To: "Tim Truman" <tim@advam.com>
cc: "'Dave Dutcher'" <dave@tridecap.com>, pgsql-performance@postgresql.org
Subject: Re: Forcing the use of particular execution plans 
In-reply-to: <000001c6e6a0$5903b990$0340050a@sdg> 
References: <000001c6e6a0$5903b990$0340050a@sdg>
Comments: In-reply-to "Tim Truman" <tim@advam.com>
	message dated "Tue, 03 Oct 2006 13:29:37 +0930"
Date: Tue, 03 Oct 2006 00:19:39 -0400
Message-ID: <19401.1159849179@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

"Tim Truman" <tim@advam.com> writes:
> Here is an "explain analyze" for the query that performs slowly,

This shows that the planner is exactly correct in thinking that all
the runtime is going into the seqscan on transaction:

> "Aggregate  (cost=88256.32..88256.32 rows=1 width=0) (actual
> time=55829.000..55829.000 rows=1 loops=1)"
> ...
> "                          ->  Seq Scan on "transaction" t
> (cost=0.00..87061.04 rows=1630 width=349) (actual time=234.000..55797.000
> rows=200 loops=1)"
> "                                Filter: ((transaction_date >=
> '2005-01-01'::date) AND (transaction_date <= '2006-09-25'::date) AND
> ((credit_card_no)::text ~~ '4564%549'::text))"

Since that component of the plan was identical in your two original
plans ("desired" and "undesired") it seems pretty clear that you have
not correctly identified what your problem is.

			regards, tom lane