Date: Wed, 23 Apr 2008 09:58:10 +0200
From: "A. Kretschmer" <andreas.kretschmer@schollglas.com>
To: pgsql-performance@postgresql.org
Subject: Re: mysterious difference in speed when combining two queries with OR
Message-ID: <20080423075810.GB8401@a-kretschmer.de>
References: <20080423072303.GF11886@amin>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20080423072303.GF11886@amin>
User-Agent: Mutt/1.5.9i

am  Wed, dem 23.04.2008, um  9:23:07 +0200 mailte Hans Ekbrand folgendes:
> I cannot understand why the following two queries differ so much in execution time (almost ten times)

wild guess: different execution plans.


Can you show us the plans? (EXPLAIN ANALYSE SELECT ...)

> 
> Query A (two queries)
> 
> select distinct moment.mid from moment,timecard where parent = 45 and (pid=17 and timecard.mid = moment.mid) order by moment.mid;
> select distinct moment.mid from moment,timecard where parent = 45 and (pbar = 0) order by moment.mid;
> 
> Query B (combining the two with OR)
> 
> select distinct moment.mid from moment,timecard where parent = 45 and ((pid=17 and timecard.mid = moment.mid) or (pbar = 0)) order by moment.mid;
>
> [ snip ]
> 
> I should say that this is on postgresql 7.4.16 (debian stable).

Uhh. Why not a recent version? We have 8.3.0...


> 
> Can query B be rewritten so that it would execute faster?

Quick and dirty: use both selects (query A) combined with UNION.

I guess, with a recent version the planner can use a bitmap index scan
to perform Query B faster.


Andreas
-- 
Andreas Kretschmer
Kontakt:  Heynitz: 035242/47150,   D1: 0160/7141639 (mehr: -> Header)
GnuPG-ID:   0x3FFF606C, privat 0x7F4584DA   http://wwwkeys.de.pgp.net