X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [64.117.224.130]) by svr1.postgresql.org (Postfix) with ESMTP id 274C5D1B52B for ; Wed, 17 Sep 2003 00:21:49 +0000 (GMT) Received: from svr1.postgresql.org ([64.117.224.193]) by localhost (neptune.hub.org [64.117.224.130]) (amavisd-new, port 10024) with ESMTP id 83278-03 for ; Tue, 16 Sep 2003 21:21:19 -0300 (ADT) Received: from mail.osdl.org (fw.osdl.org [65.172.181.6]) by svr1.postgresql.org (Postfix) with ESMTP id 1D6BBD1B4F0 for ; Tue, 16 Sep 2003 21:21:16 -0300 (ADT) Received: from ibm-e.pdx.osdl.net (ibm-e.pdx.osdl.net [172.20.1.71]) by mail.osdl.org (8.11.6/8.11.6) with ESMTP id h8H0L4o11077; Tue, 16 Sep 2003 17:21:04 -0700 Subject: Re: [GENERAL] how to get accurate values in pg_statistic From: Mary Edie Meredith To: pgsql-performance , osdldbt-general Content-Type: text/plain Organization: Open Source Development Lab Message-Id: <1063758041.5730.549.camel@ibm-e.pdx.osdl.net> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.2.2 Date: 16 Sep 2003 17:21:04 -0700 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at postgresql.org X-Archive-Number: 200309/239 X-Sequence-Number: 3605 Yesterday Jenny and I started to look at plan changes with different seed and default_statistics_sample changes. Since we have 21 plans to check, it takes a long time to determine if the plans were different. We had to do it visually with xxdiff. Diff will always show a difference since the costs are almost always different. Is there any option to remove the cost numbers from the plan so we can just use "diff" to automate the plan comparisons? Otherwise it will be very tedious to do this experiment. On Sun, 2003-09-07 at 17:32, Tom Lane wrote: > Greg Stark writes: > > Perhaps the default of 10 is simply way > > too small and should be raised? > > I've suspected since the default existed that it might be too small ;-). > No one's yet done any experiments to try to establish a better default, > though. I suppose the first hurdle is to find a representative dataset. > > regards, tom lane > > ---------------------------(end of broadcast)--------------------------- > TIP 8: explain analyze is your friend -- Mary Edie Meredith Open Source Development Lab