From: Tom Lane <tgl@sss.pgh.pa.us>
To: nh.dev@sturdy.ai
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #17721: A completely unused CTE negatively affect Query Plan
In-reply-to: <17721-a660b93948b165a7@postgresql.org>
References: <17721-a660b93948b165a7@postgresql.org>
Comments: In-reply-to PG Bug reporting form <noreply@postgresql.org>
	message dated "Wed, 14 Dec 2022 17:41:37 +0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2784035.1671042328.1@sss.pgh.pa.us>
Date: Wed, 14 Dec 2022 13:25:28 -0500
Message-ID: <2784036.1671042328@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/2784036.1671042328%40sss.pgh.pa.us>
Precedence: bulk

PG Bug reporting form <noreply@postgresql.org> writes:
> I've boiled down an issue we have in production to a simple query that can
> demonstrate it.  I've run this on 13,14 and 15 locally in a docker container
> with the same results.  If a CTE that is completely unexecuted exists in a
> subquery (or view in our production case) it affects the query plan VERY
> negatively.  The first explain shows a sequential scan, where the second
> explain shows an index scan, just by the existence of the (obviously)
> unexecuted CTE.

The presence of the CTE prevents flattening of the subquery.
If you don't like it, don't attach the CTE right there.

			regards, tom lane