From: Tom Lane <tgl@sss.pgh.pa.us>
To: christoph.berg@credativ.de
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #17123: ALTER TABLE ... NO INHERIT leaves orphaned statistics
 behind
In-reply-to: <17123-734abc3934f500b6@postgresql.org>
References: <17123-734abc3934f500b6@postgresql.org>
Comments: In-reply-to PG Bug reporting form <noreply@postgresql.org>
	message dated "Mon, 26 Jul 2021 12:54:22 -0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <86585.1627309764.1@sss.pgh.pa.us>
Date: Mon, 26 Jul 2021 10:29:24 -0400
Message-ID: <86586.1627309764@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/86586.1627309764%40sss.pgh.pa.us>
Precedence: bulk

PG Bug reporting form <noreply@postgresql.org> writes:
> Removing the last/only inheritance child from a table using ALTER TABLE NO
> INHERIT leaves orphaned statistics behind that ANALYZE doesn't remove.
> Original customer complaint on 10.16, but 15devel has the same issue.

Good catch.  I imagine DETACH PARTITION has related issues?

It seems like there are basically two ways to fix this:

1. Make ALTER TABLE NO INHERIT/DETACH PARTITION get rid of
no-longer-wanted stats.

2. Make ANALYZE check for and remove any inappropriate stats.

I'm kind of inclined to prefer #1, on the grounds that #2 would
add almost-always-useless cycles to every ANALYZE run forevermore.
On the other hand, #2 seems a bit more robust, in that we'd not
have to remember to consider this issue every time somebody invents
a new flavor of DDL that affects these things.

Thoughts?

			regards, tom lane