To: Ben Chobot <bench@silentmedia.com>
cc: "Kevin Grittner" <Kevin.Grittner@wicourts.gov>,
	pgsql-performance@postgresql.org
Subject: Re: 8.1 -> 8.4 regression 
In-reply-to: <5FDB08C1-3835-4B80-835D-B44F483C3227@silentmedia.com> 
References: <4B791AFF020000250002F2DD@gw.wicourts.gov>
	<8D572A50-01C7-464B-8D0D-FCF19C53BC9C@silentmedia.com>
	<1376.1266254788@sss.pgh.pa.us>
	<5FDB08C1-3835-4B80-835D-B44F483C3227@silentmedia.com>
Comments: In-reply-to Ben Chobot <bench@silentmedia.com>
	message dated "Mon, 15 Feb 2010 09:35:13 -0800"
Date: Wed, 17 Feb 2010 00:34:37 -0500
Message-ID: <19449.1266384877@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Ben Chobot <bench@silentmedia.com> writes:
> Awesome, that did the trick. Thanks Tom! So I understand better, why is my case not the normal, better case?

Well, the short answer is that the 8.4 changes here are in the nature of
two steps forward and one step back.  The long-term goal is to increase
the planner's ability to choose among different join orders; but we're
getting rid of one restriction at a time, and sometimes the interactions
of those restrictions produce unwanted results like the older code being
able to find a better plan than the new code can.

> (I assume the long-term fix is post-9.0, right?)

Yeah, fraid so.  I've been mostly buried in non-planner work in the 9.0
cycle, but hope to get back to this and other problems in the next
cycle.

			regards, tom lane