Re: Automatic function replanning

Christopher Browne wrote:
> > Not unless you do something that would cause the planner to make
> > different choices next time. (Such as changing the ANALYZE statistics,
> > perhaps.) The TODO item is OK as stated, it's just talking about
> > mechanism and not the things that might trigger the mechanism.
>
> A mechanism might be to bump up the stats stored for pg_autovacuum,
> which would encourage a table to be re-ANALYZEd.
>
> That may not immediately change ANALYZE statistics, but it's
> something...
>
> Even if there is NO feedback mechanism on statistics, if we know the
> plan was pretty bad, it is surely at least *a* feedback to invalidate
> the plan.

Something has to cause the replan to be better, and that is one
possibility.

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073