On 01/31/2013 04:41 AM, Timour Katchaounov wrote:

Igor,

Let me know if I can push the patch below. It fixes the bug according to our last discussion. If the number of group_list elements was known, I could add a test already in the beginning of the function, however such a counter doesn't exist, and I didn't want to complicate the code just for this special case. That's why the test is done where it is now.

Timour

------------------------------------------------------------ revno: 3622 revision-id: timour@askmonty.org-20130131120040-oan9zei8ojyco178 parent: sergii@pisem.net-20130128081223-mp9rsd3t9soz8lly fixes bug(s): https://mariadb.atlassian.net/browse/MDEV-641 committer: timour@askmonty.org branch nick: 5.3-md641 timestamp: Thu 2013-01-31 14:00:40 +0200 message: Analysis: Range analysis discoveres that the query can be executed via loose index scan for GROUP BY. Later, GROUP BY analysis fails to confirm that the GROUP operation can be computed via an index because there is no logic to handle duplicate field references in the GROUP clause. As a result the optimizer the optimizer produces an inconsistent plan. It constructs a temporary table, but on the other hand the group fields are not set to point there.

Solution: Make loose scan analysis work in sync with order by analysis. In the case of duplicate columns loose scan will not be applicable. This limitation will be lifted in 10.0 by removing duplicate columns.