From: Tom Lane <tgl@sss.pgh.pa.us>
To: Ancoron Luciferis <ancoron.luciferis@googlemail.com>
cc: pgsql-performance@lists.postgresql.org
Subject: Re: Custom opclass for column statistics?
In-reply-to: <b0590dd1-5bfb-ad62-c4d0-a7057c04a68b@googlemail.com>
References: <b0590dd1-5bfb-ad62-c4d0-a7057c04a68b@googlemail.com>
Comments: In-reply-to Ancoron Luciferis <ancoron.luciferis@googlemail.com>
	message dated "Sat, 06 Jul 2019 11:02:27 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <26688.1562421538.1@sss.pgh.pa.us>
Date: Sat, 06 Jul 2019 09:58:58 -0400
Message-ID: <26689.1562421538@sss.pgh.pa.us>
Precedence: bulk

Ancoron Luciferis <ancoron.luciferis@googlemail.com> writes:
> I've been wondering whether it is possible somehow to have the standard
> column statistics to respect a certain operator class?

In principle, pg_statistic can represent stats for a non-default opclass.
Teaching ANALYZE to collect such stats when appropriate, and then teaching
the planner to use them when appropriate, is left as an exercise for the
reader.

I think the "when appropriate" bit is actually the hardest part of that.
Possibly, if you were satisfied with a relatively manual approach,
you could proceed by using CREATE STATISTICS to declare interest in
keeping standard stats for a non-default sort order.  Not sure what to
do if you want it to be automatic, because I don't think people would
hold still for having ANALYZE collect stats for any random non-default
opclass automatically.  Maybe a new opclass property?

			regards, tom lane