Re: Use extended statistics to estimate (Var op Var) clauses

On Sun, 13 Jun 2021 at 21:28, Tomas Vondra
<tomas(dot)vondra(at)enterprisedb(dot)com> wrote:
>
> Here is a slightly updated version of the patch
>
> The main issue I ran into
> is the special case clauselist_selectivity, which does
>
> if (list_length(clauses) == 1)
> return clause_selectivity_ext(...);
>
> which applies to cases like "WHERE a < b" which can now be handled by
> extended statistics, thanks to this patch. But clause_selectivity_ext
> only used to call restriction_selectivity for these clauses, which does
> not use extended statistics, of course.
>
> I considered either getting rid of the special case, passing everything
> through extended stats, including cases with a single clause. But that
> ends up affecting e.g. OR clauses, so I tweaked clause_selectivity_ext a
> bit, which seems like a better approach.

Yeah, I looked at this a few months ago, while looking at the other
extended stats stuff, and I came to the same conclusion that solving
this issue by tweaking clause_selectivity_ext() is the best approach.

I haven't looked at the patch in much detail yet, but I think the
basic approach looks OK.

There are a few comments that need updating, e.g.:
- In statext_is_compatible_clause_internal(), before the "if
(is_opclause(clause))" test.
- The description of the arguments for examine_opclause_args().

I wonder if "expronleftp" for examine_opclause_args() should be
"constonrightp", or some such -- as it stands it's being set to false
for an Expr Op Expr clause, which doesn't seem right because there
*is* an expression on the left.

Regards,
Dean