Date: Mon, 30 Apr 2007 19:23:34 +0200
From: "Steinar H. Gunderson" <sgunderson@bigfoot.com>
To: pgsql-performance@postgresql.org
Subject: Re: Query performance problems with partitioned tables
Message-ID: <20070430172334.GA669@uio.no>
Mail-Followup-To: pgsql-performance@postgresql.org
References: <4635D678.3050205@xss.co.at> <87ps5mgghp.fsf@meuh.mnc.lan>
	<4635EF3A.1070204@xss.co.at>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <4635EF3A.1070204@xss.co.at>
User-Agent: Mutt/1.5.13 (2006-08-11)

On Mon, Apr 30, 2007 at 03:29:30PM +0200, Andreas Haumer wrote:
> This already gives a row matching the given WHERE clause.
> It makes no sense to scan the other tables, as the query
> asks for one row only and all the other tables have timestamps
> larger than all the timestamps in table t_mv_200601 (according
> to the CHECK constraints for the partion tables)

So for each row, it has to check all CHECK constraints to see if it has
enough rows? That sounds fairly inefficient.

I wonder if the planner could copy the limit down through the Append, though
-- it certainly doesn't need more than one row from each partition. It sounds
slightly cumbersome to try to plan such a thing, though...

/* Steinar */
-- 
Homepage: http://www.sesse.net/