Date: Thu, 22 Mar 2007 10:41:51 -0400
From: Michael Stone <mstone+postgres@mathom.us>
Subject: Re: Performance of count(*)
In-reply-to: <b42b73150703220639x40dbac7en29318778b43e1e4c@mail.gmail.com>
To: pgsql-performance@postgresql.org
Mail-followup-to: pgsql-performance@postgresql.org
Message-id: <20070322144149.GF11402@mathom.us>
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii; format=flowed
Content-disposition: inline
References: <Pine.LNX.4.62.0703221149210.13747@wr-linux02>
	<b42b73150703220639x40dbac7en29318778b43e1e4c@mail.gmail.com>
User-Agent: Mutt/1.5.13 (2006-08-11)

On Thu, Mar 22, 2007 at 09:39:18AM -0400, Merlin Moncure wrote:
>You can get the approximate count by selecting reltuples from
>pg_class.  It is valid as of last analyze.

Of course, that only works if you're not using any WHERE clause. 
Here's a (somewhat ugly) example of getting an approximate count based 
off the statistics info, which will work for more complicated queries:
http://archives.postgresql.org/pgsql-sql/2005-08/msg00046.php
The ugliness is that you have to provide the whole query as a 
parameter to the function, instead of using it as a drop-in replacement 
for count. I assume that the TODO item is to provide the latter, but for 
now this method can be useful for UI type stuff where you just want to 
know whether there's "a little" or "a lot".

Mike Stone