It is triple bug with one test suite:
1. Incorrect outer table detection
2. Incorrect leaf table processing for multi-update (should be full like for usual updates and inserts)
3. ON condition fix_fields() fould be called for all tables of the query.
Both bugs are caused by the same problem: the function optimize_cond() should
update the value of *cond_equal rather than the value of join->cond_equal,
because it is called not only for the WHERE condition, but for the HAVING
condition as well.
- With big_tables=ON, materialized table will use Aria (or MyISAM) SE, which
allows prefix key reads. However, the temp.table has rec_per_key=NULL which
causes the optimizer to crash when attempting to read index statistics for a
prefix index read.
- Fixed by providing a rec_per_key array with zeros (i.e. "no statistics data")
The calls of the function remove_eq_conds() may change the and/or structure
of the where conditions. So JOIN::equal_cond should be updated for non-recursive
calls of remove_eq_conds().
update_used_tables for the the where condition to update cached
indicators of constant subexpressions. It should be done before further
possible simplification of the where condition.
This change caused simplification of the executed where conditions
in many test cases.
- The problem was that JOIN::prepare() tried to set TABLE::maybe_null
for a table in join. Non-merged semi-join tables 1) are present as
join's base tables on second EXECUTE, but 2) do not yet have a TABLE
object.
Worked around the problem by putting mixed_implicit_grouping into JOIN
object, and then passing it to JTBM tables in setup_jtbm_semi_joins().
The field JOIN::select_lex->where should be updated after the call
of remove_eq_conds() in the function make_join_statistics(). This
matters for subselects.
Old code in create_tmp_table(), that created an extra one-byte field (recinfo)
before every NULL-able grouping field (Field) in the tmp table, did not actually work.
Because the matching code in end_update(), that was supposed to update this byte,
was using a wrong offset, updating the first byte of the Field, not a byte before it.
Normally this wasn't an issue, because the Field value (written later in end_update)
was overwriting this byte anyway. But in this bug the Field was Field_null, with zero
length, so end_update() was overwriting the first byte of the following field.
And the following field was not-nullable constant, which was stored only once in
create_tmp_table and never updated later.
Fixed by removing the code that didn't do any useful work anyway.
MDEV-5337: Wrong result in mariadb 5.5.32 with ORDER BY + LIMIT when
index_condition_pushdown=on
- in test_if_skip_sort_order(), correct the condition under which
we have the code that restores the previously pushed index condition.
Main fix was to not cache derivied tables as they may be temporary tables that are deleted before the next query.
This was a bit tricky as Item_field::fix_fields depended on cached_tables to be set to resolve some columns.
mysql-test/r/sp-bugs.result:
Added test case
mysql-test/t/sp-bugs.test:
Added test case
sql/item.cc:
Fixed fix_outer_field to handle case where found field did not have in cached_table
Idea is that if cached_table is not avaliable, use from_field->table->pos_in_table_list instead
sql/records.cc:
Also accept INTERNAL_TMP_TABLE for memmap
sql/sql_base.cc:
More DBUG_PRINT
Fixed that setup_natural_join_row_types() is not run twice.
Original code modified context->first_name_resolution_table also for second executions.
This was wrong as this could give wrong results if some joins had been optimized away between calls.
sql/sql_derived.cc:
Mark derived tables as internal temporary tables (INTERNAL_TMP_TABLE), not as NON_TRANSACTIONAL_TMP_TABLE.
This is more correct as the tables are not visible by the end user.
sql/sql_insert.cc:
Reset pos_in_table_list before calling fix_fields.
One of the consequences of the change of not caching all generated tables in Item_ident is that
pos_in_table_list needs to be correct in calls to fix_fields.
sql/sql_lex.cc:
More DBUG_PRINT
sql/sql_parse.cc:
Don't cache derivied tables as they may be temporary tables that are deleted before the next query
sql/sql_select.cc:
Reset table_vector. This was required as some code checked the vector to see if temporary tables had already been created.
sql/table.cc:
Mark tables with field translations as cacheable (as these will not disapper between stmt executions.
THD::thd->activate_stmt_arena_if_needed() should be used to temporary activating statement arena instead of direct usage of THD::set_n_backup_active_arena() because possible such scenario:
1) func1 saves current arena and activates copy1 of statement arena
2) func2 saves copy1 of statement arena setup by func1 and activates copy2
3) some changes made for copy 2
4) func2 stores changed copy2 back to statenet arena and activates copy1
5) func1 store unchanged copy1 back to statemnt arena (rewrite changed copy 2 so changes become lost) and activates arena which was before.
THD::thd->activate_stmt_arena_if_needed() should be used to temporary activating statement arena instead of direct usage of THD::set_n_backup_active_arena() because possible such scenario:
1) func1 saves current arena and activates copy1 of statement arena
2) func2 saves copy1 of statement arena setup by func1 and activates copy2
3) some changes made for copy 2
4) func2 stores changed copy2 back to statenet arena and activates copy1
5) func1 store unchanged copy1 back to statemnt arena (rewrite changed copy 2 so changes become lost) and activates arena which was before.
- When a JOIN has both "optimization tabs" (JOIN_TABs used to
read the base tables and do the join operation) and also
has "execution tabs" (a JOIN_TAB that is to produce result set
that is sent to the client), do not forget to call JOIN_TAB::cleanup()
for the execution JOIN_TAB.
This only happend when using an ORDER BY on a primary key part, where all other key parts where constant.
Remove of duplicated expressions in ORDER BY (as the old code did this in some strange cases)
mysql-test/r/group_by.result:
Fixed results to take into account that duplicate order by parts are now deleted
mysql-test/r/group_by_innodb.result:
Ensure extended keys are on
mysql-test/r/innodb_ext_key.result:
More tests
mysql-test/r/order_by.result:
More tests
mysql-test/t/group_by.test:
Fixed results to take into account that duplicate order by parts are now deleted
mysql-test/t/group_by_innodb.test:
Ensure extended keys are on
mysql-test/t/innodb_ext_key.test:
More tests
mysql-test/t/order_by.test:
More tests
sql/sql_select.cc:
Fixed bug where we looked at extended key parts when we shouldn't
Remove of duplicated expressions in ORDER BY
sql/table.cc:
Indentation fixes
The earlier pushed fix for the bug was incomplete. It did not remove
the main cause of the problem: the function remove_eq_conds()
removed always true multiple equalities from any conjunct, but did not
adjust the list of them stored in Item_cond_and::cond_equal.current_level.
Simplified the test case for the bug and moved it to another test file.
The fix triggered changes in EXPLAIN EXTENDED for some queries.
- Don't pull out a table out of a semi-join if it is on the inner side of an outer join.
- Make join->sort_by_table= get_sort_by_table(...) call after const table detection
is done. That way, the value of join->sort_by_table will match the actual execution.
Which will allow the code in setup_semijoin_dups_elimination() (search for
"Make sure that possible sorting of rows from the head table is not to be employed."
to see that "Using filesort" is going to be used together with Duplicate Elimination (
and change it to Using temporary + Using filesort)
- make_join_readinfo() has the code that forces use of "Using temporary;
Using filesort" when join buffering is in use.
That code didn't handle all cases, in particular it didn't hande the case
where ORDER BY originally has tables from multiple columns, but the
optimizer eventually figures out that doing filesort() on one table
will be sufficient. Adjusted the code to handle that case.
Apply fix suggested by Igor:
- When eliminate_item_equal() generates pair-wise equalities from a
multi-equality, do generate a "bridge" equality between the first
field inside SJM nest and the field that's first in the overall multi-equality.
Analysis:
st_select_lex_unit::prepare() computes can_skip_order_by as TRUE.
As a result join->prepare() gets called with order == NULL, and
doesn't do name resolution for the inner ORDER clause. Due to this
the prepare phase doesn't detect that the query references non-exiting
function and field.
Later join->optimize() calls update_used_tables() for a non-resolved
Item_field, which understandably has no Field object. This call results
in a crash.
Solution:
Resolve unnecessary ORDER BY clauses to detect if they reference non-exising
objects. Then remove such clauses from the JOIN object.
