To: mark@mark.mielke.cc
cc: Bruce Momjian <bruce@momjian.us>, Jie Zhang <jzhang@greenplum.com>,
	Hannu Krosing <hannu@skype.net>, Gavin Sherry <swm@linuxworld.com.au>,
	pgsql-hackers@postgresql.org, Luke Lonergan <LLonergan@greenplum.com>
Subject: Re: On-disk bitmap index patch 
In-reply-to: <20060725015805.GA30087@mark.mielke.cc> 
References: <C0EAB84F.8FB4%jzhang@greenplum.com>
	<200607250104.k6P14St25388@momjian.us>
	<20060725015805.GA30087@mark.mielke.cc>
Comments: In-reply-to mark@mark.mielke.cc
	message dated "Mon, 24 Jul 2006 21:58:05 -0400"
Date: Tue, 25 Jul 2006 00:36:42 -0400
Message-ID: <5085.1153802202@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

mark@mark.mielke.cc writes:
> Reading 1/4, for a larger table, has a good chance of being faster than
> reading 4/4 of the table. :-)

Really?

If you have to hit one tuple out of four, it's pretty much guaranteed
that you will need to fetch every heap page.  So using an index provides
zero I/O savings on the heap side, and any fetches needed to read the
index are pure cost.  Now you have to demonstrate that the CPU costs
involved in processing the index are significantly cheaper than the cost
of just testing the WHERE qual at every heap tuple --- not a bet that's
likely to win at a one-in-four ratio.

> Will it be worth it or not? I won't know until I try it. :-)

Agreed.

			regards, tom lane