From: Tom Lane <tgl@sss.pgh.pa.us>
To: Benny Kramek <benny@medflyt.com>
cc: pgsql-performance@lists.postgresql.org
Subject: Re: Slow performance with trivial self-joins
In-reply-to: 
 <CAGPCyEZWEVjkDtuHipOaPKbO2WXdLUuNBhj818-wEuhi4KLbzA@mail.gmail.com>
References: 
 <CAGPCyEZWEVjkDtuHipOaPKbO2WXdLUuNBhj818-wEuhi4KLbzA@mail.gmail.com>
Comments: In-reply-to Benny Kramek <benny@medflyt.com>
	message dated "Mon, 03 Feb 2020 15:38:15 -0500"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <28147.1580764220.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 03 Feb 2020 16:10:20 -0500
Message-ID: <28148.1580764220@sss.pgh.pa.us>
Precedence: bulk

Benny Kramek <benny@medflyt.com> writes:
> I expect the query plan to be identical for both of the below queries (a=
nd I
> expect the performance to also be identical).

[ shrug... ] Your expectation is mistaken.  There is no code in Postgres
to eliminate useless self-joins.  People have been fooling around with
a patch to do so [1], but I'm unsure whether it'll ever get committed,
or whether we even want the feature.  It seems not unlikely that the
time wasted trying to identify useless self-joins (in queries where
the optimization doesn't actually apply) would outweigh the win when
it does apply.  So there's a limit to how much effort the server should
spend trying to clean up after poorly-written queries.

			regards, tom lane

[1] https://www.postgresql.org/message-id/flat/64486b0b-0404-e39e-322d-080=
1154901f3@postgrespro.ru