Date: Tue, 25 Sep 2007 11:31:33 +0200
From: "Steinar H. Gunderson" <sgunderson@bigfoot.com>
To: Carlo Stonebanks <stonec.register@sympatico.ca>
Cc: 'Ow Mun Heng' <Ow.Mun.Heng@wdc.com>, pgsql-performance@postgresql.org
Subject: Re: REPOST: Nested loops row estimates always too high
Message-ID: <20070925093133.GA27052@uio.no>
Mail-Followup-To: Carlo Stonebanks <stonec.register@sympatico.ca>,
	'Ow Mun Heng' <Ow.Mun.Heng@wdc.com>,
	pgsql-performance@postgresql.org
References: <fcu215$2em3$1@news.hub.org>
	<1190616376.17050.51.camel@neuromancer.home.net>
	<fd8ukj$s38$1@news.hub.org>
	<1190681440.22945.8.camel@neuromancer.home.net>
	<53877A06E07146D2898619A7025C6107@serenity>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <53877A06E07146D2898619A7025C6107@serenity>
User-Agent: Mutt/1.5.16 (2007-06-11)

On Tue, Sep 25, 2007 at 12:53:55AM -0400, Carlo Stonebanks wrote:
> My problem is that I think that SRF's are causing my problems. The SRF's
> gets an automatic row estimate of 1000 rows. Add a condition to it, the
> planner guesses 333 rows. Even at 333, this is an overestimate of the number
> of rows returned.
> 
> I'm really disappointed - SRF's are a great way to place the enterprise's
> db-centric business logic at the server.

Fortunately, in 8.3 you can attach a row estimate to the function yourself,
which should most likely fix your problem. Look forward to the first beta :-)

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