Date: Fri, 22 May 2015 02:48:11 +0200
From: Andres Freund <andres@anarazel.de>
To: Alvaro Herrera <alvherre@2ndquadrant.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Peter Geoghegan <pg@heroku.com>,
 Pg Hackers <pgsql-hackers@postgresql.org>
Subject: Re: Re: [COMMITTERS] pgsql: Add support for INSERT ... ON
 CONFLICT DO NOTHING/UPDATE.
Message-ID: <20150522004811.GD2028@alap3.anarazel.de>
References: <E1YqZCW-0000e6-Od@gemulon.postgresql.org>
 <20150520215816.GI5885@postgresql.org>
 <CAM3SWZRUhr1806zsF28CVRrB=b4t8PWv8d9x29z02xTQoFDyWA@mail.gmail.com>
 <20150520220911.GK5885@postgresql.org>
 <CAM3SWZQXP3Z=iXQ_+HwoqG7417JYf6HKDTmgk-4Xt05OEL_S_Q@mail.gmail.com>
 <CAM3SWZT0Rex4CpZo+re_m2Dh7wunMNsbD-9ZSWWx=JSewW3OVg@mail.gmail.com>
 <20150521011229.GX27868@alap3.anarazel.de>
 <10660.1432171328@sss.pgh.pa.us>
 <20150521013059.GY27868@alap3.anarazel.de>
 <20150521232841.GO5885@postgresql.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20150521232841.GO5885@postgresql.org>
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

On 2015-05-21 20:28:41 -0300, Alvaro Herrera wrote:
> That said, I'm not sure about having it be the same, either: first, I
> don't think we need to update the fe-exec.c code at all -- I mean, all
> the things I see there are very old compatibility stuff; reporting the
> OID of the just-inserted row?  Seriously, who has OIDs in user tables
> anymore?  Surely we wouldn't try to do that for INSERT ON CONFLICT DO
> UPDATE at all ... if anyone wants that functionality, they can use
> RETURNING, which is far saner.

The oid currently is reported for UPSERT... I agree it's not worth much,
but it seems pointless to break it for a single command.

> As for PQcmdTuples, what would a single number returned mean?  Are we
> returning the sum of both inserted and updated tuples?

Yes.

> Isn't this a bit odd?

Imo it's pretty much in line with what's done with INSTEAD OF, FDWs and
such.

> If the number is useful, then perhaps we should distinguish the cases;
> and if it's not useful, why not just return zero?

There's libraries/frameworks checking if an insert succeeded by looking
at that number, and it seems like a bad idea to needlessly break those.


So I think we're good with ripping it out. Peter?


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers