Message-ID: <40FD2A71.2030104@ricker.us>
Date: Tue, 20 Jul 2004 10:21:37 -0400
From: Vic Ricker <vic@ricker.us>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US;
	rv:1.7) Gecko/20040616 MultiZilla/ StumbleUpon/1.994
MIME-Version: 1.0
To: SZUCS Gabor <surrano@mailbox.hu>
Cc: pgsql-sql@postgresql.org
Subject: Re: How do I convice postgres to use an index?
References: <Pine.LNX.4.44.0407151659420.12214-100000@matrix.gatewaynet.com>
	<050001c46d6d$0fb2dea0$0403a8c0@fejleszt4>
In-Reply-To: <050001c46d6d$0fb2dea0$0403a8c0@fejleszt4>
Content-Type: text/plain; charset=ISO-8859-7; format=flowed
Content-Transfer-Encoding: 7bit

The plan showed that Postgres did the proper cast on the timestamp.  I 
think part of the problem is that I need to increase the memory 
allocated to effective_cache_size for the optimizer choose the proper 
method.  (Thanks to Richard Huxton for help.)  I've read that if it 
doesn't have enough ram, it will forego the index for a sequential 
scan..  In my case, that's a very poor decision on the optimizer's 
part.  Disabling enable_seqscan seems to generally fix the problem but 
I'm afraid that it might degrade performance elsewhere.  I have 
expermiented with the effective_cache_size and some other settings but 
haven't had as much luck.  I think I need more physical ram.  Will try 
that soon.

-Vic


SZUCS Gabor wrote:

>re-checked; it's WITHOUT in both version, but it's irrelevant if you give
>the full spec. Well, then maybe it was a difference between 7.2 and 7.3, but
>again, it's irrelevant in your case. Have you tried the typecast?
>
>  
>