From: Mark Stosberg <mark@summersault.com>
Subject: tip: faster sorting for proximity queries by using cube_distance()
Date: Wed, 07 Feb 2007 16:30:16 -0500
Organization: Hub.Org Networking Services
Lines: 23
Message-ID: <eqdge4$291m$1@news.hub.org>
References: <eq2m5o$8cr$1@sea.gmane.org> <20070204051644.GA3754@wolff.to>	
	<eq81l5$2na3$1@news.hub.org>
	<b42b73150702051215s6fe341e2re43872ab84abb185@mail.gmail.com>
	<eq8d0b$26cd$1@news.hub.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
User-Agent: Thunderbird 1.5.0.8 (X11/20061109)
In-Reply-To: <eq8d0b$26cd$1@news.hub.org>
To: pgsql-performance@postgresql.org

Hello,

I wanted to share something else I learned in my proximity search work.
 One my requirements is to order by the distance that matches are found
from the center point.

When did this using earth_distance(), the benefit of the earth_box()
technique over the old geo_distance became minimal as I approached a
250mi radius.

Switching to sorting by cube_distance() offered a huge benefit, allowing
the earth_distance() query to run in about 100ms vs 300ms for the
geo_distance() equivalent.

I checked the results that cube_distance() produced versus
earth_distance(). cube_distance() is always (not surprisingly) a little
smaller, but the difference seems only grows to about a mile for a 250
mile radius. That's an acceptable margin of error for this application,
and may be for others as well.

   Mark