Date: Sat, 22 Aug 2009 14:34:57 -0500
From: Kenneth Marshall <ktm@rice.edu>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Mark Cave-Ayland <mark.cave-ayland@siriusit.co.uk>,
	pgsql-hackers@postgresql.org
Subject: Re: Another try at reducing repeated detoast work for
	PostGIS
Message-ID: <20090822193457.GA4268@it.is.rice.edu>
References: <4201.1250530640@sss.pgh.pa.us> <4A8A7287.1080208@siriusit.co.uk>
	<25799.1250616857@sss.pgh.pa.us> <4A8FD328.2030408@siriusit.co.uk>
	<10470.1250959181@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <10470.1250959181@sss.pgh.pa.us>
User-Agent: Mutt/1.5.16 (2007-06-09)

On Sat, Aug 22, 2009 at 12:39:41PM -0400, Tom Lane wrote:
> Mark Cave-Ayland <mark.cave-ayland@siriusit.co.uk> writes:
> > So in conclusion, I think that patch looks good and that the extra time 
> > I was seeing was due to RECHECK being applied to the && operator, and 
> > not the time being spent within the index scan itself.
> 
> Thanks, I appreciate the followup.
> 
> I plan to go ahead and apply the patch to HEAD --- it doesn't conflict
> with Heikki's pending patch AFAICS, and no one has suggested an
> alternative that seems likely to get implemented soon.
> 
> I am a bit tempted to apply it to 8.4 as well; otherwise the PostGIS
> people are likely to start cluttering their code with this
> add-a-dummy-function workaround, which would be unproductive in the long
> run.  Comments?
> 
> 			regards, tom lane
> 
+1 for applying it to 8.4 as well.

Cheers,
Ken