Date: Thu, 4 Oct 2007 14:14:11 -0400
From: Bill Moran <wmoran@collaborativefusion.com>
To: Ben <bench@silentmedia.com>
Cc: pgsql-performance@postgresql.org
Subject: Re: quickly getting the top N rows
Message-Id: <20071004141411.9509fe66.wmoran@collaborativefusion.com>
In-Reply-To: <Pine.LNX.4.64.0710041053590.30864@localhost.localdomain>
References: <Pine.LNX.4.64.0710041053590.30864@localhost.localdomain>
Organization: Collaborative Fusion
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

In response to Ben <bench@silentmedia.com>:

> If I have this:
> 
> create table foo (bar int primary key);
> 
> ...then in my ideal world, Postgres would be able to use that index on bar 
> to help me with this:
> 
> select bar from foo order by bar desc limit 20;
> 
> But in my experience, PG8.2 is doing a full table scan on foo, then 
> sorting it, then doing the limit. I have a more complex primary key, but I 
> was hoping the same concept would still apply. Am I doing something wrong, 
> or just expecting something that doesn't exist?

Show us the explain.

However, 2 guesses:
1) You never analyzed the table, thus PG has awful statistics and
   doesn't know how to pick a good plan.
2) You have so few rows in the table that a seq scan is actually
   faster than an index scan, which is why PG uses it instead.

-- 
Bill Moran
Collaborative Fusion Inc.
http://people.collaborativefusion.com/~wmoran/

wmoran@collaborativefusion.com
Phone: 412-422-3463x4023