Message-ID: <47B2B058.7020804@archonet.com>
Date: Wed, 13 Feb 2008 08:54:48 +0000
From: Richard Huxton <dev@archonet.com>
User-Agent: Thunderbird 2.0.0.9 (X11/20071031)
MIME-Version: 1.0
To: "Dean Gibson (DB Administrator)" <postgresql@ultimeth.com>
CC: pgsql-performance@postgresql.org
Subject: Re: Optimizing No matching record Queries
References: <47B2112C.5070707@livedatagroup.com>
	<47B234C1.1090707@ultimeth.com>
In-Reply-To: <47B234C1.1090707@ultimeth.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Dean Gibson (DB Administrator) wrote:
> The questions are:
> 
> 1. Why in the planner scanning the entire idx_listing_entrydate, when 
> I'd think it should be scanning the entire 
> pk_listingstatus_listingstatusid ?

It's looking at the ORDER BY and sees that the query needs the 10 most 
recent, so tries searching by date. That's sensible where you are going 
to have a lot of matches for fklistingsourceid.

Which suggests that statistics for "fklistingsourceid" aren't high 
enough, like Greg suggested. If that doesn't help, the index on 
(fklistingsourceid,entrydate) that Stephen might well do so.

> 2. Why is "Index Scan using pk_listingstatus_listingstatusid on 
> listingstatus listingsta1_  (cost=0.00..0.27 rows=1 width=4) (never 
> executed)" ?

Because nothing comes out of the first index-scan.

-- 
   Richard Huxton
   Archonet Ltd