Horton

Re: Failed Assert in pgstat_assoc_relation

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
Cc: vignesh C <vignesh21@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2022-12-02T05:48:48Z
Lists: pgsql-hackers 
Andres Freund <andres@anarazel.de> writes:
> On 2022-12-02 00:08:20 -0500, Tom Lane wrote:
>> Just the table-to-view hack.  I'm not aware that there are any other
>> cases, and it seems hard to credit that there ever will be any.

> I can see some halfway credible scenarios. E.g. converting a view to a
> matview, or a table into a partition. I kind of wonder if it's worth keeping
> the change, just in case we do - it's not that easy to hit...

I'd suggest putting in an assertion that the relkind isn't changing,
instead.  When and if somebody makes a credible feature patch that'd
require relaxing that, we can see what to do.

(There's a couple of places in rewriteHandler.c that could
perhaps be simplified, too.)

			regards, tom lane

Commits

  1. Prevent pgstats from getting confused when relkind of a relation changes

  2. Remove logic for converting a table to a view.