Date: Thu, 21 Apr 2005 14:15:08 -0700 (PDT)
From: Stephan Szabo <sszabo@megazone.bigpanda.com>
To: Enrico Weigelt <weigelt@metux.de>
Cc: postgresql performance list <pgsql-performance@postgresql.org>
Subject: Re: index not used
In-Reply-To: <20050421190543.GA16548@nibiru.borg.metux.de>
Message-ID: <20050421140958.Y59173@megazone.bigpanda.com>
References: <20050421190543.GA16548@nibiru.borg.metux.de>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Thu, 21 Apr 2005, Enrico Weigelt wrote:

> I'm doing a simple lookup in a small table by an unique id, and I'm
> wondering, why explains tells me seqscan is used instead the key.
>
> The table looks like:
>
>     id	bigint		primary key,
>     a	varchar,
>     b	varchar,
>     c 	varchar
>
> and I'm quering: select * from foo where id = 2;
>
> I've got only 15 records in this table, but I wanna have it as
> fast as possible since its used (as a map between IDs and names)
> for larger queries.

Two general things:
 For 15 records, an index scan may not be faster.  For simple tests
  you can play with enable_seqscan to see, but for more complicated
  queries it's a little harder to tell.
 If you're using a version earlier than 8.0, you'll need to quote
  or cast the value you're searching for due to problems with
  cross-type comparisons (the 2 would be treated as int4).