In-Reply-To: <ab07320e0706010948m7c4eb1b8s9cdaf79b1c400125@mail.gmail.com>
References: <ab07320e0706010948m7c4eb1b8s9cdaf79b1c400125@mail.gmail.com>
Mime-Version: 1.0 (Apple Message framework v752.2)
Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed
Message-Id: <96F60F91-F3A7-4134-8DD0-E519C8941CDB@seespotcode.net>
Cc: pgsql-performance@postgresql.org
Content-Transfer-Encoding: 7bit
From: Michael Glaesemann <grzm@seespotcode.net>
Subject: Re: Seq Scan
Date: Fri, 1 Jun 2007 12:03:30 -0500
To: Tyler Durden <tylersticky@gmail.com>


On Jun 1, 2007, at 11:48 , Tyler Durden wrote:

> I'm having some problems in performance in a simple select count(id)
> from....

Unrestricted count() (i.e., no WHERE clause) will perform a  
sequential scan. If you're looking for faster ways to store table row  
count information, please search the archives, as this has been  
discussed many times before.

> # explain select (id) from table_name;
> -[ RECORD  
> 1 ]----------------------------------------------------------------
> QUERY PLAN | Seq Scan on table_name  (cost=0.00..8601.30  
> rows=266730 width=4)

The query returns the id column value for each row in the table. The  
fastest way to do this is visiting every row., i.e., a sequential  
scan. Using an index would require (1) looking in the index and (2)  
looking up the corresponding row.

Michael Glaesemann
grzm seespotcode net