Re: Add starelid, attnum to pg_stats and leverage this in pg_dump

On Mon, Mar 9, 2026 at 11:51 AM Nathan Bossart <nathandbossart(at)gmail(dot)com>
wrote:

> On Mon, Mar 09, 2026 at 10:44:58AM -0500, Sami Imseih wrote:
> >> Oh, a question I forgot to ask: why wouldn't we do the same thing for
> >> pg_stats_ext and pg_stats_ext_exprs?
> >
> > [...]
> >
> > If we do add relid there, it will be for consistency only.
>
> It might be only for consistency for now, but it strikes me as something
> that could be handy down the road (in which case it'd be nice to minimize
> the number of versions that need a fallback).
>
> --
> nathan
>

You're both right. Currently, we fetch extended stats one at a time, thus
there's no _immediate_ need to do so.

But "why wait until there is a crisis" is solid reasoning and for that I
had already coded up the change. I did have one small problem in that
Michael Paquier had hoped that we could get the expr index (-1, -2, etc) on
the expressions as that was something that we at least briefly thought we'd
need for importing expression statistics. However the existing query uses a
SELECT unnest(a), unnest(b) pattern in it, and WITH ORDINALITY is not
allowed, and the workarounds I found seemed a bit tortured. Hence, I
decided to leave that out so as not to distract from the now accepted
patch, and with that out of the way I'll happily inflict that tortured SQL
on y'all.