Re: UPDATE .. RETURNING OLD.*

Lists: pgsql-hackers
From: Marko Tiikkaja <marko(dot)tiikkaja(at)cs(dot)helsinki(dot)fi>
To: PostgreSQL development <pgsql-hackers(at)postgresql(dot)org>
Subject: UPDATE .. RETURNING OLD.*
Date: 2009-08-28 15:12:30
Message-ID: 4A97F3DE.9050005@cs.helsinki.fi
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

Hi everyone,

Today I needed a feature like $subject. The use case was: UPDATE foo SET
bar = bar + 1 WHERE id=$1, but I wanted to only do it when bar was 0. In
order to give the user an informative error message, I also needed to
distinguish the two cases: a row with id = $1 doesn't exist, and bar was
0, so I couldn't put bar != 0 into the WHERE clause. This time I got
around it by using RETURNING bar and checking that it was 1 on the
client side, but I can come up with other cases where you can't do that.

Comments?

Regards,
Marko Tiikkaja


From: David Fetter <david(at)fetter(dot)org>
To: Marko Tiikkaja <marko(dot)tiikkaja(at)cs(dot)helsinki(dot)fi>
Cc: PostgreSQL development <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: UPDATE .. RETURNING OLD.*
Date: 2009-08-28 18:52:23
Message-ID: 20090828185223.GG3886@fetter.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-hackers

On Fri, Aug 28, 2009 at 06:12:30PM +0300, Marko Tiikkaja wrote:
> Hi everyone,
>
> Today I needed a feature like $subject. The use case was: UPDATE
> foo SET bar = bar + 1 WHERE id=$1, but I wanted to only do it when
> bar was 0. In order to give the user an informative error message,
> I also needed to distinguish the two cases: a row with id = $1
> doesn't exist, and bar was 0, so I couldn't put bar != 0 into the
> WHERE clause. This time I got around it by using RETURNING bar and
> checking that it was 1 on the client side, but I can come up with
> other cases where you can't do that.
>
> Comments?

We talked about this briefly in IRC last night, and since that's not
recorded, I'd like to mention a few things here:

* OLD is already a reserved word. We could use it without fear of a
badly named database object.

* Having access to both the old and new row could make debugging
complex UPDATE queries much easier.

* There's some interesting use cases if the UPDATE...RETURNING can
also be used as a subquery. Auditing would be one.

Cheers,
David.
--
David Fetter <david(at)fetter(dot)org> http://fetter.org/
Phone: +1 415 235 3778 AIM: dfetter666 Yahoo!: dfetter
Skype: davidfetter XMPP: david(dot)fetter(at)gmail(dot)com

Remember to vote!
Consider donating to Postgres: http://www.postgresql.org/about/donate