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.
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.
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).
"the server side preparedStatement error for LIMIT placeholder",
which moves all uses of LIMIT clause from PREPARE to OPTIMIZE
and later steps.
After-review fixes.
myisam_max_extra_sort_file_size is depricated
Ensure that myisam_data_pointer_size is honoured when creating new MyISAM files
Changed default value of myisam_data_pointer_size from 4 to 6 to get rid of 'table-is-full' errors
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
