From: "Qingqing Zhou" <zhouqq@cs.toronto.edu>
Subject: Re: performance of implicit join vs. explicit conditions on inet
	queries?
Date: Tue, 1 Nov 2005 20:19:37 -0500
Organization: Hub.Org Networking Services
Lines: 30
Message-ID: <dk946n$2m3p$1@news.hub.org>
References: <20051024035809.GA18261@edmonds.ath.cx>
	<dk4p9d$30q4$1@news.hub.org> <29864.1130768650@sss.pgh.pa.us>
To: pgsql-performance@postgresql.org


"Tom Lane" <tgl@sss.pgh.pa.us> wrote
>
> No, that's completely irrelevant to his problem.  The reason we can't do
> this is that the transformation from "x << const" to a range check on x
> is a plan-time transformation; there's no mechanism in place to do it
> at runtime.  This is not easy to fix, because the mechanism that's doing
> it is primarily intended for LIKE/regex index optimization, and in that
> case a runtime pattern might well not be optimizable at all.
>

Not quite understand, sorry ...

(1) For this query (in an as-is PG syntax, which find out all rectangles lie 
in a given rectangle) :

SELECT r FROM all_rectangles
  WHERE r << rectangle('(1,9),(9,1)');

If there is a GiST/Rtree index associated with all_rectangles.r, how do 
optimizer estimate the cost to decide that we should use this index or 
not(then by a seqscan)?

(2) Does your above explaination mean that we can't use GiST for a spatial 
join operation?

Regards,
Qingqing