Base code and idea from a patch from by plinux at Taobao.
The idea is that we mark all memory that are thread specific with MY_THREAD_SPECIFIC.
Memory counting is done per thread in the my_malloc_size_cb_func callback function from my_malloc().
There are plenty of new asserts to ensure that for a debug server the counting is correct.
Information_schema.processlist gets two new columns: MEMORY_USED and EXAMINED_ROWS.
- The later is there mainly to show how query is progressing.
The following changes in interfaces was needed to get this to work:
- init_alloc_root() amd init_sql_alloc() has extra option so that one can mark memory with MY_THREAD_SPECIFIC
- One now have to use alloc_root_set_min_malloc() to set min memory to be allocated by alloc_root()
- my_init_dynamic_array() has extra option so that one can mark memory with MY_THREAD_SPECIFIC
- my_net_init() has extra option so that one can mark memory with MY_THREAD_SPECIFIC
- Added flag for hash_init() so that one can mark hash table to be thread specific.
- Added flags to init_tree() so that one can mark tree to be thread specific.
- Removed with_delete option to init_tree(). Now one should instead use MY_TREE_WITH_DELETE_FLAG.
- Added flag to Warning_info::Warning_info() if the structure should be fully initialized.
- String elements can now be marked as thread specific.
- Internal HEAP tables are now marking it's memory as MY_THREAD_SPECIFIC.
- Changed type of myf from int to ulong, as this is always a set of bit flags.
Other things:
- Removed calls to net_end() and thd->cleanup() as these are now done in ~THD()
- We now also show EXAMINED_ROWS in SHOW PROCESSLIST
- Added new variable 'memory_used'
- Fixed bug where kill_threads_for_user() was using the wrong mem_root to allocate memory.
- Removed calls to the obsoleted function init_dynamic_array()
- Use set_current_thd() instead of my_pthread_setspecific_ptr(THR_THD,...)
client/completion_hash.cc:
Updated call to init_alloc_root()
client/mysql.cc:
Updated call to init_alloc_root()
client/mysqlbinlog.cc:
init_dynamic_array() -> my_init_dynamic_array()
Updated call to init_alloc_root()
client/mysqlcheck.c:
Updated call to my_init_dynamic_array()
client/mysqldump.c:
Updated call to init_alloc_root()
client/mysqltest.cc:
Updated call to init_alloc_root()
Updated call to my_init_dynamic_array()
Fixed compiler warnings
extra/comp_err.c:
Updated call to my_init_dynamic_array()
extra/resolve_stack_dump.c:
Updated call to my_init_dynamic_array()
include/hash.h:
Added HASH_THREAD_SPECIFIC
include/heap.h:
Added flag is internal temporary table.
include/my_dir.h:
Safety fix: Ensure that MY_DONT_SORT and MY_WANT_STAT don't interfer with other mysys flags
include/my_global.h:
Changed type of myf from int to ulong, as this is always a set of bit flags.
include/my_sys.h:
Added MY_THREAD_SPECIFIC and MY_THREAD_MOVE
Added malloc_flags to DYNAMIC_ARRAY
Added extra mysys flag argument to my_init_dynamic_array()
Removed deprecated functions init_dynamic_array() and my_init_dynamic_array.._ci
Updated paramaters for init_alloc_root()
include/my_tree.h:
Added my_flags to allow one to use MY_THREAD_SPECIFIC with hash tables.
Removed with_delete. One should now instead use MY_TREE_WITH_DELETE_FLAG
Updated parameters to init_tree()
include/myisamchk.h:
Added malloc_flags to allow one to use MY_THREAD_SPECIFIC for checks.
include/mysql.h:
Added MYSQL_THREAD_SPECIFIC_MALLOC
Used 'unused1' to mark memory as thread specific.
include/mysql.h.pp:
Updated file
include/mysql_com.h:
Used 'unused1' to mark memory as thread specific.
Updated parameters for my_net_init()
libmysql/libmysql.c:
Updated call to init_alloc_root() to mark memory thread specific.
libmysqld/emb_qcache.cc:
Updated call to init_alloc_root()
libmysqld/lib_sql.cc:
Updated call to init_alloc_root()
mysql-test/r/create.result:
Updated results
mysql-test/r/user_var.result:
Updated results
mysql-test/suite/funcs_1/datadict/processlist_priv.inc:
Update to handle new format of SHOW PROCESSLIST
mysql-test/suite/funcs_1/datadict/processlist_val.inc:
Update to handle new format of SHOW PROCESSLIST
mysql-test/suite/funcs_1/r/is_columns_is.result:
Update to handle new format of SHOW PROCESSLIST
mysql-test/suite/funcs_1/r/processlist_priv_no_prot.result:
Updated results
mysql-test/suite/funcs_1/r/processlist_val_no_prot.result:
Updated results
mysql-test/t/show_explain.test:
Fixed usage of debug variable so that one can run test with --debug
mysql-test/t/user_var.test:
Added test of memory_usage variable.
mysys/array.c:
Added extra my_flags option to init_dynamic_array() and init_dynamic_array2() so that one can mark memory with MY_THREAD_SPECIFIC
All allocated memory is marked with the given my_flags.
Removed obsolete function init_dynamic_array()
mysys/default.c:
Updated call to init_alloc_root()
Updated call to my_init_dynamic_array()
mysys/hash.c:
Updated call to my_init_dynamic_array_ci().
Allocated memory is marked with MY_THREAD_SPECIFIC if HASH_THREAD_SPECIFIC is used.
mysys/ma_dyncol.c:
init_dynamic_array() -> my_init_dynamic_array()
Added #if to get rid of compiler warnings
mysys/mf_tempdir.c:
Updated call to my_init_dynamic_array()
mysys/my_alloc.c:
Added extra parameter to init_alloc_root() so that one can mark memory with MY_THREAD_SPECIFIC
Extend MEM_ROOT with a flag if memory is thread specific.
This is stored in block_size, to keep the size of the MEM_ROOT object identical as before.
Allocated memory is marked with MY_THREAD_SPECIFIC if used with init_alloc_root()
mysys/my_chmod.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_chsize.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_copy.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_create.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_delete.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_error.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_fopen.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_fstream.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_getwd.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_lib.c:
Updated call to init_alloc_root()
Updated call to my_init_dynamic_array()
Updated DBUG_PRINT because of change of myf type
mysys/my_lock.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_malloc.c:
Store at start of each allocated memory block the size of the block and if the block is thread specific.
Call malloc_size_cb_func, if set, with the memory allocated/freed.
Updated DBUG_PRINT because of change of myf type
mysys/my_open.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_pread.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_read.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_redel.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_rename.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_seek.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_sync.c:
Updated DBUG_PRINT because of change of myf type
mysys/my_thr_init.c:
Ensure that one can call my_thread_dbug_id() even if thread is not properly initialized.
mysys/my_write.c:
Updated DBUG_PRINT because of change of myf type
mysys/mysys_priv.h:
Updated parameters to sf_malloc and sf_realloc()
mysys/safemalloc.c:
Added checking that for memory marked with MY_THREAD_SPECIFIC that it's the same thread that is allocation and freeing the memory.
Added sf_malloc_dbug_id() to allow MariaDB to specify which THD is handling the memory.
Added my_flags arguments to sf_malloc() and sf_realloc() to be able to mark memory with MY_THREAD_SPECIFIC.
Added sf_report_leaked_memory() to get list of memory not freed by a thread.
mysys/tree.c:
Added flags to init_tree() so that one can mark tree to be thread specific.
Removed with_delete option to init_tree(). Now one should instead use MY_TREE_WITH_DELETE_FLAG.
Updated call to init_alloc_root()
All allocated memory is marked with the given malloc flags
mysys/waiting_threads.c:
Updated call to my_init_dynamic_array()
sql-common/client.c:
Updated call to init_alloc_root() and my_net_init() to mark memory thread specific.
Updated call to my_init_dynamic_array().
Added MYSQL_THREAD_SPECIFIC_MALLOC so that client can mark memory as MY_THREAD_SPECIFIC.
sql-common/client_plugin.c:
Updated call to init_alloc_root()
sql/debug_sync.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/event_scheduler.cc:
Removed calls to net_end() as this is now done in ~THD()
Call set_current_thd() to ensure that memory is assigned to right thread.
sql/events.cc:
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/filesort.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/filesort_utils.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/ha_ndbcluster.cc:
Updated call to init_alloc_root()
Updated call to my_net_init()
Removed calls to net_end() and thd->cleanup() as these are now done in ~THD()
sql/ha_ndbcluster_binlog.cc:
Updated call to my_net_init()
Updated call to init_sql_alloc()
Removed calls to net_end() and thd->cleanup() as these are now done in ~THD()
sql/ha_partition.cc:
Updated call to init_alloc_root()
sql/handler.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
Added missing call to my_dir_end()
sql/item_func.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/item_subselect.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/item_sum.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/log.cc:
More DBUG
Updated call to init_alloc_root()
sql/mdl.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/mysqld.cc:
Added total_memory_used
Updated call to init_alloc_root()
Move mysql_cond_broadcast() before my_thread_end()
Added mariadb_dbug_id() to count memory per THD instead of per thread.
Added my_malloc_size_cb_func() callback function for my_malloc() to count memory.
Move initialization of mysqld_server_started and mysqld_server_initialized earlier.
Updated call to my_init_dynamic_array().
Updated call to my_net_init().
Call my_pthread_setspecific_ptr(THR_THD,...) to ensure that memory is assigned to right thread.
Added status variable 'memory_used'.
Updated call to init_alloc_root()
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/mysqld.h:
Added set_current_thd()
sql/net_serv.cc:
Added new parameter to my_net_init() so that one can mark memory with MY_THREAD_SPECIFIC.
Store in net->thread_specific_malloc if memory is thread specific.
Mark memory to be thread specific if requested.
sql/opt_range.cc:
Updated call to my_init_dynamic_array()
Updated call to init_sql_alloc()
Added MY_THREAD_SPECIFIC to allocated memory.
sql/opt_subselect.cc:
Updated call to init_sql_alloc() to mark memory thread specific.
sql/protocol.cc:
Fixed compiler warning
sql/records.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/rpl_filter.cc:
Updated call to my_init_dynamic_array()
sql/rpl_handler.cc:
Updated call to my_init_dynamic_array2()
sql/rpl_handler.h:
Updated call to init_sql_alloc()
sql/rpl_mi.cc:
Updated call to my_init_dynamic_array()
sql/rpl_tblmap.cc:
Updated call to init_alloc_root()
sql/rpl_utility.cc:
Updated call to my_init_dynamic_array()
sql/slave.cc:
Initialize things properly before calling functions that allocate memory.
Removed calls to net_end() as this is now done in ~THD()
sql/sp_head.cc:
Updated call to init_sql_alloc()
Updated call to my_init_dynamic_array()
Added parameter to warning_info() that it should be fully initialized.
sql/sp_pcontext.cc:
Updated call to my_init_dynamic_array()
sql/sql_acl.cc:
Updated call to init_sql_alloc()
Updated call to my_init_dynamic_array()
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_admin.cc:
Added parameter to warning_info() that it should be fully initialized.
sql/sql_analyse.h:
Updated call to init_tree() to mark memory thread specific.
sql/sql_array.h:
Updated call to my_init_dynamic_array() to mark memory thread specific.
sql/sql_audit.cc:
Updated call to my_init_dynamic_array()
sql/sql_base.cc:
Updated call to init_sql_alloc()
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_cache.cc:
Updated comment
sql/sql_class.cc:
Added parameter to warning_info() that not initialize it until THD is fully created.
Updated call to init_sql_alloc()
Mark THD::user_vars has to be thread specific.
Updated call to my_init_dynamic_array()
Ensure that memory allocated by THD is assigned to the THD.
More DBUG
Always acll net_end() in ~THD()
Assert that all memory signed to this THD is really deleted at ~THD.
Fixed set_status_var_init() to not reset memory_used.
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_class.h:
Added MY_THREAD_SPECIFIC to allocated memory.
Added malloc_size to THD to record allocated memory per THD.
sql/sql_delete.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/sql_error.cc:
Added 'initialize' parameter to Warning_info() to say if should allocate memory for it's structures.
This is used by THD::THD() to not allocate memory until THD is ready.
Added Warning_info::free_memory()
sql/sql_error.h:
Updated Warning_info() class.
sql/sql_handler.cc:
Updated call to init_alloc_root() to mark memory thread specific.
sql/sql_insert.cc:
More DBUG
sql/sql_join_cache.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/sql_lex.cc:
Updated call to my_init_dynamic_array()
sql/sql_lex.h:
Updated call to my_init_dynamic_array()
sql/sql_load.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/sql_parse.cc:
Removed calls to net_end() and thd->cleanup() as these are now done in ~THD()
Ensure that examined_row_count() is reset before query.
Fixed bug where kill_threads_for_user() was using the wrong mem_root to allocate memory.
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
Don't restore thd->status_var.memory_used when restoring thd->status_var
sql/sql_plugin.cc:
Updated call to init_alloc_root()
Updated call to my_init_dynamic_array()
Don't allocate THD on the stack, as this causes problems with valgrind when doing thd memory counting.
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_prepare.cc:
Added parameter to warning_info() that it should be fully initialized.
Updated call to init_sql_alloc() to mark memory thread specific.
sql/sql_reload.cc:
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_select.cc:
Updated call to my_init_dynamic_array() and init_sql_alloc() to mark memory thread specific.
Added MY_THREAD_SPECIFIC to allocated memory.
More DBUG
sql/sql_servers.cc:
Updated call to init_sql_alloc() to mark memory some memory thread specific.
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_show.cc:
Updated call to my_init_dynamic_array()
Mark my_dir() memory thread specific.
Use my_pthread_setspecific_ptr(THR_THD,...) to mark that allocated memory should be allocated to calling thread.
More DBUG.
Added malloc_size and examined_row_count to SHOW PROCESSLIST.
Added MY_THREAD_SPECIFIC to allocated memory.
Updated call to init_sql_alloc()
Added parameter to warning_info() that it should be fully initialized.
sql/sql_statistics.cc:
Fixed compiler warning
sql/sql_string.cc:
String elements can now be marked as thread specific.
sql/sql_string.h:
String elements can now be marked as thread specific.
sql/sql_table.cc:
Updated call to init_sql_alloc() and my_malloc() to mark memory thread specific
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
Fixed compiler warning
sql/sql_test.cc:
Updated call to my_init_dynamic_array() to mark memory thread specific.
sql/sql_trigger.cc:
Updated call to init_sql_alloc()
sql/sql_udf.cc:
Updated call to init_sql_alloc()
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/sql_update.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
sql/table.cc:
Updated call to init_sql_alloc().
Mark memory used by temporary tables, that are not for slave threads, as MY_THREAD_SPECIFIC
Updated call to init_sql_alloc()
sql/thr_malloc.cc:
Added my_flags argument to init_sql_alloc() to be able to mark memory as MY_THREAD_SPECIFIC.
sql/thr_malloc.h:
Updated prototype for init_sql_alloc()
sql/tztime.cc:
Updated call to init_sql_alloc()
Updated call to init_alloc_root() to mark memory thread specific.
my_pthread_setspecific_ptr(THR_THD,...) -> set_current_thd()
sql/uniques.cc:
Updated calls to init_tree(), my_init_dynamic_array() and my_malloc() to mark memory thread specific.
sql/unireg.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
storage/csv/ha_tina.cc:
Updated call to init_alloc_root()
storage/federated/ha_federated.cc:
Updated call to init_alloc_root()
Updated call to my_init_dynamic_array()
Ensure that memory allocated by fedarated is registered for the system, not for the thread.
storage/federatedx/federatedx_io_mysql.cc:
Updated call to my_init_dynamic_array()
storage/federatedx/ha_federatedx.cc:
Updated call to init_alloc_root()
Updated call to my_init_dynamic_array()
storage/heap/ha_heap.cc:
Added MY_THREAD_SPECIFIC to allocated memory.
storage/heap/heapdef.h:
Added parameter to hp_get_new_block() to be able to do thread specific memory tagging.
storage/heap/hp_block.c:
Added parameter to hp_get_new_block() to be able to do thread specific memory tagging.
storage/heap/hp_create.c:
- Internal HEAP tables are now marking it's memory as MY_THREAD_SPECIFIC.
- Use MY_TREE_WITH_DELETE instead of removed option 'with_delete'.
storage/heap/hp_open.c:
Internal HEAP tables are now marking it's memory as MY_THREAD_SPECIFIC.
storage/heap/hp_write.c:
Added new parameter to hp_get_new_block()
storage/maria/ma_bitmap.c:
Updated call to my_init_dynamic_array()
storage/maria/ma_blockrec.c:
Updated call to my_init_dynamic_array()
storage/maria/ma_check.c:
Updated call to init_alloc_root()
storage/maria/ma_ft_boolean_search.c:
Updated calls to init_tree() and init_alloc_root()
storage/maria/ma_ft_nlq_search.c:
Updated call to init_tree()
storage/maria/ma_ft_parser.c:
Updated call to init_tree()
Updated call to init_alloc_root()
storage/maria/ma_loghandler.c:
Updated call to my_init_dynamic_array()
storage/maria/ma_open.c:
Updated call to my_init_dynamic_array()
storage/maria/ma_sort.c:
Updated call to my_init_dynamic_array()
storage/maria/ma_write.c:
Updated calls to my_init_dynamic_array() and init_tree()
storage/maria/maria_pack.c:
Updated call to init_tree()
storage/maria/unittest/sequence_storage.c:
Updated call to my_init_dynamic_array()
storage/myisam/ft_boolean_search.c:
Updated call to init_tree()
Updated call to init_alloc_root()
storage/myisam/ft_nlq_search.c:
Updated call to init_tree()
storage/myisam/ft_parser.c:
Updated call to init_tree()
Updated call to init_alloc_root()
storage/myisam/ft_stopwords.c:
Updated call to init_tree()
storage/myisam/mi_check.c:
Updated call to init_alloc_root()
storage/myisam/mi_write.c:
Updated call to my_init_dynamic_array()
Updated call to init_tree()
storage/myisam/myisamlog.c:
Updated call to init_tree()
storage/myisam/myisampack.c:
Updated call to init_tree()
storage/myisam/sort.c:
Updated call to my_init_dynamic_array()
storage/myisammrg/ha_myisammrg.cc:
Updated call to init_sql_alloc()
storage/perfschema/pfs_check.cc:
Rest current_thd
storage/perfschema/pfs_instr.cc:
Removed DBUG_ENTER/DBUG_VOID_RETURN as at this point my_thread_var is not allocated anymore, which can cause problems.
support-files/compiler_warnings.supp:
Disable compiler warning from offsetof macro.
and small collateral changes
mysql-test/lib/My/Test.pm:
somehow with "print" we get truncated writes sometimes
mysql-test/suite/perfschema/r/digest_table_full.result:
md5 hashes of statement digests differ, because yacc token codes are different in mariadb
mysql-test/suite/perfschema/r/dml_handler.result:
host table is not ported over yet
mysql-test/suite/perfschema/r/information_schema.result:
host table is not ported over yet
mysql-test/suite/perfschema/r/nesting.result:
this differs, because we don't rewrite general log queries, and multi-statement
packets are logged as a one entry. this result file is identical to what mysql-5.6.5
produces with the --log-raw option.
mysql-test/suite/perfschema/r/relaylog.result:
MariaDB modifies the binlog index file directly, while MySQL 5.6 has a feature "crash-safe binlog index" and modifies a special "crash-safe" shadow copy of the index file and then moves it over. That's why this test shows "NONE" index file writes in MySQL and "MANY" in MariaDB.
mysql-test/suite/perfschema/r/server_init.result:
MariaDB initializes the "manager" resources from the "manager" thread, and starts this thread only when --flush-time is not 0. MySQL 5.6 initializes "manager" resources unconditionally on server startup.
mysql-test/suite/perfschema/r/stage_mdl_global.result:
this differs, because MariaDB disables query cache when query_cache_size=0. MySQL does not
do that, and this causes useless mutex locks and waits.
mysql-test/suite/perfschema/r/statement_digest.result:
md5 hashes of statement digests differ, because yacc token codes are different in mariadb
mysql-test/suite/perfschema/r/statement_digest_consumers.result:
md5 hashes of statement digests differ, because yacc token codes are different in mariadb
mysql-test/suite/perfschema/r/statement_digest_long_query.result:
md5 hashes of statement digests differ, because yacc token codes are different in mariadb
mysql-test/suite/rpl/r/rpl_mixed_drop_create_temp_table.result:
will be updated to match 5.6 when alfranio.correia@oracle.com-20110512172919-c1b5kmum4h52g0ni and anders.song@greatopensource.com-20110105052107-zoab0bsf5a6xxk2y are merged
mysql-test/suite/rpl/r/rpl_non_direct_mixed_mixing_engines.result:
will be updated to match 5.6 when anders.song@greatopensource.com-20110105052107-zoab0bsf5a6xxk2y is merged
Analysis:
When the method JOIN::choose_subquery_plan() decided to apply
the IN-TO-EXISTS strategy, it set the unit and select_lex
uncacheable flag to UNCACHEABLE_DEPENDENT_INJECTED unconditionally.
As result, even if IN-TO-EXISTS injected non-correlated predicates,
the subquery was still treated as correlated.
Solution:
Set the subquery as correlated only if the injected predicate(s) depend
on the outer query.
The patch enables back constant subquery execution during
query optimization after it was disabled during the development
of MWL#89 (cost-based choice of IN-TO-EXISTS vs MATERIALIZATION).
The main idea is that constant subqueries are allowed to be executed
during optimization if their execution is not expensive.
The approach is as follows:
- Constant subqueries are recursively optimized in the beginning of
JOIN::optimize of the outer query. This is done by the new method
JOIN::optimize_constant_subqueries(). This is done so that the cost
of executing these queries can be estimated.
- Optimization of the outer query proceeds normally. During this phase
the optimizer may request execution of non-expensive constant subqueries.
Each place where the optimizer may potentially execute an expensive
expression is guarded with the predicate Item::is_expensive().
- The implementation of Item_subselect::is_expensive has been extended
to use the number of examined rows (estimated by the optimizer) as a
way to determine whether the subquery is expensive or not.
- The new system variable "expensive_subquery_limit" controls how many
examined rows are considered to be not expensive. The default is 100.
In addition, multiple changes were needed to make this solution work
in the light of the changes made by MWL#89. These changes were needed
to fix various crashes and wrong results, and legacy bugs discovered
during development.
- Let fix_semijoin_strategies_for_picked_join_order() set
POSITION::prefix_record_count for POSITION records that it copies from
SJ_MATERIALIZATION_INFO::tables.
(These records do not have prefix_record_count set, because they are optimized
as joins-inside-semijoin-nests, without full advance_sj_state() processing).
Analysis:
The reason for the wrong result is the interaction between constant
optimization (in this case 1-row table) and subquery optimization.
- First the outer query is optimized, and 'make_join_statistics' finds that
table t2 has one row, reads that row, and marks the whole table as constant.
This also means that all fields of t2 are constant.
- Next, we optimize the subquery in the end of the outer 'make_join_statistics'.
The field 'f2' is considered constant, with value '3'. The subquery predicate
is rewritten as the constant TRUE.
- The outer query execution detects early that the whole query result is empty
and calls 'return_zero_rows'. Since the query is with implicit grouping, we
have to produce one row with special values for the aggregates (depending on
each aggregate function), and NULL values for all non-aggregate fields. This
function calls 'no_rows_in_result' to set each aggregate function to the
default value when it aggregates over an empty result, and then calls
'send_data', which in turn evaluates each Item in the SELECT list.
- When evaluation reaches the subquery predicate, it executes the subquery
with field 'f2' having a constant value '3', and the subquery produces the
incorrect result '7'.
Solution:
Implement Item::no_rows_in_result for all subquery predicates. In order to
make this work, it is also needed to make all val_* methods of all subquery
predicates respect the Item_subselect::forced_const flag. Otherwise subqueries
are executed anyways, and override the default value set by no_rows_in_result
with whatever result is produced from the subquery evaluation.
- When doing join optimization, pre-sort the tables so that they mimic the execution
order we've had with 'semijoin=off'.
- That way, we will not get regressions when there are two query plans (the old and the
new) that have indentical costs but different execution times (because of factors that
the optimizer was not able to take into account).
- The problem was with execution strategy for cases where FirstMatch's inner tables
were interleaved with outer-uncorrelated tables.
- I was unable to find any cases where such join orders would be practically useful,
so fixed it by disabling them.
include/mysql_com.h:
remove "shutdown levels" that aren't shutdown levels from mysql_enum_shutdown_level
mysys/my_addr_resolve.c:
my_snprintf in 5.5 (but not in 5.3) supports %p
sql/item_func.cc:
use a method (that exists only in 5.5) instead of directly accessing a member
sql/item_subselect.cc:
use a method (that exists only in 5.5) instead of directly accessing a member
sql/opt_subselect.cc:
use a method (that exists only in 5.5) instead of directly accessing a member
sql/sql_select.cc:
use a method (that exists only in 5.5) instead of directly accessing a member
- Fix equality propagation to work with SJM nests and OR clauses (full descirption of problem and
solution in the comment in the patch)
(The second commit with post-review fixes)
- The problem was that
= we've picked a LooseScan that used full index scan (tab->type==JT_ALL) on certain index.
= there was also a quick select (tab->quick!=NULL), that used other indexes.
= some old code assumes that (tab->type==JT_ALL && tab->quick) -> means that the
quick select should be used, which is not true.
Fixed by discarding the quick select as soon as we know we're using LooseScan
without using the quick select.
- Remove all references of MAX_TABLES from JOIN struct and make these dynamic
- Updated Join_plan_state to allocate just as many elements as it's needed
sql/opt_subselect.cc:
Optimized version of Join_plan_state
sql/sql_select.cc:
Set join->positions and join->best_positions dynamicly
Don't call update_virtual_fields() if table->vfield is not set.
sql/sql_select.h:
Remove all references of MAX_TABLES from JOIN struct and Join_plan_state and make these dynamic
- Avoid needless load/stores in my_hash_sort_simple due to possible aliasing
- Avoid expensive Join_plan_state constructor in choose_subquery_plan when no subquery
- Avoid calling update_virtual_fields for every row when no virtual fields.
- The problem was that convert_subq_to_jtbm() attached the semi-join
TABLE_LIST object into the wrong list: they used to attach it to the
end of parent_lex->leaf_tables.head()->next_local->...->next_local.
This was apparently inccorect, as one can construct an example where
JTBM nest is attached to a table that is inside some mergeable VIEW, which
breaks (causes crash) for name resolution on the subsequent statement
re-execution.
- Solution: Attach to the "right" list. The "wording" was copied from
st_select_lex::handle_derived.
The result of materialization of the right part of an IN subquery predicate
is placed into a temporary table. Each row of the materialized table is
distinct. A unique key over all fields of the temporary table is defined and
created. It allows to perform key look-ups into the table.
The table created for a materialized subquery can be accessed by key as
any other table. The function best_access-path search for the best access
to join a table to a given partial join. With some where conditions this
function considers a possibility of a ref_or_null access. If such access
employs the unique key on the temporary table then when estimating
the cost this access the function tries to use the array rec_per_key. Yet,
such array is not built for this unique key. This causes a crash of the server.
Rows returned by the subquery that contain nulls don't have to be placed
into temporary table, as they cannot be match any row produced by the
left part of the subquery predicate. So all fields of the temporary table
can be defined as non-nullable. In this case any ref_or_null access
to the temporary table does not make any sense and it does not make sense
to estimate such an access.
The fix makes sure that the temporary table for a materialized IN subquery
is defined with columns that are all non-nullable. The also ensures that
any row with nulls returned by the subquery is not placed into the
temporary table.
- If LooseScan is used with quick select, require that quick select produces
data in key order (this disables use of MRR, which can return data in arbitrary order).
- Disable use of join cache when we're using FirstMatch strategy, and the join
order is such that subquery's inner tables are interleaved with outer. Join
buffering code is incapable of handling such join orders.
- The testcase requires use of @@debug_optimizer_prefer_join_prefix to hit the bug,
but I'm pushing it anyway (including the mention of the variable in .test file),
so that it can be found and enabled when/if we get something comparable in the
main tree.
The problem was that LooseScan execution code assumed that tab->key holds
the index used for looseScan. This is only true when range or full index
scan are used. In case of ref access, the index is in tab->ref.key (and
tab->index==0 which explains how LooseScan passed tests with ref access: they
used one index)
Fixed by setting/using loosescan_key, which always the correct index#.
fixed several defects in the greedy optimization:
1) The greedy optimizer calculated the 'compare-cost' (CPU-cost)
for iterating over the partial plan result at each level in
the query plan as 'record_count / (double) TIME_FOR_COMPARE'
This cost was only used locally for 'best' calculation at each
level, and *not* accumulated into the total cost for the query plan.
This fix added the 'CPU-cost' of processing 'current_record_count'
records at each level to 'current_read_time' *before* it is used as
'accumulated cost' argument to recursive
best_extension_by_limited_search() calls. This ensured that the
cost of a huge join-fanout early in the QEP was correctly
reflected in the cost of the final QEP.
To get identical cost for a 'best' optimized query and a
straight_join with the same join order, the same change was also
applied to optimize_straight_join() and get_partial_join_cost()
2) Furthermore to get equal cost for 'best' optimized query and a
straight_join the new code substrcated the same '0.001' in
optimize_straight_join() as it had been already done in
best_extension_by_limited_search()
3) When best_extension_by_limited_search() aggregated the 'best' plan a
plan was 'best' by the check :
'if ((search_depth == 1) || (current_read_time < join->best_read))'
The term '(search_depth == 1' incorrectly caused a new best plan to be
collected whenever the specified 'search_depth' was reached - even if
this partial query plan was more expensive than what we had already
found.
