Re: Columns correlation and adaptive query optimization
Konstantin Knizhnik <k.knizhnik@postgrespro.ru>
From: Konstantin Knizhnik <k.knizhnik@postgrespro.ru>
To: legrand legrand <legrand_legrand@hotmail.com>,
pgsql-hackers@postgresql.org
Date: 2019-10-15T07:46:38Z
Lists: pgsql-hackers
Attachments
- join_selectivity.patch (text/x-patch) patch
- auto_explain_create_statistics.patch (text/x-patch) patch
On 15.10.2019 1:20, legrand legrand wrote: > Hello Konstantin, > > What you have proposed regarding join_selectivity and multicolumn statistics > is a very good new ! > > Regarding your auto_explain modification, maybe an "advisor" mode would also > be helpfull (with auto_explain_add_statistics_threshold=-1 for exemple). > This would allow to track which missing statistic should be tested (manually > or in an other environment). > > In my point of view this advice should be an option of the EXPLAIN command, > that should also permit > auto_explain module to propose "learning" phase. Thank you for good suggestion. Advisor mode is really good idea. I have added "auto_explain.suggest_only" GUC. When it is switched on, suggested CREATE STATISTICS statement is just printed in log but not actually created. -- Konstantin Knizhnik Postgres Professional: http://www.postgrespro.com The Russian Postgres Company