Message-ID: <4D66F5B0.5080502@agliodbs.com>
Date: Thu, 24 Feb 2011 16:20:00 -0800
From: Josh Berkus <josh@agliodbs.com>
User-Agent: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-US;
	rv:1.9.1b3pre) Gecko/20090223 Thunderbird/3.0b2
MIME-Version: 1.0
To: Dave Crooke <dcrooke@gmail.com>
CC: Kevin Grittner <Kevin.Grittner@wicourts.gov>,
	Michael Glaesemann <grzm@seespotcode.net>, pgsql-performance@postgresql.org
Subject: Re: Picking out the most recent row using a time stamp
 column
References: <AANLkTinTgU-jfPOhrfteBvhAhgG_8zknpk13N9L4QMap@mail.gmail.com>
	<08B3368E-E623-4E9C-83F5-AB05B471AF4C@seespotcode.net>
	<4D666A0F020000250003AFCE@gw.wicourts.gov>
	<AANLkTimSmjGrCqwZWRXU1_=FW0adbWdxLXMK0n=Vw+Cc@mail.gmail.com>
In-Reply-To: <AANLkTimSmjGrCqwZWRXU1_=FW0adbWdxLXMK0n=Vw+Cc@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit

On 2/24/11 3:38 PM, Dave Crooke wrote:
> Thanks to all .... I had a tickling feeling at the back of my mind that
> there was a neater answer here. For the record, times (all from
> in-memory cached data, averaged over a bunch of runs):
> 
> Dependent subquery = 117.9 seconds
> Join to temp table = 2.7 sec
> DISTINCT ON = 2.7 sec

But wait, there's more!  You haven't tested the Windowing Function
solution.  I'll bet it's even faster.


SELECT id_key, time_stamp, value
FROM (
	SELECT id_key, time_stamp, value,
		row_number()
		OVER ( PARTITION BY id_key
			ORDER BY time_stamp DESC)
 		as ranking
	FROM thetable
      ) as filtered_table
WHERE ranking = 1

-- 
                                  -- Josh Berkus
                                     PostgreSQL Experts Inc.
                                     http://www.pgexperts.com