Message-ID: <00d701c46a78$1d675eb0$0403a8c0@fejleszt4>
From: =?iso-8859-1?Q?SZUCS_G=E1bor?= <surrano@mailbox.hu>
To: <pgsql-sql@postgresql.org>
References: <40F44635.6050202@ricker.us>
Subject: Re: How do I convice postgres to use an index?
Date: Thu, 15 Jul 2004 16:29:17 +0200
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit

Try casting the constant to the type of the field, i.e.

  WHERE timestamp >= '7/12/2004'::"timestamp without time zone"

(iirc the quotes are necessary)

Also, I'd try to avoid naming attributes like (built-in) types. (iirc
"timestamp" is a type without time zone in 7.3, and with time zone in 7.4 :)
"timestamptz" is the opposite in both cases)

G.
%----------------------- cut here -----------------------%
\end

----- Original Message ----- 
From: "Vic Ricker" <vic@ricker.us>
Sent: Tuesday, July 13, 2004 10:29 PM


> I apologize for the following stupid question.  I have been doing some
> searching and haven't found anything really helpful.
>
> The problem is that postgres (7.4.2) keeps choosing to do a sequential
> scan on a table when an index scan would be significantly faster.
>
> The queries that I'm using look at daily statistics from events logged
> by our Checkpoint firewall and generate graphs.  Since they are bit
> complicated, I simplified it to "select count(*) from log where
> timestamp>='7/12/2004'" for testing.