From: Bruce Momjian <pgman@candle.pha.pa.us>
Message-Id: <200208261954.g7QJsmW18141@candle.pha.pa.us>
Subject: Re: LIMIT 1 FOR UPDATE or FOR UPDATE LIMIT 1?
In-Reply-To: <20020826215027.A15218@ford.rockstorm.se>
To: magnus.enbom@rockstorm.se
Date: Mon, 26 Aug 2002 15:54:48 -0400 (EDT)
Cc: Jan Wieck <JanWieck@Yahoo.com>, pgsql-sql@postgresql.org
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=US-ASCII

Magnus Enbom wrote:
> On Mon, Aug 26, 2002 at 02:42:26PM -0400, Jan Wieck wrote:
> > Bruce Momjian wrote:
> > > 
> > > I found this email from April.  It properly points out that our
> > > LIMIT/FOR UPDATE ordering doesn't match MySQL's, and MySQL's looks more
> > > correct, specifically that the FOR UPDATE is after the LIMIT.  Our
> > > grammar is:
> > 
> > How do you define "correct" for "non-standard" features? And why don't
> > you ask Monty first to change to our "de-facto-standard"? ;-)
> 
> Already done that. ;-)
> He said he would look into it(having MySQL accept both behaviors), but if 
> it would require a big change of their grammar(for a value of big), he'd rather
> not. He also pointed out(as Bruce and Tom have done) that our(PG) way is 
> kind of backwards. 
> If you look at Oracle, you can see that they also have it last:
> 
> select :== subquery -> for_update_clause ;
> 
> OTOH, Oracle doesn't have LIMIT, but that's another story...
> 

Yep, we clearly have it backwards.  Now, how to address it:

	1) leave it unchanged
	2) allow only new ordering
	3) allow both orderings for one release
	4) allow both ordering forever
	

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073