Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Subject: Re: index vs. seq scan choice? 
Date: Thu, 7 Jun 2007 15:52:16 -0700
Message-ID: 
 <8C5B026B51B6854CBE88121DBF097A86DEA6D3@ehost010-33.exch010.intermedia.net>
In-Reply-To: <415.1181255628@sss.pgh.pa.us>
Thread-Topic: [GENERAL] index vs. seq scan choice? 
Thread-Index: AcepVA8eDuRDOS1oRqO0qeupEXUy5AAATr5g
References: 
 <8C5B026B51B6854CBE88121DBF097A86C3A30D@ehost010-33.exch010.intermedia.net>
	<27828.1180055291@sss.pgh.pa.us>
	<8C5B026B51B6854CBE88121DBF097A86DEA6B4@ehost010-33.exch010.intermedia.net>
	<4668804F.6090201@commandprompt.com>
	<8C5B026B51B6854CBE88121DBF097A86DEA6C5@ehost010-33.exch010.intermedia.net>
	<415.1181255628@sss.pgh.pa.us>
From: "George Pavlov" <gpavlov@mynewplace.com>
To: "Tom Lane" <tgl@sss.pgh.pa.us>, "Joshua D. Drake" <jd@commandprompt.com>,
	<pgsql-general@postgresql.org>

> From: Tom Lane
> "George Pavlov" <gpavlov@mynewplace.com> writes:
> >> From: Joshua D. Drake [mailto:jd@commandprompt.com]
> >> In those rare cases wouldn't it make more sense to just set
> >> enable_seqscan to off; run query; set enable_seqscan to on;
>=20
> > 1. these cases are not that rare (to me);
>=20
> It strikes me that you probably need to adjust the planner cost
> parameters to reflect reality on your system.  Usually dropping
> random_page_cost is the way to bias the thing more in favor of
> index scans.

Thanks, Tom, I will try that. Seems better than fiddling with
enable_seqscan around every query/transaction.

Joshua, I fail to understand why setting and unsetting enable_seqscan on
a per query/transaction basis is in any way preferable to query hints?
Don't get me wrong, I don't like the idea of hints, and I have read the
archives on the subject and I agree with the philosophy, but if the
optimization toolkit for routine application queries is going to include
setting config parameters that just smacks of hints by another name...

George