Date: Tue, 14 Oct 2025 17:03:12 -0500
From: Nathan Bossart <nathandbossart@gmail.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: andrewbille@gmail.com, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #19086: pg_dump --data-only selects and do not uses index
 definitions for the dumped tables.
Message-ID: <aO7IoIEnQDbtFzrs@nathan>
References: <19086-871ff11017d03dc5@postgresql.org>
 <aO6mUj9H2WpEFylG@nathan>
 <1000041.1760471415@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1000041.1760471415@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/aO7IoIEnQDbtFzrs%40nathan>
Precedence: bulk

On Tue, Oct 14, 2025 at 03:50:15PM -0400, Tom Lane wrote:
> Even if it's been seen in the field, it hardly qualifies as a
> justification for complicating pg_dump's behavior.  There's no reason
> to think that catalog corruption preferentially affects indexes, or
> does so in this particular way.  Should we also not collect data on
> functions, types, etc etc?

FWIW the getIndexes() query does tend to be one of the slowest, even with
intact system indexes.  I've no concrete proposals, but there might be some
room for improvement.  I don't think we gain all that much by simply
avoiding the query in probably-somewhat-rare use-cases.  IMHO it ought to
be reworked for efficiency.

-- 
nathan