GitLab

This problem could be observed for queries with nested outer joins
for which the not_exist optimization were applicable. 
The problem was caused by the code of the patch for bug #49322
that erroneously forced the return to the previous nested loop
level when the join algorithm successfully builds a partial record
for an embedded outer to which the not_exist optimization could be
applied.
Actually the immediate return to the previous nested loops level
is correct only if this partial record is rejected by a predicate
pushed down to one of the inner tables of this outer join. Otherwise
attempts to find extensions of this record must be made.

the optimizer switch.

Do not make substitution of a single-row table if it is an inner
table of an outer join with on expression containing an expensive
subquery.

An aggregating query over an empty set of a join of two tables
with a rejecting HAVING clause erroneously could return a row.
It could happen in the cases when the optimizer made a conclusion
that the aggregating set was empty.
Wrong results were produced because the server missed initial
setting for aggregation functions in the mentioned cases.
      

BUG##806524: Assertion `join->best_read < 1.7976931348623157e+308 with table_elimination=on and derived_merge=on 
reset_nj_counters() used to rely on the fact that join nests have 
table->table==NULL. This ceased to be true wit new derived table
optimizations. Use test for table->nested_join!=NULL instead.

The problem was that optimizer removes some outer references (it they are
constant for example) and the list of outer items built during prepare phase is
not actual during execution phase when we need it as the cache parameters.
First solution was use pointer on pointer on outer reference Item and
initialize temporary table on demand. This solved most problem except case
when optimiser also reduce Item which contains outer references ('OR' in
this bug test suite).

The solution is to build the list of outer reference items on execution
phase (after optimization) on demand (just before temporary table creation)
by walking Item tree and finding outer references among Item_ident
(Item_field/Item_ref) and Item_sum items.

Removed depends_on list (because it is not neede any mnore for the cache, in the place where it was used it replaced with upper_refs).

Added processor (collect_outer_ref_processor) and get_cache_parameters() methods to collect outer references (or other expression parameters in future).

mysql-test/r/subselect_cache.result:
  A new test added.
mysql-test/r/subselect_scache.result:
  Changes in creating the cache and its paremeters order or adding arguments of aggregate function (which is a parameter also, but this has no influence on the result).
mysql-test/t/subselect_cache.test:
  Added a new test.
sql/item.cc:
  depends_on removed.
  
  Added processor (collect_outer_ref_processor) and get_cache_parameters() methods to collect outer references.
  
  Item_cache_wrapper collect parameters befor initialization of its cache.
sql/item.h:
  depends_on removed.
  
  Added processor (collect_outer_ref_processor) and get_cache_parameters() methods to collect outer references.
sql/item_cmpfunc.cc:
  depends_on removed.
  
  Added processor (collect_outer_ref_processor) to collect outer references.
sql/item_cmpfunc.h:
  Added processor (collect_outer_ref_processor) to collect outer references.
sql/item_subselect.cc:
  depends_on removed.
  Added processor get_cache_parameters() method to collect outer references.
sql/item_subselect.h:
  depends_on removed.
  Added processor get_cache_parameters() method to collect outer references.
sql/item_sum.cc:
  Added processor (collect_outer_ref_processor) method to collect outer references.
sql/item_sum.h:
  Added processor (collect_outer_ref_processor) and get_cache_parameters() methods to collect outer references.
sql/opt_range.cc:
  depends_on removed.
sql/sql_base.cc:
  depends_on removed.
sql/sql_class.h:
  New iterator added.
sql/sql_expression_cache.cc:
  Build of list of items resolved in outer query done just before creating expression cache on the first execution of the subquery which removes influence of optimizer removing items (all optimization already done).
sql/sql_expression_cache.h:
  Build of list of items resolved in outer query done just before creating expression cache on the first execution of the subquery which removes influence of optimizer removing items (all optimization already done).
sql/sql_lex.cc:
  depends_on removed.
sql/sql_lex.h:
  depends_on removed.
sql/sql_list.h:
  Added add_unique method to add only unique elements to the list.
sql/sql_select.cc:
  Support of new Item list added.
sql/sql_select.h:
  Support of new Item list added.

This bug could lead to wrong result sets for a query over a
materialized derived table or view accessed by a multi-component
key.
It happened because the function get_next_field_for_derived_key
was supposed to update its argument, and it did not do it.

Also:
1. simplified the code of the function mysql_derived_merge_for_insert.
2. moved merge of views/dt for multi-update/delete to the prepare stage.
3. the list of the references to the candidates for semi-join now is
   allocated in the statement memory.

Analysis:

For some of the re-executions of the correlated subquery the
where clause is false. In these cases the execution of the
subquery detects that it must generate a NULL row because of
implicit grouping. In this case the subquery execution reaches
the following code in do_select():

        while ((table= li++))
          mark_as_null_row(table->table);

This code marks all rows in the table as complete NULL rows.
In the example, when evaluating the field t2.f10 for the second
row, all bits of Field::null_ptr[0] are set by the previous call
to mark_as_null_row(). Then the call to Field::is_null()
returns true, resulting in a NULL for the MAX function.

Thus the lines above are not suitable for subquery re-execution
because mark_as_null_row() changes the NULL bits of each table
field, and there is no logic to restore these fields.

Solution:

The call to mark_as_null_row() was added by the fix for bug
lp:613029. Therefore removing the fix for lp:613029 corrects
this wrong result. A...

The bitmap of used tables must be evaluated for the select list of every
materialized derived table / view and saved in a dedicated field.
This is also applied to materialized subqueries.

The value of THD::used tables should be re-evaluated after merges
of views and derived tables into the main query. 
Now it's done in the function SELECT_LEX::update_used_tables.
The re-evaluation of the 'used_table' bitmaps for the items
in HAVING, GROUP BY and ORDER BY clauses has been added as well.
 

The bug was caused by an incorrect code of the function
Item_direct_view_ref::replace_equal_field introduced in the
patch for bugs 717577, 724942. The function erroneously
returned the wrapped field instead of the Item_direct_view_ref
object itself in the cases when no replacement happened.

The bug masked two other minor bugs that could result in not
quite correct output of the EXPLAIN command for some queries.
They were fixed in the patch as well.
 

The function generate_derived_keys_for_table incorrectly handled
the cases when a materialized view or derived table could be accessed
by different keys on the same fields if these keys depended on the
same tables.
  

Analysis:
This bug consists of two related problems that are
result of too early evaluation of single-row subqueries
during the optimization phase of the outer query.

Several optimizer code paths try to evaluate single-row
subqueries in order to produce a constant and use that
constant for further optimzation.

When the execution of the subquery peforms destructive
changes to the representation of the subquery, and these
changes are not anticipated by the subsequent optimization
phases of the outer query, we tipically get a crash or
failed assert.

Specifically, in this bug the inner-most suqbuery with
DISTINCT triggers a substitution of the original JOIN
object by a single-table JOIN object with a temp table
needed to perform the DISTINCT operation (created by
JOIN::make_simple_join).

This substitution breaks EXPLAIN because:
a) in the first example JOIN::cleanup no longer can
reach the original table of the innermost subquery, and
close all indexes, and
b) in this second test query, EXPLAIN attempts to print
the name of the internal temp table, and crashes because
the temp table has no name (NULL pointer instead).

Solution:
a) fully disable subquery evaluation during optimization
in all cases - both for constant propagation and range
optimization, and
b) change JOIN::join_free() to perform cleanup irrespective
of EXPLAIN or not.

If no index is used to access a materialized derived table or view
then the value of TABLE_REF::key for this table must be (-1).

- JOIN::prepare would have set JOIN::table_count to incorrect value (bad merge of MWL 106)
- optimize_keyuse() would use table-bit as table number
  (the change in optimize_keyuse is also the reason for query plan changes. Not 
   expected to have much effect because only handles cases of no index statistics)
- st_select_lex::register_dependency_item() ignored the fact that some of the 
  selects on the dependency paths could have been merged to their parents (because they 
  were mergeable VIEWs)
- Undo the incorrect fix in Item_subselect::recalc_used_tables(): do not call 
  fix_after_pullout() for Item_subselect::Ref_to_outside members. 

If the expression for a derived table contained a clause LIMIT 0
SELECT from such derived table incorrectly returned a non-empty set.

Fixed by ensuring JOIN::do_send_rows to be updated after the call
of st_select_lex_unit::set_limit that sets the value of 
JOIN::unit->select_limit_cnt.

Due to this bug in the function generate_derived_keys_for_table some
key definitions to access materialized derived tables or materialized
views were constructed with invalid info for their key parts.
This could make the server crash when it optimized queries using 
materialized derived tables or materialized views. 

- The crash was because a NOT NULL table column inside the subquery was considered NULLable
  because the code thought it was on the inner side of an outer join nest.
- Fixed by making correct distinction between tables inside outer join nests and inside semi-join nests.

Split status variable Rows_read to Rows_read and Rows_tmp_read so that one can see how much real data is read.
Same was done with with Handler_update and Handler_write.
Fixed bug in MEMORY tables where some variables was counted twice.
Added new internal handler call 'ha_close()' to have one place to gather statistics.
Fixed bug where thd->open_options was set to wrong value when doing admin_recreate_table()


mysql-test/r/status.result:
  Updated test results and added new tests
mysql-test/r/status_user.result:
  Udated test results
mysql-test/t/status.test:
  Added new test for temporary table status variables
sql/ha_partition.cc:
  Changed to call ha_close() instead of close()
sql/handler.cc:
  Added internal_tmp_table variable for easy checking of temporary tables.
  Added new internal handler call 'ha_close()' to have one place to gather statistics.
  Gather statistics for internal temporary tables.
sql/handler.h:
  Added handler variables internal_tmp_table, rows_tmp_read.
  Split function update_index_statistics() to two.
  Added ha_update_tmp_row() for faster tmp table handling with more statistics.
sql/item_sum.cc:
  ha_write_row() -> ha_write_tmp_row()
sql/multi_range_read.cc:
  close() -> ha_close()
sql/mysqld.cc:
  New status variables: Rows_tmp_read, Handler_tmp_update and Handler_tmp_write
sql/opt_range.cc:
  close() -> ha_close()
sql/sql_base.cc:
  close() -> ha_close()
sql/sql_class.cc:
  Added handling of rows_tmp_read
sql/sql_class.h:
  Added new satistics variables.
  rows_read++  ->  update_rows_read() to be able to correctly count reads to internal temp tables.
  Added handler::ha_update_tmp_row()
sql/sql_connect.cc:
  Added comment
sql/sql_expression_cache.cc:
  ha_write_row() -> ha_write_tmp_row()
sql/sql_select.cc:
  close() -> ha_close()
  ha_update_row() -> ha_update_tmp_row()
sql/sql_show.cc:
  ha_write_row() -> ha_write_tmp_row()
sql/sql_table.cc:
  Fixed bug where thd->open_options was set to wrong value when doing admin_recreate_table()
sql/sql_union.cc:
  ha_write_row() -> ha_write_tmp_row()
sql/sql_update.cc:
  ha_write_row() -> ha_write_tmp_row()
sql/table.cc:
  close() -> ha_close()
storage/heap/ha_heap.cc:
  Removed double counting of statistic variables.
  close -> ha_close() to get tmp table statistics.
storage/maria/ha_maria.cc:
  close -> ha_close() to get tmp table statistics.

Ensured valid calculations of the estimates stored in JOIN_TAB::used_fieldlength.

The following were missing in the patch for mwl106:
- KEY_PART_INFO::fieldnr were not set for generated keys to access
  tmp tables storing the rows of materialized derived tables/views
- TABLE_SHARE::column_bitmap_size was not set for tmp tables storing
  the rows of materialized derived tables/views.
These could cause crashes or memory overwrite.

If a view/derived table is non-mergeable then the definition of the tmp table
to store the rows for it is created at the prepare stage. In this case if the
view definition uses outer joins and a view column belongs to an inner table
of one of them then the column should be considered as nullable independently
on nullability of the underlying column. If the underlying column happens to be
defined as non-nullable then the function create_tmp_field_from_item rather 
than the function create_tmp_field_from_field should be employed to create
the definition of the interesting column in the tmp table.
 

mysql-test/r/join.result:
  Test case for LP:798597
mysql-test/t/join.test:
  Test case for LP:798597
sql/sql_select.cc:
  In simplify_joins we reset table->maybe_null for outer join tables that can't ever be NULL.
  This caused a conflict between the previously calculated items and the group_buffer against the fields
  in the temporary table that are created as not null thanks to the optimization.
  The fix is to correct the group by items to also be not_null so that they match the used fields and keys.

The function setup_tables should handle table_list elements for
semijoin materialized tables in a special way when executing
a prepared statement for the second time.

- evaluate_null_complemented_join_record() should perform FirstMatch checks.

- Do make the DuplicateWeedout check for outer joins.

The patch for bugs 717577 and 724942 has missed to make adjustments for the
call item_equal->add_const(const_item, orig_field_item) in the function
check_simple_equality that builds multiple equality for a field and a constant.
As a result, when this field happens to be a view field and the corresponding
Item_field object F is wrapped in an Item_direct_view_ref object R the object
F is placed in the multiple equality instead of the object R.
A substitution of an equal item for F potentially can cause very serious
problems and in some cases can lead to crashes of the server.

- Make make_outerjoin_info() correctly process semi-join nests
- Make make_join_select() attach conditions to the right places.

- Added regression test with queries over the WORLD database.
- Discovered and fixed several bugs in the related cost calculation
  functionality both in the semijoin and non-semijon subquery code.
- Added DBUG printing of the cost variables used to decide between
  IN-EXISTS and MATERIALIZATION.

(Changeset sp1r-igor@olga.mysql.com-20070526173301-38848 moved this loop from one place 
to another, then the merge of sp1r-gshchepa/uchum@gleb.loc-20070527192244-26330  have 
kept both copies).

- move attempt to evaluate join->exec_const_cond() out of the "Extract constant part of each ON expression" loop
  (it got there by mistake when merging).

- Handle the case when the subquery's join is degenerate and so has zero tables.

from an ad hoc set of limitations
to a correct rule

Fixed reference to not initialized memory detected by valgrind

sql/sql_select.cc:
  A bit better fix for tmp-table problem:
  Use only dynamic_record format for group by and distinct.
storage/maria/ma_create.c:
  DYNAMIC_RECORD format doesn't pack VARCHAR fields.
  This change fixes a non-fatal uninitialized memory copy.