Re: using extended statistics to improve join estimates

On Wed, Jan 19, 2022 at 06:18:09PM +0800, Julien Rouhaud wrote:
> On Tue, Jan 04, 2022 at 03:55:50PM -0800, Andres Freund wrote:
> > On 2022-01-01 18:21:06 +0100, Tomas Vondra wrote:
> > > Here's an updated patch, rebased and fixing a couple typos reported by
> > > Justin Pryzby directly.
> >
> > FWIW, cfbot reports a few compiler warnings:
>
> Also the patch doesn't apply anymore:
>
> http://cfbot.cputube.org/patch_36_3055.log
> === Applying patches on top of PostgreSQL commit ID 74527c3e022d3ace648340b79a6ddec3419f6732 ===
> === applying patch ./0001-Estimate-joins-using-extended-statistics-20220101.patch
> patching file src/backend/optimizer/path/clausesel.c
> patching file src/backend/statistics/extended_stats.c
> Hunk #1 FAILED at 30.
> Hunk #2 succeeded at 102 (offset 1 line).
> Hunk #3 succeeded at 2619 (offset 9 lines).
> 1 out of 3 hunks FAILED -- saving rejects to file src/backend/statistics/extended_stats.c.rej

Rebased over 269b532ae and muted compiler warnings.

Tomas - is this patch viable for pg15 , or should move to the next CF ?

In case it's useful, I ran this on cirrus with my branch for code coverage.
https://cirrus-ci.com/task/5816731397521408
https://api.cirrus-ci.com/v1/artifact/task/5816731397521408/coverage/coverage/00-index.html

statext_find_matching_mcv() has poor coverage.
statext_clauselist_join_selectivity() has poor coverage for the "stats2" case.

In mcv.c: mcv_combine_extended() and mcv_combine_simple() have poor coverage
for the "else if" cases (does it matter?)

Not related to this patch:
build_attnums_array() isn't being hit.

Same at statext_is_compatible_clause_internal()
1538 0 : *exprs = lappend(*exprs, clause);

statext_mcv_[de]serialize() aren't being hit for cstrings.

--
Justin