key_part++)
key_part->field->move_field_offset(-ptrdiff);
}
Import changeset
2000-07-31 21:29:14 +02:00
}
Bug#22351 - handler::index_next_same() call to key_cmp_if_same() uses the wrong buffer handler::index_next_same() did not take into account that the internally called function key_cmp_if_same() uses the fixed buffer table->record[0] for key comparison instead of the buffer provided by the caller of handler::index_next_same(). Added code to temporarily redirect table->record[0] and the fields used for the key to the record buffer provided by the caller of handler::index_next_same(). The test case is in partition.test already.
2007-11-07 09:30:41 +01:00
DBUG_RETURN(error);
Import changeset
2000-07-31 21:29:14 +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
void handler::get_dynamic_partition_info(PARTITION_INFO *stat_info,
uint part_id)
WL#2506: Information Schema tables for PARTITIONing added I_S 'PARTITIONS' table
2006-01-10 16:44:04 +01:00
{
info(HA_STATUS_CONST | HA_STATUS_TIME | HA_STATUS_VARIABLE |
HA_STATUS_NO_LOCK);
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
stat_info->records= stats.records;
stat_info->mean_rec_length= stats.mean_rec_length;
stat_info->data_file_length= stats.data_file_length;
stat_info->max_data_file_length= stats.max_data_file_length;
stat_info->index_file_length= stats.index_file_length;
stat_info->delete_length= stats.delete_length;
stat_info->create_time= stats.create_time;
stat_info->update_time= stats.update_time;
stat_info->check_time= stats.check_time;
stat_info->check_sum= 0;
WL#2506: Information Schema tables for PARTITIONing added I_S 'PARTITIONS' table
2006-01-10 16:44:04 +01:00
if (table_flags() & (ulong) HA_HAS_CHECKSUM)
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
stat_info->check_sum= checksum();
WL#2506: Information Schema tables for PARTITIONing added I_S 'PARTITIONS' table
2006-01-10 16:44:04 +01:00
return;
}
Import changeset
2000-07-31 21:29:14 +02:00
/****************************************************************************
** Some general functions that isn't in the handler class
****************************************************************************/
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Initiates table-file and calls appropriate database-creator.
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
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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
0 ok
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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
Added NDB storage engine
2004-04-15 09:14:14 +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 ha_create_table(THD *thd, const char *path,
const char *db, const char *table_name,
HA_CREATE_INFO *create_info,
Import changeset
2000-07-31 21:29:14 +02:00
bool update_create_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
int error= 1;
Import changeset
2000-07-31 21:29:14 +02:00
TABLE table;
Some small portability fixes. Added support for lower_case_table_names=2, which is to be used on case insensitive file systems. This tells MySQL to preserve the used case of filenames and database names to make it esier to move files between cases sensitive can case insensitive file systems (like Windows and Linux)
2003-12-30 12:14:21 +01:00
char name_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
const char *name;
TABLE_SHARE share;
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_ENTER("ha_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
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
init_tmp_table_share(thd, &share, db, 0, table_name, path);
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 (open_table_def(thd, &share, 0) ||
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
open_table_from_share(thd, &share, "", 0, (uint) READ_ALL, 0, &table,
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;
Import changeset
2000-07-31 21:29:14 +02:00
if (update_create_info)
update_create_info_from_table(create_info, &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
Bug#37402: Mysql cant read partitioned table with capital letter in the name Problem was that ha_partition had HA_FILE_BASED flag set (since it uses a .par file), but after open it uses the first partitions flags, which results in different case handling for create and for open. Solution was to change the underlying partition name so it was consistent. (Only happens when lower_case_table_names = 2, i.e. Mac OS X and storage engines without HA_FILE_BASED, like InnoDB and Memory.) (Recommit after adding rename of check_lowercase_names to get_canonical_filename, and moved it from handler.h to mysql_priv.h) NOTE: if a mixed case name for a partitioned table was created when lower_case_table_name = 2 it should be renamed or dropped before using the updated version (See bug#37402 for more info)
2008-07-11 01:14:13 +02:00
name= get_canonical_filename(table.file, share.path.str, name_buff);
Some small portability fixes. Added support for lower_case_table_names=2, which is to be used on case insensitive file systems. This tells MySQL to preserve the used case of filenames and database names to make it esier to move files between cases sensitive can case insensitive file systems (like Windows and Linux)
2003-12-30 12:14:21 +01:00
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
error= table.file->ha_create(name, &table, create_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
VOID(closefrm(&table, 0));
Changed to use my_global.h Fixed problem with LIKE with latin1_de Added parsing support of UNSIGNED LONG LONG
2001-09-14 01:54:33 +02:00
if (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
{
strxmov(name_buff, db, ".", table_name, NullS);
my_error(ER_CANT_CREATE_TABLE, MYF(ME_BELL+ME_WAITTANG), name_buff, error);
}
err:
free_table_share(&share);
Import changeset
2000-07-31 21:29:14 +02:00
DBUG_RETURN(error != 0);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Try to discover table from 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
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
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 found, write the frm file to disk.
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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 Table did not exists
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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
0 Table created ok
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
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
> 0 Error, table existed but could not be created
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +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 ha_create_table_from_engine(THD* thd, const char *db, const char *name)
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
{
true,false -> TRUE, FALSE Simple fixes/optimization of things discovered during review of new pushed code
2004-10-14 17:03:46 +02:00
int error;
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 *frmblob;
size_t frmlen;
Bug#44834 strxnmov is expected to behave as you'd expect The problem: described in the bug report. The fix: --increase buffers where it's necessary (buffers which are used in stxnmov) --decrease buffer lengths which are used
2009-06-19 10:24:43 +02:00
char path[FN_REFLEN + 1];
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
HA_CREATE_INFO create_info;
TABLE 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
TABLE_SHARE share;
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
DBUG_ENTER("ha_create_table_from_engine");
Portability fixes Fixes while reviewing new pushed code NULL as argument to encrypt/decrypt should return NULL without a warning
2005-06-27 19:31:00 +02:00
DBUG_PRINT("enter", ("name '%s'.'%s'", db, name));
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +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
bzero((uchar*) &create_info,sizeof(create_info));
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
if ((error= ha_discover(thd, db, name, &frmblob, &frmlen)))
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
{
Portability fixes Fixes while reviewing new pushed code NULL as argument to encrypt/decrypt should return NULL without a warning
2005-06-27 19:31:00 +02:00
/* Table could not be discovered and thus not created */
BUG10365 Cluster handler uses non-standard error code - Updated after review
2005-06-16 15:17:47 +02:00
DBUG_RETURN(error);
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
}
true,false -> TRUE, FALSE Simple fixes/optimization of things discovered during review of new pushed code
2004-10-14 17:03:46 +02:00
/*
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
Table exists in handler and could be discovered
frmblob and frmlen are set, write the frm to disk
true,false -> TRUE, FALSE Simple fixes/optimization of things discovered during review of new pushed code
2004-10-14 17:03:46 +02:00
*/
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
Bug#44834 strxnmov is expected to behave as you'd expect The problem: described in the bug report. The fix: --increase buffers where it's necessary (buffers which are used in stxnmov) --decrease buffer lengths which are used
2009-06-19 10:24:43 +02:00
build_table_filename(path, sizeof(path) - 1, db, name, "", 0);
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
// Save the frm file
Portability fixes Fixes while reviewing new pushed code NULL as argument to encrypt/decrypt should return NULL without a warning
2005-06-27 19:31:00 +02:00
error= writefrm(path, frmblob, frmlen);
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(frmblob, MYF(0));
Portability fixes Fixes while reviewing new pushed code NULL as argument to encrypt/decrypt should return NULL without a warning
2005-06-27 19:31:00 +02:00
if (error)
BUG10365 Cluster handler uses non-standard error code - Updated after review
2005-06-16 15:17:47 +02:00
DBUG_RETURN(2);
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
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
init_tmp_table_share(thd, &share, db, 0, name, path);
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 (open_table_def(thd, &share, 0))
{
DBUG_RETURN(3);
}
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 (open_table_from_share(thd, &share, "" ,0, 0, 0, &table, 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
{
free_table_share(&share);
BUG10365 Cluster handler uses non-standard error code - Updated after review
2005-06-16 15:17:47 +02:00
DBUG_RETURN(3);
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
}
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
update_create_info_from_table(&create_info, &table);
BUG#14514 Creating table with packed key fails silently
2005-11-02 15:53:04 +01:00
create_info.table_options|= HA_OPTION_CREATE_FROM_ENGINE;
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
Bug#37402: Mysql cant read partitioned table with capital letter in the name Problem was that ha_partition had HA_FILE_BASED flag set (since it uses a .par file), but after open it uses the first partitions flags, which results in different case handling for create and for open. Solution was to change the underlying partition name so it was consistent. (Only happens when lower_case_table_names = 2, i.e. Mac OS X and storage engines without HA_FILE_BASED, like InnoDB and Memory.) (Recommit after adding rename of check_lowercase_names to get_canonical_filename, and moved it from handler.h to mysql_priv.h) NOTE: if a mixed case name for a partitioned table was created when lower_case_table_name = 2 it should be renamed or dropped before using the updated version (See bug#37402 for more info)
2008-07-11 01:14:13 +02:00
get_canonical_filename(table.file, path, path);
A pre-requisite for the fix for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of state" Make private all class handler methods (PSEA API) that may modify data. Introduce and deploy public ha_* wrappers for these methods in all sql/. This necessary to keep track of all data modifications in sql/, which is in turn necessary to be able to optimize two-phase commit of those transactions that do not modify data.
2007-12-20 19:16:55 +01:00
error=table.file->ha_create(path, &table, &create_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
VOID(closefrm(&table, 1));
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
DBUG_RETURN(error != 0);
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
}
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
void st_ha_check_opt::init()
{
flags= sql_flags= 0;
sort_buffer_size = current_thd->variables.myisam_sort_buff_size;
}
/*****************************************************************************
Key cache handling.
This code is only relevant for ISAM/MyISAM tables
key_cache->cache may be 0 only in the case where a key cache is not
initialized or when we where not able to init the key cache in a previous
call to ha_init_key_cache() (probably out of memory)
*****************************************************************************/
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Init a key cache if it has not been initied before.
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
int ha_init_key_cache(const char *name, KEY_CACHE *key_cache)
Import changeset
2000-07-31 21:29:14 +02:00
{
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
DBUG_ENTER("ha_init_key_cache");
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
if (!key_cache->key_cache_inited)
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
{
pthread_mutex_lock(&LOCK_global_system_variables);
Bug#39494 : key_buffer_size > 4GB does not work on 64 bit Windows Cache size is truncated via 32bit ulong in ha_init_key_cache() and ha_resize_key_cache() This change fixes the cast to size_t instead of ulong. This cast is safe, because key_buffer_size parameter is limited to SIZE_T_MAX
2008-11-19 16:02:38 +01:00
size_t tmp_buff_size= (size_t) key_cache->param_buff_size;
Bug#17332 - changing key_buffer_size on a running server can crash under load Resizing a key cache while it was in heavy use could crash the server. There were several race conditions. I reworked some of the algorithms to fix the race conditions. No test case. Repeating the crashes requires heavy concurrent load on the key cache. A test script is attached to the bug report. More explanations to the changes are contained in a text file attached to the bug report.
2007-01-31 18:49:07 +01:00
uint tmp_block_size= (uint) key_cache->param_block_size;
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
uint division_limit= key_cache->param_division_limit;
uint age_threshold= key_cache->param_age_threshold;
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
pthread_mutex_unlock(&LOCK_global_system_variables);
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
DBUG_RETURN(!init_key_cache(key_cache,
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
tmp_block_size,
tmp_buff_size,
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
division_limit, age_threshold));
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
}
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
DBUG_RETURN(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
}
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
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Resize key cache.
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
int ha_resize_key_cache(KEY_CACHE *key_cache)
Many files: Fix after manual merge
2003-10-02 08:37:05 +02:00
{
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
DBUG_ENTER("ha_resize_key_cache");
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
if (key_cache->key_cache_inited)
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
{
pthread_mutex_lock(&LOCK_global_system_variables);
Bug#39494 : key_buffer_size > 4GB does not work on 64 bit Windows Cache size is truncated via 32bit ulong in ha_init_key_cache() and ha_resize_key_cache() This change fixes the cast to size_t instead of ulong. This cast is safe, because key_buffer_size parameter is limited to SIZE_T_MAX
2008-11-19 16:02:38 +01:00
size_t tmp_buff_size= (size_t) key_cache->param_buff_size;
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
long tmp_block_size= (long) key_cache->param_block_size;
uint division_limit= key_cache->param_division_limit;
uint age_threshold= key_cache->param_age_threshold;
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
pthread_mutex_unlock(&LOCK_global_system_variables);
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
DBUG_RETURN(!resize_key_cache(key_cache, tmp_block_size,
tmp_buff_size,
division_limit, age_threshold));
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
}
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
DBUG_RETURN(0);
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
}
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
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
Change parameters for key cache (like size)
*/
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
int ha_change_key_cache_param(KEY_CACHE *key_cache)
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
{
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
if (key_cache->key_cache_inited)
{
pthread_mutex_lock(&LOCK_global_system_variables);
uint division_limit= key_cache->param_division_limit;
uint age_threshold= key_cache->param_age_threshold;
pthread_mutex_unlock(&LOCK_global_system_variables);
change_key_cache_param(key_cache, division_limit, age_threshold);
}
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
return 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
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Free memory allocated by a key cache.
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
int ha_end_key_cache(KEY_CACHE *key_cache)
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
{
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
end_key_cache(key_cache, 1); // Can never fail
Manual merge after improving concurrency for key cache reassignment
2003-10-02 05:20:17 +02:00
return 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
}
Import changeset
2000-07-31 21:29:14 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Move all tables from one key cache to another one.
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Merge key cache structures to one Fixed compiler warnings (IRIX C compiler and VC++)
2003-11-20 21:06:25 +01:00
int ha_change_key_cache(KEY_CACHE *old_key_cache,
KEY_CACHE *new_key_cache)
Import changeset
2000-07-31 21:29:14 +02:00
{
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
mi_change_key_cache(old_key_cache, new_key_cache);
return 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
}
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Try to discover one table from handler(s).
true,false -> TRUE, FALSE Simple fixes/optimization of things discovered during review of new pushed code
2004-10-14 17:03:46 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
-1 Table did not exists
@retval
0 OK. In this case *frmblob and *frmlen are set
@retval
>0 error. frmblob and frmlen may not be set
Added NDB storage engine
2004-04-15 09:14:14 +02:00
*/
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
struct st_discover_args
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
{
const char *db;
const char *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
uchar **frmblob;
size_t *frmlen;
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
};
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
static my_bool discover_handlerton(THD *thd, plugin_ref plugin,
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
void *arg)
{
st_discover_args *vargs= (st_discover_args *)arg;
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
handlerton *hton= plugin_data(plugin, handlerton *);
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
if (hton->state == SHOW_OPTION_YES && hton->discover &&
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
(!(hton->discover(hton, thd, vargs->db, vargs->name,
vargs->frmblob,
vargs->frmlen))))
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
return TRUE;
return FALSE;
}
true,false -> TRUE, FALSE Simple fixes/optimization of things discovered during review of new pushed code
2004-10-14 17:03:46 +02:00
int ha_discover(THD *thd, const char *db, const char *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
uchar **frmblob, size_t *frmlen)
Added NDB storage engine
2004-04-15 09:14:14 +02:00
{
BUG10365 Cluster handler uses non-standard error code - Updated after review
2005-06-16 15:17:47 +02:00
int error= -1; // Table does not exist in any handler
Added NDB storage engine
2004-04-15 09:14:14 +02:00
DBUG_ENTER("ha_discover");
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
DBUG_PRINT("enter", ("db: %s, name: %s", db, name));
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
st_discover_args args= {db, name, frmblob, frmlen};
Bug #12055 NDB temp tables created by ALTER TABLE are usable - do not discover temporary files and make them visible
2005-07-20 18:40:55 +02:00
if (is_prefix(name,tmp_file_prefix)) /* skip temporary tables */
DBUG_RETURN(error);
Cleanup of unused variables. Fixed "discover" in the handler API. Fixed problem where handlerton was not zero'ed. I need to look around, I suspect this problem is more widespread.
2006-08-25 22:31:15 +02:00
if (plugin_foreach(thd, discover_handlerton,
MYSQL_STORAGE_ENGINE_PLUGIN, &args))
error= 0;
Added NDB storage engine
2004-04-15 09:14:14 +02:00
if (!error)
Bug#21554 (sp_cache.cc: violates C++ aliasing rules) The problem reported is a compile bug, reported by the development GCC team with GCC 4.2. The original issue can no longer be reproduced in MySQL 5.1, since the configure script no longer define HAVE_ATOMIC_ADD, which caused the Linux atomic functions to be used (and cause a problem with an invalid cast). This patch implements some code cleanup for 5.1 only, which was identified during the investigation of this issue. With this patch, statistics maintained in THD::status_var are by definition owned by the running thread, and do not need to be protected against race conditions. These statistics are maintained by the status_var_* helpers, which do not require any lock.
2007-05-22 21:41:40 +02:00
status_var_increment(thd->status_var.ha_discover_count);
Added NDB storage engine
2004-04-15 09:14:14 +02:00
DBUG_RETURN(error);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Call this function in order to give the handler the possiblity
to ask engine if there are any new tables that should be written to disk
Changed WL#1424 to use the function ha_find_files. This is a simpler implementation and all handler specific code is hidden in the appropriate handler.
2004-09-21 12:13:58 +02:00
or any dropped tables that need to be removed from disk
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
*/
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
struct st_find_files_args
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
{
const char *db;
const char *path;
const char *wild;
bool dir;
WL#3732 Information schema optimization
2007-08-03 00:14:05 +02:00
List<LEX_STRING> *files;
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
};
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
static my_bool find_files_handlerton(THD *thd, plugin_ref plugin,
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
void *arg)
{
st_find_files_args *vargs= (st_find_files_args *)arg;
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
handlerton *hton= plugin_data(plugin, handlerton *);
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
if (hton->state == SHOW_OPTION_YES && hton->find_files)
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
if (hton->find_files(hton, thd, vargs->db, vargs->path, vargs->wild,
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
vargs->dir, vargs->files))
return TRUE;
return FALSE;
}
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
Changed WL#1424 to use the function ha_find_files. This is a simpler implementation and all handler specific code is hidden in the appropriate handler.
2004-09-21 12:13:58 +02:00
int
ha_find_files(THD *thd,const char *db,const char *path,
WL#3732 Information schema optimization
2007-08-03 00:14:05 +02:00
const char *wild, bool dir, List<LEX_STRING> *files)
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
{
int error= 0;
Changed WL#1424 to use the function ha_find_files. This is a simpler implementation and all handler specific code is hidden in the appropriate handler.
2004-09-21 12:13:58 +02:00
DBUG_ENTER("ha_find_files");
Fixed compiler warnings Fixed compile-pentium64 scripts Fixed wrong estimate of update_with_key_prefix in sql-bench Merge bk-internal.mysql.com:/home/bk/mysql-5.1 into mysql.com:/home/my/mysql-5.1 Fixed unsafe define of uint4korr() Fixed that --extern works with mysql-test-run.pl Small trivial cleanups This also fixes a bug in counting number of rows that are updated when we have many simultanous queries Move all connection handling and command exectuion main loop from sql_parse.cc to sql_connection.cc Split handle_one_connection() into reusable sub functions. Split create_new_thread() into reusable sub functions. Added thread_scheduler; Preliminary interface code for future thread_handling code. Use 'my_thread_id' for internal thread id's Make thr_alarm_kill() to depend on thread_id instead of thread Make thr_abort_locks_for_thread() depend on thread_id instead of thread In store_globals(), set my_thread_var->id to be thd->thread_id. Use my_thread_var->id as basis for my_thread_name() The above changes makes the connection we have between THD and threads more soft. Added a lot of DBUG_PRINT() and DBUG_ASSERT() functions Fixed compiler warnings Fixed core dumps when running with --debug Removed setting of signal masks (was never used) Made event code call pthread_exit() (portability fix) Fixed that event code doesn't call DBUG_xxx functions before my_thread_init() is called. Made handling of thread_id and thd->variables.pseudo_thread_id uniform. Removed one common 'not freed memory' warning from mysqltest Fixed a couple of usage of not initialized warnings (unlikely cases) Suppress compiler warnings from bdb and (for the moment) warnings from ndb
2007-02-23 12:13:55 +01:00
DBUG_PRINT("enter", ("db: '%s' path: '%s' wild: '%s' dir: %d",
db, path, wild ? wild : "NULL", dir));
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
st_find_files_args args= {db, path, wild, dir, files};
plugin_foreach(thd, find_files_handlerton,
MYSQL_STORAGE_ENGINE_PLUGIN, &args);
/* The return value is not currently used */
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
DBUG_RETURN(error);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Ask handler if the table exists in engine.
@retval
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
HA_ERR_NO_SUCH_TABLE Table does not exist
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
HA_ERR_TABLE_EXIST Table exists
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
\# Error code
*/
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
struct st_table_exists_in_engine_args
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
{
const char *db;
const char *name;
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
int err;
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
};
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
static my_bool table_exists_in_engine_handlerton(THD *thd, plugin_ref plugin,
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
void *arg)
{
st_table_exists_in_engine_args *vargs= (st_table_exists_in_engine_args *)arg;
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
handlerton *hton= plugin_data(plugin, handlerton *);
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
int err= HA_ERR_NO_SUCH_TABLE;
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
if (hton->state == SHOW_OPTION_YES && hton->table_exists_in_engine)
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
err = hton->table_exists_in_engine(hton, thd, vargs->db, vargs->name);
vargs->err = err;
if (vargs->err == HA_ERR_TABLE_EXIST)
return TRUE;
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
return FALSE;
}
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
int ha_table_exists_in_engine(THD* thd, const char* db, const char* name)
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
{
BUG#10365 Cluster handler uses non-standard error codes - Added better error messages when trying to open a table that can't be discovered or unpacked. The most likely cause of this is that it does not have any frm data, probably since it has been created from NdbApi or is a NDB system table. - Separated functionality that was in ha_create_table_from_engine into two functions. One that checks if the table exists and another one that tries to create the table from the engine.
2005-06-08 13:31:59 +02:00
DBUG_ENTER("ha_table_exists_in_engine");
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
DBUG_PRINT("enter", ("db: %s, name: %s", db, name));
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
st_table_exists_in_engine_args args= {db, name, HA_ERR_NO_SUCH_TABLE};
plugin_foreach(thd, table_exists_in_engine_handlerton,
This pulls two function calls which should have been handlerton calls out of handler.cc.
2006-08-30 21:28:25 +02:00
MYSQL_STORAGE_ENGINE_PLUGIN, &args);
BUG#18676 In order to coincide with 5.0 mysqld error code after bug#18676, Map the 4009 ndb error code to 157 mysql error code
2007-04-16 09:15:47 +02:00
DBUG_PRINT("exit", ("error: %d", args.err));
DBUG_RETURN(args.err);
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
}
wl2325 wl2324
2006-01-12 19:51:02 +01:00
#ifdef HAVE_NDB_BINLOG
/*
TODO: change this into a dynamic struct
List<handlerton> does not work as
1. binlog_end is called when MEM_ROOT is gone
2. cannot work with thd MEM_ROOT as memory should be freed
*/
#define MAX_HTON_LIST_ST 63
struct hton_list_st
{
handlerton *hton[MAX_HTON_LIST_ST];
uint sz;
};
struct binlog_func_st
{
enum_binlog_func fn;
void *arg;
};
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
wl2325 wl2324
2006-01-12 19:51:02 +01:00
Listing handlertons first to avoid recursive calls and deadlock
*/
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
static my_bool binlog_func_list(THD *thd, plugin_ref plugin, void *arg)
wl2325 wl2324
2006-01-12 19:51:02 +01:00
{
hton_list_st *hton_list= (hton_list_st *)arg;
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
handlerton *hton= plugin_data(plugin, handlerton *);
wl2325 wl2324
2006-01-12 19:51:02 +01:00
if (hton->state == SHOW_OPTION_YES && hton->binlog_func)
{
uint sz= hton_list->sz;
if (sz == MAX_HTON_LIST_ST-1)
{
/* list full */
return FALSE;
}
hton_list->hton[sz]= hton;
hton_list->sz= sz+1;
}
return FALSE;
}
static my_bool binlog_func_foreach(THD *thd, binlog_func_st *bfn)
{
hton_list_st hton_list;
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
uint i, sz;
wl2325 wl2324
2006-01-12 19:51:02 +01:00
hton_list.sz= 0;
plugin_foreach(thd, binlog_func_list,
MYSQL_STORAGE_ENGINE_PLUGIN, &hton_list);
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
for (i= 0, sz= hton_list.sz; i < sz ; i++)
hton_list.hton[i]->binlog_func(hton_list.hton[i], thd, bfn->fn, bfn->arg);
wl2325 wl2324
2006-01-12 19:51:02 +01:00
return FALSE;
}
int ha_reset_logs(THD *thd)
{
binlog_func_st bfn= {BFN_RESET_LOGS, 0};
binlog_func_foreach(thd, &bfn);
return 0;
}
void ha_reset_slave(THD* thd)
{
binlog_func_st bfn= {BFN_RESET_SLAVE, 0};
binlog_func_foreach(thd, &bfn);
}
void ha_binlog_wait(THD* thd)
{
binlog_func_st bfn= {BFN_BINLOG_WAIT, 0};
binlog_func_foreach(thd, &bfn);
}
int ha_binlog_end(THD* thd)
{
binlog_func_st bfn= {BFN_BINLOG_END, 0};
binlog_func_foreach(thd, &bfn);
return 0;
}
int ha_binlog_index_purge_file(THD *thd, const char *file)
{
binlog_func_st bfn= {BFN_BINLOG_PURGE_FILE, (void *)file};
binlog_func_foreach(thd, &bfn);
Solaris compile fixes.
2006-01-14 11:06:44 +01:00
return 0;
wl2325 wl2324
2006-01-12 19:51:02 +01:00
}
struct binlog_log_query_st
{
enum_binlog_command binlog_command;
const char *query;
uint query_length;
const char *db;
const char *table_name;
};
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
static my_bool binlog_log_query_handlerton2(THD *thd,
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
handlerton *hton,
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
void *args)
wl2325 wl2324
2006-01-12 19:51:02 +01:00
{
struct binlog_log_query_st *b= (struct binlog_log_query_st*)args;
if (hton->state == SHOW_OPTION_YES && hton->binlog_log_query)
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->binlog_log_query(hton, thd,
wl2325 wl2324
2006-01-12 19:51:02 +01:00
b->binlog_command,
b->query,
b->query_length,
b->db,
b->table_name);
return FALSE;
}
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
static my_bool binlog_log_query_handlerton(THD *thd,
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 plugin,
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
void *args)
{
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
return binlog_log_query_handlerton2(thd, plugin_data(plugin, handlerton *), args);
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
}
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
void ha_binlog_log_query(THD *thd, handlerton *hton,
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
enum_binlog_command binlog_command,
wl2325 wl2324
2006-01-12 19:51:02 +01:00
const char *query, uint query_length,
const char *db, const char *table_name)
{
struct binlog_log_query_st b;
b.binlog_command= binlog_command;
b.query= query;
b.query_length= query_length;
b.db= db;
b.table_name= table_name;
Bug #17038, distribution of schema operation to multiple binlogs missing/multiple entries, partial fix - log alter table directly in server instead of in handler - acknowledge alter table _after_ all binlog events have been processed
2006-02-06 11:47:12 +01:00
if (hton == 0)
plugin_foreach(thd, binlog_log_query_handlerton,
MYSQL_STORAGE_ENGINE_PLUGIN, &b);
else
binlog_log_query_handlerton2(thd, hton, &b);
wl2325 wl2324
2006-01-12 19:51:02 +01:00
}
#endif
WL1424 Multiple MySQL Servers: SHOW TABLES etc. should detect new and delete old tables.
2004-09-13 14:46:38 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
Read the first row of a multi-range set.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param found_range_p Returns a pointer to the element in 'ranges' that
corresponds to the returned row.
@param ranges An array of KEY_MULTI_RANGE range descriptions.
@param range_count Number of ranges in 'ranges'.
@param sorted If result should be sorted per key.
@param buffer A HANDLER_BUFFER for internal handler usage.
@note
- Record is read into table->record[0].
- *found_range_p returns a valid value only if read_multi_range_first()
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
returns 0.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
- Sorting is done within each range. If you want an overall sort, enter
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
'ranges' with sorted ranges.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
0 OK, found a row
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
HA_ERR_END_OF_FILE No rows in range
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
\# Error code
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
*/
int handler::read_multi_range_first(KEY_MULTI_RANGE **found_range_p,
KEY_MULTI_RANGE *ranges, uint range_count,
bool sorted, HANDLER_BUFFER *buffer)
{
int result= HA_ERR_END_OF_FILE;
DBUG_ENTER("handler::read_multi_range_first");
multi_range_sorted= sorted;
multi_range_buffer= buffer;
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
table->mark_columns_used_by_index_no_reset(active_index, table->read_set);
table->column_bitmaps_set(table->read_set, table->write_set);
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
for (multi_range_curr= ranges, multi_range_end= ranges + range_count;
multi_range_curr < multi_range_end;
multi_range_curr++)
{
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
result= read_range_first(multi_range_curr->start_key.keypart_map ?
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
&multi_range_curr->start_key : 0,
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
multi_range_curr->end_key.keypart_map ?
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
&multi_range_curr->end_key : 0,
test(multi_range_curr->range_flag & EQ_RANGE),
multi_range_sorted);
if (result != HA_ERR_END_OF_FILE)
break;
}
*found_range_p= multi_range_curr;
DBUG_PRINT("exit",("result %d", result));
DBUG_RETURN(result);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
Read the next row of a multi-range set.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param found_range_p Returns a pointer to the element in 'ranges' that
corresponds to the returned row.
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
- Record is read into table->record[0].
- *found_range_p returns a valid value only if read_multi_range_next()
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
returns 0.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
0 OK, found a row
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
HA_ERR_END_OF_FILE No (more) rows in range
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
\# Error code
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
*/
int handler::read_multi_range_next(KEY_MULTI_RANGE **found_range_p)
{
Bug#58057: 5.1 libmysql/libmysql.c unused variable/compile failure Bug#57995: Compiler flag change build error on OSX 10.4: my_getncpus.c Bug#57996: Compiler flag change build error on OSX 10.5 : bind.c Bug#57994: Compiler flag change build error : my_redel.c Bug#57993: Compiler flag change build error on FreeBsd 7.0 : regexec.c Bug#57992: Compiler flag change build error on FreeBsd : mf_keycache.c Bug#57997: Compiler flag change build error on OSX 10.6: debug_sync.cc Fix assorted compiler generated warnings.
2010-11-10 22:14:47 +01:00
int UNINIT_VAR(result);
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
DBUG_ENTER("handler::read_multi_range_next");
/* We should not be called after the last call returned EOF. */
DBUG_ASSERT(multi_range_curr < multi_range_end);
do
{
/* Save a call if there can be only one row in range. */
if (multi_range_curr->range_flag != (UNIQUE_RANGE | EQ_RANGE))
{
result= read_range_next();
/* On success or non-EOF errors jump to the end. */
if (result != HA_ERR_END_OF_FILE)
break;
}
else
{
Fixed bug #31310. Locked rows of the InnoDB storage was silently skipped in the read-committed isolation level. QUICK_RANGE_SELECT for unique ranges lacks second (blocking) read of the record that was read semi-consistently and just skip it. The handler::read_multi_range_next method has been modified to retry previous unique range if the previous read was semi-consistent.
2007-10-08 22:07:15 +02:00
if (was_semi_consistent_read())
goto scan_it_again;
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
/*
We need to set this for the last range only, but checking this
condition is more expensive than just setting the result code.
*/
result= HA_ERR_END_OF_FILE;
}
Fixed bug #31310. Locked rows of the InnoDB storage was silently skipped in the read-committed isolation level. QUICK_RANGE_SELECT for unique ranges lacks second (blocking) read of the record that was read semi-consistently and just skip it. The handler::read_multi_range_next method has been modified to retry previous unique range if the previous read was semi-consistent.
2007-10-08 22:07:15 +02:00
multi_range_curr++;
scan_it_again:
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
/* Try the next range(s) until one matches a record. */
Fixed bug #31310. Locked rows of the InnoDB storage was silently skipped in the read-committed isolation level. QUICK_RANGE_SELECT for unique ranges lacks second (blocking) read of the record that was read semi-consistently and just skip it. The handler::read_multi_range_next method has been modified to retry previous unique range if the previous read was semi-consistent.
2007-10-08 22:07:15 +02:00
for (; multi_range_curr < multi_range_end; multi_range_curr++)
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
{
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
result= read_range_first(multi_range_curr->start_key.keypart_map ?
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
&multi_range_curr->start_key : 0,
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
multi_range_curr->end_key.keypart_map ?
WL#2126 - Multi_read_range. Added the required structures and functions for handing over multiple key ranges to the table handler.
2004-12-23 21:45:10 +01:00
&multi_range_curr->end_key : 0,
test(multi_range_curr->range_flag & EQ_RANGE),
multi_range_sorted);
if (result != HA_ERR_END_OF_FILE)
break;
}
}
while ((result == HA_ERR_END_OF_FILE) &&
(multi_range_curr < multi_range_end));
*found_range_p= multi_range_curr;
DBUG_PRINT("exit",("handler::read_multi_range_next: result %d", result));
DBUG_RETURN(result);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Read first row between two ranges.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
Store ranges for future calls to read_range_next.
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param start_key Start key. Is 0 if no min range
@param end_key End key. Is 0 if no max range
@param eq_range_arg Set to 1 if start_key == end_key
@param sorted Set to 1 if result should be sorted per key
@note
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Record is read into table->record[0]
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
0 Found row
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
HA_ERR_END_OF_FILE No rows in range
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
\# Error code
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
*/
int handler::read_range_first(const key_range *start_key,
const key_range *end_key,
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
bool eq_range_arg, bool sorted)
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
{
int result;
DBUG_ENTER("handler::read_range_first");
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
eq_range= eq_range_arg;
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
end_range= 0;
if (end_key)
{
end_range= &save_end_range;
save_end_range= *end_key;
key_compare_result_on_equal= ((end_key->flag == HA_READ_BEFORE_KEY) ? 1 :
(end_key->flag == HA_READ_AFTER_KEY) ? -1 : 0);
}
range_key_part= table->key_info[active_index].key_part;
if (!start_key) // Read first record
result= index_first(table->record[0]);
else
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
result= index_read_map(table->record[0],
start_key->key,
start_key->keypart_map,
start_key->flag);
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
if (result)
Removed unnecessary condition
2004-09-07 16:21:29 +02:00
DBUG_RETURN((result == HA_ERR_KEY_NOT_FOUND)
? HA_ERR_END_OF_FILE
: result);
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
DBUG_RETURN (compare_key(end_range) <= 0 ? 0 : HA_ERR_END_OF_FILE);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Read next row between two ranges.
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@note
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Record is read into table->record[0]
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
0 Found row
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
HA_ERR_END_OF_FILE No rows in range
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
\# Error code
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
*/
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
int handler::read_range_next()
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
{
int result;
DBUG_ENTER("handler::read_range_next");
if (eq_range)
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
{
/* We trust that index_next_same always gives a row in range */
DBUG_RETURN(index_next_same(table->record[0],
end_range->key,
end_range->length));
}
result= index_next(table->record[0]);
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
if (result)
DBUG_RETURN(result);
DBUG_RETURN(compare_key(end_range) <= 0 ? 0 : HA_ERR_END_OF_FILE);
}
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Compare if found key (in row) is over max-value.
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@param range range to compare to row. May be 0 for no range
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@seealso
key.cc::key_cmp()
@return
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
The return value is SIGN(key_in_row - range_key):
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
- 0 : Key is equal to range or 'range' == 0 (no range)
- -1 : Key is less than range
- 1 : Key is larger than range
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
*/
int handler::compare_key(key_range *range)
{
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
int cmp;
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
if (!range)
return 0; // No max range
key_cmp -> key_cmp_if_same New records_in_range() interface (similar to read_range()) Macros for faster bitmap handling Simplify read_range() code (#WL1786) New general key_cmp() function to compare keys
2004-05-16 13:48:32 +02:00
cmp= key_cmp(range_key_part, range->key, range->length);
if (!cmp)
cmp= key_compare_result_on_equal;
return cmp;
Moved reading of ranges from opt_range.cc to handler.cc This gives the handler more optimization possiblities and is needed for NDB cluster Fixed not-initialized memory error detected by valgrind
2004-04-08 12:58:06 +02:00
}
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
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
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
int handler::index_read_idx_map(uchar * buf, uint index, const uchar * key,
key_part_map keypart_map,
enum ha_rkey_function find_flag)
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
{
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
int error, error1;
error= index_init(index, 0);
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
if (!error)
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
{
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
error= index_read_map(buf, key, keypart_map, find_flag);
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
error1= index_end();
}
return error ? error : error1;
handler interface cleanups: more logical table/index_flags return HA_ERR_WRONG_COMMAND instead of abstract methods where appropriate max_keys and other limits renamed to max_supported_keys/etc max_keys/etc are now wrappers to max_supported_keys/etc ha_index_init/ha_rnd_init/ha_index_end/ha_rnd_end are now wrappers to real {index,rnd}_{init,end} to enforce strict pairing
2004-06-23 12:29:05 +02:00
}
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
/**
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
Returns a list of all known extensions.
No mutexes, worst case race is a minor surplus memory allocation
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
We have to recreate the extension map if mysqld is restarted (for example
within libmysqld)
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
Doxygenization of comments.
2007-10-11 19:29:09 +02:00
@retval
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
pointer pointer to TYPELIB structure
*/
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
static my_bool exts_handlerton(THD *unused, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
{
List<char> *found_exts= (List<char> *) arg;
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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
handler *file;
if (hton->state == SHOW_OPTION_YES && hton->create &&
This removes the passing of global hton to engine instance.
2006-09-30 02:19:02 +02:00
(file= hton->create(hton, (TABLE_SHARE*) 0, current_thd->mem_root)))
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
{
List_iterator_fast<char> it(*found_exts);
const char **ext, *old_ext;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
for (ext= file->bas_ext(); *ext; ext++)
{
while ((old_ext= it++))
{
if (!strcmp(old_ext, *ext))
break;
}
if (!old_ext)
found_exts->push_back((char *) *ext);
it.rewind();
}
delete file;
}
return FALSE;
}
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
TYPELIB *ha_known_exts(void)
{
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
if (!known_extensions.type_names || mysys_usage_id != known_extensions_id)
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
{
List<char> found_exts;
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
const char **ext, *old_ext;
known_extensions_id= mysys_usage_id;
This the 4-th patch in scope of CS patch (BUG#11986). The patch contains the following changes: - Introduce auxilary functions to convenient work with character sets: - resolve_charset(); - resolve_collation(); - get_default_db_collation(); - Introduce lex_string_set(); - Refactor Table_trigger_list::process_triggers() & sp_head::execute_trigger() to be consistent with other code; - Move reusable code from add_table_for_trigger() into build_trn_path(), check_trn_exists() and load_table_name_for_trigger() to be used in the following patch. - Rename triggers_file_ext and trigname_file_ext into TRN_EXT and TRG_EXT respectively.
2007-06-14 17:23:55 +02:00
found_exts.push_back((char*) TRG_EXT);
found_exts.push_back((char*) TRN_EXT);
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
plugin_foreach(NULL, exts_handlerton,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
MYSQL_STORAGE_ENGINE_PLUGIN, &found_exts);
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
ext= (const char **) my_once_alloc(sizeof(char *)*
(found_exts.elements+1),
MYF(MY_WME | MY_FAE));
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
Remove compiler warnings and remove not used variables (Found during build process)
2005-02-25 15:53:22 +01:00
DBUG_ASSERT(ext != 0);
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
known_extensions.count= found_exts.elements;
known_extensions.type_names= ext;
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
List_iterator_fast<char> it(found_exts);
Added possibility to detect if libmysqld is restarted (Needed to check if memory allocated with mysql_once_init() has been freed)
2004-11-10 23:36:18 +01:00
while ((old_ext= it++))
*ext++= old_ext;
*ext= 0;
Bug#6031 - To drop database you have to execute DROP DATABASE command twice. DROP DATABASE failed because of file ext not in TYPELIB of known extensions. General solution - construct a TYPELIB at runtime instead of a static list.
2004-11-10 16:07:11 +01:00
}
return &known_extensions;
}
Many files: Semi-synchronous replication for InnoDB type tables; before telling the client that a commit has been processed, wait that the replication thread has returned from my_net_send() where it sends the binlog to the slave; note that TCP/IP, even with the TCP_NODELAY option does not guarantee that the slave has RECEIVED the data - this is just heuristic at the moment; this is useful in failover: in almost all cases, every transaction that has returned from the commit has been sent and processed in the slave, which makes failover to the slave simpler if the master crashes; the code does not work yet as is, because MySQL should call innobase_report_binlog_offset_and_commit() in a commit; we will most probably return that call to 5.0.x, to make InnoDB Hot Backup and group commit to work again; XA code broke them temporarily in 5.0.3
2005-04-20 18:27:46 +02:00
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
static bool stat_print(THD *thd, const char *type, uint type_len,
const char *file, uint file_len,
const char *status, uint status_len)
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
{
Protocol *protocol= thd->protocol;
protocol->prepare_for_resend();
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
protocol->store(type, type_len, system_charset_info);
protocol->store(file, file_len, system_charset_info);
protocol->store(status, status_len, system_charset_info);
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
if (protocol->write())
return TRUE;
return FALSE;
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +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
static my_bool showstat_handlerton(THD *thd, plugin_ref plugin,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
void *arg)
{
enum ha_stat_type stat= *(enum ha_stat_type *) arg;
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
handlerton *hton= plugin_data(plugin, handlerton *);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (hton->state == SHOW_OPTION_YES && hton->show_status &&
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
hton->show_status(hton, thd, stat_print, stat))
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return TRUE;
return FALSE;
}
bool ha_show_status(THD *thd, handlerton *db_type, enum ha_stat_type stat)
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
{
List<Item> field_list;
Protocol *protocol= thd->protocol;
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
bool result;
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
field_list.push_back(new Item_empty_string("Type",10));
field_list.push_back(new Item_empty_string("Name",FN_REFLEN));
field_list.push_back(new Item_empty_string("Status",10));
if (protocol->send_fields(&field_list,
Protocol::SEND_NUM_ROWS | Protocol::SEND_EOF))
return TRUE;
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (db_type == NULL)
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
{
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
result= plugin_foreach(thd, showstat_handlerton,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
MYSQL_STORAGE_ENGINE_PLUGIN, &stat);
}
else
{
if (db_type->state != SHOW_OPTION_YES)
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
{
after review fixes
2006-05-31 18:07:32 +02:00
const LEX_STRING *name=&hton2plugin[db_type->slot]->name;
result= stat_print(thd, name->str, name->length,
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
"", 0, "DISABLED", 8) ? 1 : 0;
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
else
handlerton cleanup: duplicate fields removed, st_mysql_storage_engine added to support run-time handlerton initialization (no compiler warnings), handler API is now tied to MySQL version, handlerton->plugin mapping added (slot-based), dummy default_hton removed, plugin-type-specific initialization generalized, built-in plugins are now initialized too, --default-storage-engine no longer needs a list of storage engines in handle_options(). mysql-test-run.pl bugfixes
2006-05-28 14:51:01 +02:00
result= db_type->show_status &&
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
db_type->show_status(db_type, thd, stat_print, stat) ? 1 : 0;
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
}
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
if (!result)
Rename send_eof() to my_eof() for consistency with my_ok() and my_error()
2008-02-19 13:58:08 +01:00
my_eof(thd);
Finalize storage engine plugins Give BerkeleyDB savepoints Remove "enum db_type" from most of the code
2005-12-21 19:18:40 +01:00
return result;
Make storage engines "pluggable", handlerton work
2005-11-07 16:25:06 +01:00
}
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
/*
Function to check if the conditions for row-based binlogging is
correct for the table.
A row in the given table should be replicated if:
WL#3023 (RBR: Use locks in a statement-like manner): More adaptions to make it work with existing code base.
2006-03-03 09:49:46 +01:00
- Row-based replication is enabled in the current thread
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
- The binlog is enabled
WL#3023 (RBR: Use locks in a statement-like manner): More adaptions to make it work with existing code base.
2006-03-03 09:49:46 +01:00
- It is not a temporary table
- The binary log is open
- The database the table resides in shall be binlogged (binlog_*_db rules)
Fix to not replicate mysql.event table Test fixes
2006-03-09 15:59:29 +01:00
- table is not mysql.event
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
*/
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
static bool check_table_binlog_row_based(THD *thd, TABLE *table)
{
if (table->s->cached_row_logging_check == -1)
Fix template problem with Sun compiler by using anonymous namespace instead of static for referenged function.
2006-01-09 11:08:23 +01:00
{
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
int const check(table->s->tmp_table == NO_TMP_TABLE &&
binlog_filter->db_ok(table->s->db.str));
table->s->cached_row_logging_check= check;
}
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
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
DBUG_ASSERT(table->s->cached_row_logging_check == 0 ||
table->s->cached_row_logging_check == 1);
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
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
return (thd->current_stmt_binlog_row_based &&
table->s->cached_row_logging_check &&
(thd->options & OPTION_BIN_LOG) &&
mysql_bin_log.is_open());
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
Bug#19995 (Extreneous table maps generated for statements that do not generate rows): Switched to writing out table maps for tables that are locked when the first row in a statement is seen.
2006-05-31 19:21:52 +02:00
Write table maps for all (manually or automatically) locked tables
to the binary log.
Bug#19995 (Extreneous table maps generated for statements that does not generate rows): Comment change.
2006-06-02 15:47:06 +02:00
SYNOPSIS
write_locked_table_maps()
thd Pointer to THD structure
DESCRIPTION
This function will generate and write table maps for all tables
that are locked by the thread 'thd'. Either manually locked
(stored in THD::locked_tables) and automatically locked (stored
in THD::lock) are considered.
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
Bug#19995 (Extreneous table maps generated for statements that does not generate rows): Comment change.
2006-06-02 15:47:06 +02:00
RETURN VALUE
0 All OK
1 Failed to write all table maps
SEE ALSO
THD::lock
THD::locked_tables
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
*/
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
static int write_locked_table_maps(THD *thd)
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
{
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
DBUG_ENTER("write_locked_table_maps");
DBUG_PRINT("enter", ("thd: 0x%lx thd->lock: 0x%lx thd->locked_tables: 0x%lx "
"thd->extra_lock: 0x%lx",
(long) thd, (long) thd->lock,
(long) thd->locked_tables, (long) thd->extra_lock));
Fixing code to avoid compile problem on Solaris (aCC).
2006-06-15 10:11:50 +02:00
Bug #34707: Row based replication: slave creates table within wrong database The failure was caused by executing a CREATE-SELECT statement that creates a table in another database than the current one. In row-based logging, the CREATE statement was written to the binary log without the database, hence creating the table in the wrong database, causing the following inserts to fail since the table didn't exist in the given database. Fixed the bug by adding a parameter to store_create_info() that will make the function print the database name before the table name and used that in the calls that write the CREATE statement to the binary log. The database name is only printed if it is different than the currently selected database. The output of SHOW CREATE TABLE has not changed and is still printed without the database name.
2008-10-08 11:15:00 +02:00
DBUG_PRINT("debug", ("get_binlog_table_maps(): %d", thd->get_binlog_table_maps()));
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
if (thd->get_binlog_table_maps() == 0)
{
MYSQL_LOCK *locks[3];
locks[0]= thd->extra_lock;
locks[1]= thd->lock;
locks[2]= thd->locked_tables;
for (uint i= 0 ; i < sizeof(locks)/sizeof(*locks) ; ++i )
Bug#19995 (Extreneous table maps generated for statements that do not generate rows): Switched to writing out table maps for tables that are locked when the first row in a statement is seen.
2006-05-31 19:21:52 +02:00
{
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
MYSQL_LOCK const *const lock= locks[i];
if (lock == NULL)
continue;
TABLE **const end_ptr= lock->table + lock->table_count;
for (TABLE **table_ptr= lock->table ;
table_ptr != end_ptr ;
++table_ptr)
Bug#19995 (Extreneous table maps generated for statements that do not generate rows): Switched to writing out table maps for tables that are locked when the first row in a statement is seen.
2006-05-31 19:21:52 +02:00
{
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
TABLE *const table= *table_ptr;
DBUG_PRINT("info", ("Checking table %s", table->s->table_name.str));
if (table->current_lock == F_WRLCK &&
check_table_binlog_row_based(thd, table))
Fixing code to avoid compile problem on Solaris (aCC).
2006-06-15 10:11:50 +02:00
{
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
int const has_trans= table->file->has_transactions();
int const error= thd->binlog_write_table_map(table, has_trans);
/*
If an error occurs, it is the responsibility of the caller to
roll back the transaction.
*/
if (unlikely(error))
DBUG_RETURN(1);
Fixing code to avoid compile problem on Solaris (aCC).
2006-06-15 10:11:50 +02:00
}
Bug#19995 (Extreneous table maps generated for statements that do not generate rows): Switched to writing out table maps for tables that are locked when the first row in a statement is seen.
2006-05-31 19:21:52 +02:00
}
}
}
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
DBUG_RETURN(0);
}
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 #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
typedef bool Log_func(THD*, TABLE*, bool, MY_BITMAP*,
uint, const uchar*, const uchar*);
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
static int binlog_log_row(TABLE* table,
const uchar *before_record,
const uchar *after_record,
Log_func *log_func)
{
if (table->no_replicate)
return 0;
bool error= 0;
THD *const thd= table->in_use;
if (check_table_binlog_row_based(thd, table))
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 #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
MY_BITMAP cols;
/* Potential buffer on the stack for the bitmap */
uint32 bitbuf[BITMAP_STACKBUF_SIZE/sizeof(uint32)];
uint n_fields= table->s->fields;
my_bool use_bitbuf= n_fields <= sizeof(bitbuf)*8;
Bug#19995 (Extreneous table maps generated for statements that do not generate rows): Switched to writing out table maps for tables that are locked when the first row in a statement is seen.
2006-05-31 19:21:52 +02:00
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
/*
If there are no table maps written to the binary log, this is
the first row handled in this statement. In that case, we need
to write table maps for all locked tables to the binary log.
*/
if (likely(!(error= bitmap_init(&cols,
use_bitbuf ? bitbuf : NULL,
(n_fields + 7) & ~7UL,
FALSE))))
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
{
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
bitmap_set_all(&cols);
if (likely(!(error= write_locked_table_maps(thd))))
error= (*log_func)(thd, table, table->file->has_transactions(),
&cols, table->s->fields,
before_record, after_record);
if (!use_bitbuf)
bitmap_free(&cols);
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
}
}
Bug #34458 Extreneous use of templates in server code Patch to remove white-space left over after removing anonymous namespace.
2008-02-15 13:06:28 +01:00
return error ? HA_ERR_RBR_LOGGING_FAILED : 0;
Fixing code to avoid compile problem on Solaris (aCC).
2006-06-15 10:11:50 +02:00
}
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
int handler::ha_external_lock(THD *thd, int lock_type)
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
{
WL#3023 (Use locks in a statement-like manner): Table maps are now written on aquiring locks to tables and released at the end of each logical statement.
2006-02-16 08:30:53 +01:00
DBUG_ENTER("handler::ha_external_lock");
WL#3146 "less locking in auto_increment": this is a cleanup patch for our current auto_increment handling: new names for auto_increment variables in THD, new methods to manipulate them (see sql_class.h), some move into handler::, causing less backup/restore work when executing substatements. This makes the logic hopefully clearer, less work is is needed in mysql_insert(). By cleaning up, using different variables for different purposes (instead of one for 3 things...), we fix those bugs, which someone may want to fix in 5.0 too: BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate statement-based" BUG#20341 "stored function inserting into one auto_increment puts bad data in slave" BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE" (now if a row is updated, LAST_INSERT_ID() will return its id) and re-fixes: BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT" (already fixed differently by Ramil in 4.1) Test of documented behaviour of mysql_insert_id() (there was no test). The behaviour changes introduced are: - LAST_INSERT_ID() now returns "the first autogenerated auto_increment value successfully inserted", instead of "the first autogenerated auto_increment value if any row was successfully inserted", see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY UPDATE, see auto_increment.test. Same for mysql_insert_id(), see mysql_client_test.c. - LAST_INSERT_ID() does not change if no autogenerated value was successfully inserted (it used to then be 0), see auto_increment.test. - if in INSERT SELECT no autogenerated value was successfully inserted, mysql_insert_id() now returns the id of the last inserted row (it already did this for INSERT VALUES), see mysql_client_test.c. - if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X (it already did this for INSERT VALUES), see mysql_client_test.c. - NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE, the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID influences not only the first row now. Additionally, when unlocking a table we check that the thread is not keeping a next_insert_id (as the table is unlocked that id is potentially out-of-date); forgetting about this next_insert_id is done in a new handler::ha_release_auto_increment(). Finally we prepare for engines capable of reserving finite-length intervals of auto_increment values: we store such intervals in THD. The next step (to be done by the replication team in 5.1) is to read those intervals from THD and actually store them in the statement-based binary log. NDB will be a good engine to test that.
2006-07-09 17:52:19 +02:00
/*
Whether this is lock or unlock, this should be true, and is to verify that
if get_auto_increment() was called (thus may have reserved intervals or
taken a table lock), ha_release_auto_increment() was too.
*/
DBUG_ASSERT(next_insert_id == 0);
WL#3303 (RBR: Engine-controlled logging format): Moving code to check storage engine capabilities to after tables are locked. Moving code to cache table flags so that table flags are read from the storage engine at the beginning of the statement in addition to when the storage engine is opened. To handle CREATE-SELECT, the decision function is called after the table is created and it is called with all tables that are in the select part of the statement as well as the newly created table.
2007-05-29 17:13:17 +02:00
/*
We cache the table flags if the locking succeeded. Otherwise, we
keep them as they were when they were fetched in ha_open().
*/
int error= external_lock(thd, lock_type);
if (error == 0)
cached_table_flags= table_flags();
DBUG_RETURN(error);
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
}
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
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
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
Check handler usage and reset state of file to after 'open'
*/
int handler::ha_reset()
{
DBUG_ENTER("ha_reset");
Fix VC 2005 compile problem
2006-08-23 15:41:52 +02:00
/* Check that we have called all proper deallocation functions */
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
DBUG_ASSERT((uchar*) table->def_read_set.bitmap +
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
table->s->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
(uchar*) table->def_write_set.bitmap);
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(bitmap_is_set_all(&table->s->all_set));
DBUG_ASSERT(table->key_read == 0);
/* ensure that ha_index_end / ha_rnd_end has been called */
DBUG_ASSERT(inited == NONE);
/* Free cache used by filesort */
free_io_cache(table);
Bug#33375 all_set corrupted on table object - make sure to reset the read and write sets handler.cc, log_event.cc: - make sure to reset the read and write sets
2008-01-15 17:01:58 +01:00
/* reset the bitmaps to point to defaults */
table->default_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
DBUG_RETURN(reset());
}
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
int handler::ha_write_row(uchar *buf)
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
{
int error;
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
Log_func *log_func= Write_rows_log_event::binlog_row_logging_function;
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
DBUG_ENTER("handler::ha_write_row");
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
if (unlikely(error= write_row(buf)))
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
DBUG_RETURN(error);
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
if (unlikely(error= binlog_log_row(table, 0, buf, log_func)))
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
DBUG_RETURN(error); /* purecov: inspected */
DBUG_RETURN(0);
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
}
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
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
int handler::ha_update_row(const uchar *old_data, uchar *new_data)
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
{
int error;
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
Log_func *log_func= Update_rows_log_event::binlog_row_logging_function;
Bug#18436 (RBR: Replication to partition engine triggers assertion on slave side): Partial fix for resolving the problem. Swapping contents of record[0] and record[1] since this is what some storage engines expect.
2006-03-23 20:13:44 +01:00
/*
Some storage engines require that the new record is in record[0]
(and the old record is in record[1]).
*/
DBUG_ASSERT(new_data == table->record[0]);
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
if (unlikely(error= update_row(old_data, new_data)))
return error;
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
if (unlikely(error= binlog_log_row(table, old_data, new_data, log_func)))
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
return error;
return 0;
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +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
int handler::ha_delete_row(const uchar *buf)
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
{
int error;
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
Log_func *log_func= Delete_rows_log_event::binlog_row_logging_function;
A fix and a test case for Bug#12713 "Error in a stored function called from a SELECT doesn't cause ROLLBACK of statem". The idea of the fix is to ensure that we always commit the current statement at the end of dispatch_command(). In order to not issue redundant disc syncs, an optimization of the two-phase commit protocol is implemented to bypass the two phase commit if the transaction is read-only.
2008-02-19 12:43:01 +01:00
mark_trx_read_write();
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
if (unlikely(error= delete_row(buf)))
return error;
Bug #34458 Extreneous use of templates in server code Replacing a template function with a normal static function. The template parameter, which previously was the class to find a binlogging function in, is now passed as a pointer to the actual binlogging function instead. The patch requires change of indention, but that is submitted as a separate patch.
2008-02-15 12:56:44 +01:00
if (unlikely(error= binlog_log_row(table, buf, 0, log_func)))
WL#3023 (RBR: Use locks in a statment-like manner): Interface changes pushed early. Separation of public and implementation interface for external_lock() in preparation for implementation.
2006-01-26 09:25:37 +01:00
return error;
return 0;
WL#1012: All changes as one single changeset. This includes both code and test cases.
2005-12-22 06:39:02 +01:00
}
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +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
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
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
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
use_hidden_primary_key() 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
*/
void handler::use_hidden_primary_key()
{
/* fallback to use all columns in the table to identify row */
table->use_all_columns();
}
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
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
Dummy function which accept information about log files which is not need
by handlers
*/
void signal_log_not_needed(struct handlerton, char *log_file)
{
DBUG_ENTER("signal_log_not_needed");
DBUG_PRINT("enter", ("logfile '%s'", log_file));
DBUG_VOID_RETURN;
}
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
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
#ifdef TRANS_LOG_MGM_EXAMPLE_CODE
/*
Example of transaction log management functions based on assumption that logs
placed into a directory
*/
#include <my_dir.h>
#include <my_sys.h>
int example_of_iterator_using_for_logs_cleanup(handlerton *hton)
{
void *buffer;
int res= 1;
struct handler_iterator iterator;
struct handler_log_file_data data;
if (!hton->create_iterator)
return 1; /* iterator creator is not supported */
This patch adds handlerton passing to functions. NDB and Innodb still require a global hanlderton in the main code due to the nature of the sql_cache call back function (should be solveable... another patch). Partitioning now has a flag to allow disabling of engines from being compatible with partitioning. Cleaned up heap naming convention on panic call.
2006-09-30 21:49:46 +02:00
if ((*hton->create_iterator)(hton, HA_TRANSACTLOG_ITERATOR, &iterator) !=
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
HA_ITERATOR_OK)
{
/* error during creation of log iterator or iterator is not supported */
return 1;
}
while((*iterator.next)(&iterator, (void*)&data) == 0)
{
printf("%s\n", data.filename.str);
if (data.status == HA_LOG_STATUS_FREE &&
my_delete(data.filename.str, MYF(MY_WME)))
goto err;
}
res= 0;
err:
(*iterator.destroy)(&iterator);
return res;
}
/*
Here we should get info from handler where it save logs but here is
just example, so we use constant.
IMHO FN_ROOTDIR ("/") is safe enough for example, because nobody has
rights on it except root and it consist of directories only at lest for
*nix (sorry, can't find windows-safe solution here, but it is only example).
*/
#define fl_dir FN_ROOTDIR
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
Dummy function to return log status should be replaced by function which
really detect the log status and check that the file is a log of this
handler.
*/
enum log_status fl_get_log_status(char *log)
{
MY_STAT stat_buff;
if (my_stat(log, &stat_buff, MYF(0)))
return HA_LOG_STATUS_INUSE;
return HA_LOG_STATUS_NOSUCHLOG;
}
struct fl_buff
{
LEX_STRING *names;
enum log_status *statuses;
uint32 entries;
uint32 current;
};
int fl_log_iterator_next(struct handler_iterator *iterator,
void *iterator_object)
{
struct fl_buff *buff= (struct fl_buff *)iterator->buffer;
struct handler_log_file_data *data=
(struct handler_log_file_data *) iterator_object;
if (buff->current >= buff->entries)
return 1;
data->filename= buff->names[buff->current];
data->status= buff->statuses[buff->current];
buff->current++;
return 0;
}
void fl_log_iterator_destroy(struct handler_iterator *iterator)
{
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*)iterator->buffer, MYF(MY_ALLOW_ZERO_PTR));
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
}
Changed comments before functions to Doxygen style.
2007-01-31 00:15:22 +01:00
/** @brief
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
returns buffer, to be assigned in handler_iterator struct
*/
enum handler_create_iterator_result
fl_log_iterator_buffer_init(struct handler_iterator *iterator)
{
MY_DIR *dirp;
struct fl_buff *buff;
char *name_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
uchar *ptr;
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
FILEINFO *file;
uint32 i;
/* to be able to make my_free without crash in case of error */
iterator->buffer= 0;
if (!(dirp = my_dir(fl_dir, MYF(0))))
{
return HA_ITERATOR_ERROR;
}
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 ((ptr= (uchar*)my_malloc(ALIGN_SIZE(sizeof(fl_buff)) +
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
((ALIGN_SIZE(sizeof(LEX_STRING)) +
sizeof(enum log_status) +
Bug#44834 strxnmov is expected to behave as you'd expect The problem: described in the bug report. The fix: --increase buffers where it's necessary (buffers which are used in stxnmov) --decrease buffer lengths which are used
2009-06-19 10:24:43 +02:00
+ FN_REFLEN + 1) *
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
(uint) dirp->number_off_files),
MYF(0))) == 0)
{
return HA_ITERATOR_ERROR;
}
buff= (struct fl_buff *)ptr;
buff->entries= buff->current= 0;
ptr= ptr + (ALIGN_SIZE(sizeof(fl_buff)));
buff->names= (LEX_STRING*) (ptr);
ptr= ptr + ((ALIGN_SIZE(sizeof(LEX_STRING)) *
(uint) dirp->number_off_files));
buff->statuses= (enum log_status *)(ptr);
name_ptr= (char *)(ptr + (sizeof(enum log_status) *
(uint) dirp->number_off_files));
for (i=0 ; i < (uint) dirp->number_off_files ; i++)
{
enum log_status st;
file= dirp->dir_entry + i;
if ((file->name[0] == '.' &&
((file->name[1] == '.' && file->name[2] == '\0') ||
file->name[1] == '\0')))
continue;
if ((st= fl_get_log_status(file->name)) == HA_LOG_STATUS_NOSUCHLOG)
continue;
name_ptr= strxnmov(buff->names[buff->entries].str= name_ptr,
FN_REFLEN, fl_dir, file->name, NullS);
buff->names[buff->entries].length= (name_ptr -
Bug#44834 strxnmov is expected to behave as you'd expect The problem: described in the bug report. The fix: --increase buffers where it's necessary (buffers which are used in stxnmov) --decrease buffer lengths which are used
2009-06-19 10:24:43 +02:00
buff->names[buff->entries].str);
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
buff->statuses[buff->entries]= st;
buff->entries++;
}
iterator->buffer= buff;
iterator->next= &fl_log_iterator_next;
iterator->destroy= &fl_log_iterator_destroy;
return HA_ITERATOR_OK;
}
/* An example of a iterator creator */
enum handler_create_iterator_result
fl_create_iterator(enum handler_iterator_type type,
struct handler_iterator *iterator)
{
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
switch(type) {
interface for transaction log management added to handlerton iterators creation interface added to handlerton
2006-06-01 20:47:15 +02:00
case HA_TRANSACTLOG_ITERATOR:
return fl_log_iterator_buffer_init(iterator);
default:
return HA_ITERATOR_UNSUPPORTED;
}
}
#endif /*TRANS_LOG_MGM_EXAMPLE_CODE*/
Reference in a new issue Copy permalink