statements
Currently the optimizer evaluates loose index scan only for top-level SELECT
statements
Extend loose index scan applicability by :
- Test the applicability of loose scan for each sub-select, instead of the
whole query. This change enables loose index scan for sub-queries.
- allow non-select statements with SELECT parts (like, e.g.
CREATE TABLE .. SELECT ...) to use loose index scan.
mysql-test/r/group_min_max.result:
Bug#24156: Loose index scan not used with CREATE TABLE ...SELECT and similar
statements
- test case
mysql-test/t/group_min_max.test:
Bug#24156: Loose index scan not used with CREATE TABLE ...SELECT and similar
statements
- test case
sql/opt_range.cc:
Bug#24156: Loose index scan not used with CREATE TABLE ...SELECT and similar
statements
- loose index scan will be tried over the current subselect
(lex->current_select) instead of the whole query (lex->select_lex).
- allow non-select statements with SELECT parts (like, e.g.
CREATE TABLE .. SELECT ...) to use loose index scan.
When using index for group by and range access the server isolates
a set of ranges based on the conditions over the key parts of the
index used. Then it uses only the ranges over the GROUP BY fields to
jump from one group to another. Since the GROUP BY fields may form a
prefix over the index, we may use only a prefix of the ranges produced
by the range optimizer.
Each range contains a notion on whether it includes its border values.
The problem is that when using a range prefix, the last range is open
because it assumes that there is a range on the next keypart. Thus when
we use a prefix range as it is, it excludes all border values.
The solution is when ignoring the suffix of the range conditions
(to jump over the GROUP BY prefix only) the server must change the
remaining intervals so they always contain their borders, e.g.
if the whole range was :
(1,-inf) <= (<group_by_col>,<min_max_arg_col>) < (1, 3) we must make
(1) <= (<group_by_col>) <= (1) because (a,b) < (c1,c2) means :
a < c1 OR (a = c1 AND b < c2).
mysql-test/r/group_min_max.result:
Bug #22342: No results returned for query using max and group by
- test case
mysql-test/t/group_min_max.test:
Bug #22342: No results returned for query using max and group by
- test case
sql/opt_range.cc:
Bug #22342: No results returned for query using max and group by
- open the intervals for prefix select when there are more conditions
than used for the prefix search.
sql/opt_range.h:
Bug #22342: No results returned for query using max and group by
- open the intervals for prefix select when there are more conditions
than used for the prefix search.
The problem was that store_top_level_join_columns() incorrectly assumed
that the left/right neighbor of a nested join table reference can be only
at the same level in the join tree.
The fix checks if the current nested join table reference has no immediate
left/right neighbor, and if so chooses the left/right neighbors of the
nearest upper level, where these references are != NULL.
mysql-test/r/group_min_max.result:
Test for BUG#21007.
mysql-test/t/group_min_max.test:
Test for BUG#21007.
sql/sql_base.cc:
After computing and materializing the columns of all NATURAL joins in a FROM clause,
the procedure store_top_level_join_columns() has to change the current natural join
into a leaf table reference for name resolution. For this it needs to make the left
neighbor point to the natural join table reference, and the natural join itself point
to its left neighbor.
This fix correctly determines the left/right neighbors of a table reference, even if
the neghbors are at higher levels in the nested join tree. The rule is that if a table
reference has no immediate left/right neighbors, we recursively pick the left/right
neighbor of the level(s) above.
Moved the InnoDB related tests to innodb_mysql
mysql-test/r/group_min_max.result:
Moved innodb related tests out of group_min_max
mysql-test/r/innodb_mysql.result:
Moved innodb related tests out of group_min_max
mysql-test/t/group_min_max.test:
Moved innodb related tests out of group_min_max
mysql-test/t/innodb_mysql.test:
Moved innodb related tests out of group_min_max
When converting DISTINCT to GROUP BY where the columns are from the covering
index and they are quoted twice in the SELECT list the optimizer is creating
improper processing sequence. This is because of the fact that the columns
of the covering index are not recognized as such and treated as non-index
columns.
Generally speaking duplicate columns can safely be removed from the GROUP
BY/DISTINCT list because this will not add or remove new rows in the
resulting set. Duplicates can be removed even if they are not consecutive
(as is the case for ORDER BY, where the duplicate columns can be removed
only if they are consecutive).
So we can safely transform "SELECT DISTINCT a,a FROM ... ORDER BY a" to
"SELECT a,a FROM ... GROUP BY a ORDER BY a" instead of
"SELECT a,a FROM .. GROUP BY a,a ORDER BY a". We can even transform
"SELECT DISTINCT a,b,a FROM ... ORDER BY a,b" to
"SELECT a,b,a FROM ... GROUP BY a,b ORDER BY a,b".
The fix to this bug consists of checking for duplicate columns in the SELECT
list when constructing the GROUP BY list in transforming DISTINCT to GROUP
BY and skipping the ones that are already in.
mysql-test/r/distinct.result:
test case for the bug without loose index scan
mysql-test/r/group_min_max.result:
test case for the bug
mysql-test/t/distinct.test:
test case for the bug without loose index scan
mysql-test/t/group_min_max.test:
test case for the bug
sql/sql_select.cc:
duplicates check and removal
into mysql.com:/home/timka/mysql/src/5.0-bug-16710
mysql-test/r/group_min_max.result:
Auto merged
mysql-test/t/group_min_max.test:
Auto merged
sql/item.cc:
Auto merged
sql/item.h:
Auto merged
sql/opt_range.cc:
Auto merged
The bug was due to a missed case in the detection of whether an index
can be used for loose scan. More precisely, the range optimizer chose
to use loose index scan for queries for which the condition(s) over
an index key part could not be pushed to the index together with the
loose scan.
As a result, loose index scan was selecting the first row in the
index with a given GROUP BY prefix, and was applying the WHERE
clause after that, while it should have inspected all rows with
the given prefix, and apply the WHERE clause to all of them.
The fix detects and skips such cases.
mysql-test/r/group_min_max.result:
Added test for BUG#16710.
mysql-test/t/group_min_max.test:
Added test for BUG#16710.
sql/item.cc:
Added new method [Item | Item_field]::find_item_in_field_list_processor.
sql/item.h:
Added new method [Item | Item_field]::find_item_in_field_list_processor.
sql/opt_range.cc:
Handle the case when there is no MIN/MAX aggregate function, and a
keypart of the index being considered, that is after the GROUP BY
prefix, is used in the WHERE clause and the condition where it is
used cannot be pushed to the index.
If this is the case, we rule out this index.
Reproduced in 5.0.16-nt. Tested to work in 5.0.20
mysql-test/r/group_min_max.result:
Test case for BUG#15102 : select distinct returns empty result, select count distinct > 0 (correct)
Reproduced in 5.0.16-nt. Tested to be fixed in 5.0.20
mysql-test/t/group_min_max.test:
Test case for BUG#15102 : select distinct returns empty result, select count distinct > 0 (correct)
Reproduced in 5.0.16-nt. Tested to be fixed in 5.0.20
If check_quick_select returns non-empty range then the function cost_group_min_max
cannot return 0 as an estimate of the number of retrieved records.
Yet the function erroneously returned 0 as the estimate in some situations.
mysql-test/r/group_min_max.result:
Added a test case for bug #16203.
mysql-test/t/group_min_max.test:
Added a test case for bug #16203.
into mysql.com:/home/timka/mysql/src/5.0-bug-14920
sql/sql_class.cc:
Auto merged
sql/sql_class.h:
Auto merged
sql/sql_select.cc:
Auto merged
mysql-test/r/group_min_max.result:
Manually merged test cases.
mysql-test/t/group_min_max.test:
Manually merged test cases.
The cause of the bug was the use of end_write_group instead of end_write
in the case when ORDER BY required a temporary table, which didn't take
into account the fact that loose index scan already computes the result
of MIN/MAX aggregate functions (and performs grouping).
The solution is to call end_write instead of end_write_group and to add
the MIN/MAX functions to the list of regular functions so that their
values are inserted into the temporary table.
mysql-test/r/group_min_max.result:
Test for BUG#14920
mysql-test/t/group_min_max.test:
Test for BUG#14920
sql/sql_class.cc:
Added new member to TMP_TABLE_PARAM.
sql/sql_class.h:
Added new member to TMP_TABLE_PARAM.
sql/sql_select.cc:
Enable result rows generated by loose index scan being written into
a temporary table. The change is necessary because loose index
scan already computes the result of GROUP BY and the MIN/MAX aggregate
functions. This is realized by three changes:
- create_tmp_table allocates space for aggregate functions in the
list of regular functions,
- use end_write instead of end_write group,
- copy the pointers to the MIN/MAX aggregate functions to the list
of regular functions TMP_TABLE_PARAM::items_to_copy.
sql/sql_select.h:
New parameter to create_tmp_table.
Loose index scan using only second part of multipart index was choosen, which
results in creating wrong keys and endless loop.
get_best_group_min_max() now allows loose index scan for distinct only if used
keyparts forms a prefix of the index.
mysql-test/t/group_min_max.test:
Test case for bug #13293 Wrongly used index results in endless loop.
mysql-test/r/group_min_max.result:
Test case for bug #13293 Wrongly used index results in endless loop.
sql/opt_range.cc:
Fix bug #13293 Wrongly used index results in endless loop.
get_best_group_min_max() now allows loose index scan for distinct only if used keyparts forms a prefix of the index.
mysql-test/r/group_min_max.result:
Test for BUG#12672.
mysql-test/t/group_min_max.test:
Test for BUG#12672.
sql/opt_range.cc:
Some storage managers like InnoDB always retrieve the value of
the primary key (if any) with any other key. For such storage
managers the optimizer marks an index as covering if that index
and the primary key contain all query fields. This is good for
reading data from an index, but not for lookup. The patch detects
such cases and tests whether an index is really a covering index
by itself, without the primary key. If yes, then that index can
be used for the index skip algorithm of the GROUP_MIN_MAX
access method.
The problem was that when there was no MIN or MAX function, after finding the
group prefix based on the DISTINCT or GROUP BY attributes we did not search further
for a key in the group that satisfies the equi-join conditions on attributes that
follow the group attributes. Thus we ended up with the wrong rows, and subsequent
calls to select_cond->val_int() in evaluate_join_record() were filtering those
rows. Hence - the query result set was empty.
The problem occured both for GROUP BY queries without MIN/MAX and for queries
with DISTINCT (which were internally executed as GROUP BY queries).
mysql-test/r/group_min_max.result:
Added test result for BUG#11044. Notice that the group by query is
equivalent to the distinct query and both are executed via the same
algorithm.
mysql-test/t/group_min_max.test:
Added test for BUG#11044. Notice that the group by query is
equivalent to the distinct query and both are executed via the
same algorithm.
sql/opt_range.cc:
* Use the extended prefix in QUICK_GROUP_MIN_MAX_SELECT::get_next()
to find keys that satisfy equality conditions in the case when there is
no MIN or MAX function.
* Corrected some method comments.
* Corrected debug printout of cost information.
sql/opt_range.cc:
Fix for bug #9798: Rollup crash with InnoDB in setup_sum_funcs().
Disable mix/max group by optimization for queries with rollup for now.
Note: the actual bug was hidden; if this optimization works with rollup
queries we will have to change some code:
the problem is that we call the join->make_sum_func_list() function against
changed thd->mem_root (see SQL_SELECT::test_quick_select()) which creates some
items for rollup, then we free that mem_root, then we try to reuse (free) those
items.
The problem was in that the code that analyses the applicability of the
QUICK_GROUP_MIN_MAX access method for DISTINC queries assumed that there
are no duplicate column references in the DISTINCT clause, and it added
non-exiting key parts for the duplicate column references.
The solution adds a test to check whether the select list already contained
a field with the same name. If such field was already present, then it was
already decided to use its key part for index access. In this such case we
must skip the duplicate field instead of counting it as a new field.
mysql-test/r/group_min_max.result:
Added test results for BUG#8532.
mysql-test/t/group_min_max.test:
Added tests for BUG#8532.
sql/opt_range.cc:
The problem was in that the code that analyses the applicability of the
QUICK_GROUP_MIN_MAX access method for DISTINC queries assumed that there
are no duplicate column references in the DISTINCT clause, and it added
non-exiting key parts for the duplicate column references.
The solution adds a test to check whether the select list already contained
a field with the same name. If such field was already present, then it was
already decided to use its key part for index access. In this such case we
must skip the duplicate field instead of counting it as a new field.
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
include/myisam.h:
Added const before names
mysql-test/r/group_min_max.result:
Make results repeatable
mysql-test/t/group_min_max.test:
Make results repeatable
sql/field.cc:
Created Field::make_field() and made Field_num::make_field() to call this
Use TABLE_SHARE
Use sql_strmake() instead of sql_memdup() to simplify code
sql/field.h:
Changed table_name to be pointer to table_name. This allows us to change alias for all fields by just changing one pointer.
Use TABLE_SHARE
sql/field_conv.cc:
Use TABLE_SHARE
sql/filesort.cc:
Use TABLE_SHARE
sql/ha_berkeley.cc:
Use TABLE_SHARE
sql/ha_heap.cc:
Use TABLE_SHARE
sql/ha_innodb.cc:
Use TABLE_SHARE
sql/ha_myisam.cc:
Use TABLE_SHARE
sql/ha_myisammrg.cc:
Use TABLE_SHARE
Change some pointer handling to use const char*
sql/ha_ndbcluster.cc:
Use TABLE_SHARE
sql/handler.cc:
Use TABLE_SHARE
sql/item.cc:
Use TABLE_SHARE
sql/item_func.cc:
Use TABLE_SHARE
sql/item_subselect.cc:
Use TABLE_SHARE
sql/item_sum.cc:
Use TABLE_SHARE
sql/key.cc:
Use TABLE_SHARE
sql/lock.cc:
Use TABLE_SHARE
sql/log_event.cc:
real_name -> table_name
sql/mysql_priv.h:
Use TABLE_SHARE
sql/opt_range.cc:
Use TABLE_SHARE
sql/opt_sum.cc:
Use TABLE_SHARE
sql/records.cc:
Use TABLE_SHARE
sql/repl_failsafe.cc:
real_name -> table_name
sql/slave.cc:
Use TABLE_SHARE
sql/sp.cc:
Use TABLE_SHARE
sql/sp_head.cc:
real_name -> table_name
sql/sql_acl.cc:
Use TABLE_SHARE
removed unnecessary assert
fixed indentation
changed some char * -> const char*
sql/sql_acl.h:
changed some char* -> const char*
sql/sql_base.cc:
Use TABLE_SHARE
sql/sql_cache.cc:
Use TABLE_SHARE
sql/sql_class.cc:
Use TABLE_SHARE
sql/sql_db.cc:
real_name -> table_name
sql/sql_delete.cc:
Use TABLE_SHARE
sql/sql_derived.cc:
Use TABLE_SHARE
sql/sql_handler.cc:
Use TABLE_SHARE
sql/sql_help.cc:
Use TABLE_SHARE
sql/sql_insert.cc:
Use TABLE_SHARE
sql/sql_load.cc:
Use TABLE_SHARE
sql/sql_parse.cc:
Use TABLE_SHARE
sql/sql_rename.cc:
real_name -> table_name
sql/sql_select.cc:
Use TABLE_SHARE
table->blob_fields now points to field offsets, not fields
tmp_table->table_name now points to alias name
sql/sql_show.cc:
Use TABLE_SHARE
sql/sql_table.cc:
Use TABLE_SHARE
sql/sql_test.cc:
Use TABLE_SHARE
sql/sql_trigger.cc:
Use TABLE_SHARE
sql/sql_udf.cc:
Use TABLE_SHARE
sql/sql_union.cc:
real_name -> table_name
sql/sql_update.cc:
Use TABLE_SHARE
sql/sql_view.cc:
Use TABLE_SHARE
sql/table.cc:
Split TABLE to TABLE and TABLE_SHARE
Changed blob_field to be field offsets instead of pointer to fields
Only initialize table->s->default_values with default record (not all table->record[#])
Some indentation changes
sql/table.h:
Split TABLE to TABLE and TABLE_SHARE
sql/tztime.cc:
real_name -> table_name
sql/unireg.cc:
Use TABLE_SHARE
sql/unireg.h:
Use TABLE_SHARE
mysql-test/r/group_min_max.result:
Test results for BUG#6303
mysql-test/t/group_min_max.test:
Ucommented tests for BUG#6303
sql/opt_range.cc:
Comment fix
BitKeeper/etc/ignore:
Added stamp-h1.in stamp-h2.in to the ignore list
mysql-test/t/group_min_max.test:
- Added test for BUG#6303
- Added test for MIN/MAX optimizable queries with subqueries
sql/opt_range.cc:
- check_group_min_max_predicates should not be called when there is no MIN/MAX function
- skip queries from GROUP BY with MIN/MAX optimization when there is a subselect in the WHERE clause
"Min/Max Optimization for Queries with Group By Clause"
mysql-test/r/group_min_max.result:
Added new test
mysql-test/t/group_min_max.test:
Added new test & comments
sql/item.h:
Remove processor methods that are no longer used.
sql/item_sum.cc:
Remove processor methods that are no longer used.
sql/item_sum.h:
Remove processor methods that are no longer used.
sql/opt_range.cc:
- the test procedure no longer needs to build lists and to traverse the expression trees
- all usable indexes are tested and the bes one is chosen
- added/edited function comments
sql/sql_select.cc:
Modified make_sum_func_list() so it can be reused in the test for MIN/MAX optimization.
sql/sql_select.h:
Modified make_sum_func_list() so it can be reused in the test for MIN/MAX optimization.
- after-review changes
- merged with the source tree from 204-08-27
mysql-test/r/distinct.result:
Different plans due to group-by optimization.
sql/ha_myisam.cc:
More general interface to key_copy.
sql/handler.cc:
More general interface to key_copy.
sql/item.cc:
New method to collect all Item_field objects. Used by Item::walk.
sql/item.h:
Several methods to collect different kinds of items from expression trees.
Used by Item::walk.
sql/item_sum.cc:
Added helper to collect Item_sum objects.
sql/item_sum.h:
Methods to collect and test Item_sum objects.
sql/key.cc:
More general interface to key_copy and key_restore.
sql/mysql_priv.h:
More general interface to key_copy and key_restore.
sql/opt_range.cc:
Complete implementaion of WL#1724 "Min/Max Optimization for Queries with Group By Clause".
sql/opt_range.h:
Complete implementaion of WL#1724 "Min/Max Optimization for Queries with Group By Clause".
sql/opt_sum.cc:
simple_pred is re-used in opt_range.cc
sql/sql_acl.cc:
More general interface to key_copy and key_restore.
sql/sql_handler.cc:
More general interface to key_copy.
sql/sql_insert.cc:
More general interface to key_copy.
sql/sql_select.cc:
Changes to hook the new QUICK_GROUP_MIN_MAX_SELECT due to two differences from all other
quick selects:
1)
This quick select may be created (and used) even if there is no WHERE clause.
Several places assumed that a QUICK_SELECT is constructed only if there is a WHERE clause,
which had to be changed so that QUICK_GROUP_MIN_MAX can be used.
2)
Unlike all other quick selects, this QUICK_GROUP_MIN_MAX_SELECT operates for GROUP BY queries.
Since for the caller the quick select already produces one result tuple per group, there is
no need to call end_send_group, instead we have to call end_send as for a regular quick select.
sql/sql_select.h:
simple_pred is re-used in opt_range.cc
