To: ddurst@larubber.com
Cc: pgsql-sql@postgresql.org
Subject: Re: LONG - Question on dealing w/ numerics 
In-reply-to: <36005.216.86.192.34.1043831871.squirrel@www.la-rubber.com> 
References: <34281.216.86.192.34.1043801682.squirrel@www.la-rubber.com>
	<19082.1043813296@sss.pgh.pa.us>
	<36005.216.86.192.34.1043831871.squirrel@www.la-rubber.com>
Comments: In-reply-to "David Durst" <ddurst@larubber.com>
	message dated "Wed, 29 Jan 2003 01:17:51 -0800"
Date: Wed, 29 Jan 2003 04:01:39 -0500
Message-ID: <21507.1043830899@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

"David Durst" <ddurst@larubber.com> writes:
> But I still think your point about the function selecting more than
> one line is valid.

> The problem is, the journal_line_id is not created until the insert
> occurs and there is no other unique ident than the journal_line_id.

Well, my standard answer to that would be "your WHERE clause should
select on a primary key (which could be multiple columns)".

We're pretty constrained in changing the semantics of plpgsql, because
the raison d' etre of that language is to emulate Oracle's (TM) PL/SQL
(TM), warts and all.

(Now, if you can show us that PL/SQL behaves differently in that
situation, we'll definitely be willing to change plpgsql.)

			regards, tom lane