Date: Wed, 11 Aug 2004 21:27:15 -0500
From: Bruno Wolff III <bruno@wolff.to>
To: David Stanaway <david@stanaway.net>
Cc: pgsql-sql <pgsql-sql@postgresql.org>
Subject: Re: UPDATE FROM problem,
	multiple updates of same row don't seem to work
Message-ID: <20040812022715.GB22747@wolff.to>
Mail-Followup-To: David Stanaway <david@stanaway.net>,
	pgsql-sql <pgsql-sql@postgresql.org>
References: <1092082589.7197.3.camel@dmxnocws13.dialmex.net>
	<20040811235241.GB19680@wolff.to>
	<FCEA9328-EC01-11D8-A0B2-0003930FDAB2@stanaway.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <FCEA9328-EC01-11D8-A0B2-0003930FDAB2@stanaway.net>
User-Agent: Mutt/1.5.6i

On Wed, Aug 11, 2004 at 20:50:28 -0500,
  David Stanaway <david@stanaway.net> wrote:
> 
> I had thought about that, but this is a simpler case of what I need to 
> do. The operations for each column in the update are dependent on the 
> current and new values of each row being merged.
> 
> Currently I am doing this with a cursor which is very slow! How about a 
> trigger on update? Or would this suffer the same behavior that I am 
> seeing with UPDATE FROM with a 1-many join?

There is a good chance that triggers would be faster since tablea seems
to reallly be a materialized view of an aggregate over tableb and doing
a delta calculation when a row changes is going to be faster than recalculating
the aggregate from scratch.

You won't have the same problem if you use a trigger as when you were
trying to calculate an aggregate using UPDATE.