| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Andreas Seltenreich <andreas(dot)seltenreich(at)credativ(dot)de> |
| Cc: | pgsql-bugs(at)postgresql(dot)org |
| Subject: | Re: Unnecessarily imprecise stats for very small tables leading to bad plans |
| Date: | 2018-06-06 15:28:13 |
| Message-ID: | 9101.1528298893@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-bugs |
Andreas Seltenreich <andreas(dot)seltenreich(at)credativ(dot)de> writes:
> Intuitively, I'd say pg_statistic can represent such a small dataset
> perfectly. But instead, most_common_values only contains one of the two
> distinct values. This seems like a bug to me. I looked at the code for
> why this is the case, but couldn't find an answer right-away. Any
> input?
ANALYZE has a hard-and-fast rule that a value must appear more than once
in the sample in order to possibly be considered an MCV.
Perhaps we could tinker with that if the sample is known to comprise
the entire table, but it feels a bit ad-hoc.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Frits Jalvingh | 2018-06-06 16:19:27 | Re: BUG #15225: [XX000] ERROR: invalid DSA memory alloc request size 1073741824 / Where: parallel worker |
| Previous Message | Tom Lane | 2018-06-06 14:48:09 | Re: psql crashes found when executing slash commands |