Received: from bast.unixathome.org (bast.unixathome.org [216.187.105.150]) by postgresql.org (Postfix) with ESMTP id 8D76147597B for ; Sat, 16 Mar 2002 19:18:11 -0500 (EST) Received: from wocker (wocker.unixathome.org [192.168.0.99]) by bast.unixathome.org (Postfix) with ESMTP id 27A883F28 for ; Sat, 16 Mar 2002 19:18:14 -0500 (EST) From: "Dan Langille" Organization: DVL Software Limited To: pgsql-sql Date: Sat, 16 Mar 2002 19:18:11 -0500 MIME-Version: 1.0 Subject: Re: why the big difference on this explain analyze? Reply-To: dan@langille.org In-reply-to: <20020316220751.A5EE73F28@bast.unixathome.org> X-mailer: Pegasus Mail for Windows (v4.01) Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body Message-Id: <20020317001814.27A883F28@bast.unixathome.org> X-Archive-Number: 200203/219 X-Sequence-Number: 7098 On 16 Mar 2002 at 17:07, Dan Langille wrote: > I was trying to get a query to run. Yes run. Not run fast. Just run. > ;) It was taking far too long. > > Here is the before: > > explain > fp2migration=# > fp2migration=# explain > fp2migration-# UPDATE Ports_Check SET found_in_ports = TRUE > fp2migration-# WHERE EXISTS > fp2migration-# ( SELECT * > fp2migration(# FROM categories, element, ports > fp2migration(# WHERE ports.category_id = categories.id > fp2migration(# AND ports.element_id = element.id > fp2migration(# AND categories.name = > Ports_Check.category_name fp2migration(# AND > element.name = Ports_Check.port_name); NOTICE: QUERY PLAN: > > Seq Scan on ports_check (cost=0.00..7226574.01 rows=3354 width=27) > SubPlan > -> Hash Join (cost=103.10..1077.13 rows=1 width=1073) > -> Hash Join (cost=101.47..1075.49 rows=2 width=1030) > -> Seq Scan on ports (cost=0.00..938.99 rows=6999 > width=1000) > -> Hash (cost=101.40..101.40 rows=27 width=30) > -> Index Scan using element_name on element > (cost=0.00..101.40 rows=27 width=30) > -> Hash (cost=1.62..1.62 rows=1 width=43) > -> Seq Scan on categories (cost=0.00..1.62 rows=1 > width=43) FWIW, I solved the high cost by adding an index: create index ports_element_id on ports (element_id); The plan is now: fp2migration-# UPDATE Ports_Check SET found_in_ports = TRUE fp2migration-# WHERE EXISTS fp2migration-# ( SELECT * fp2migration(# FROM categories, element, ports fp2migration(# WHERE ports.category_id = categories.id fp2migration(# AND ports.element_id = element.id fp2migration(# AND categories.name = Ports_Check.category_name fp2migration(# AND element.name = Ports_Check.port_name); NOTICE: QUERY PLAN: Seq Scan on ports_check (cost=0.00..1772151.04 rows=3354 width=39) SubPlan -> Hash Join (cost=1.63..264.13 rows=1 width=1060) -> Nested Loop (cost=0.00..262.49 rows=2 width=1017) -> Index Scan using element_name on element (cost=0.00..100.78 rows=27 width=30) -> Index Scan using ports_element_id on ports (cost=0.00..6.01 rows=1 width=987) -> Hash (cost=1.62..1.62 rows=1 width=43) -> Seq Scan on categories (cost=0.00..1.62 rows=1 width=43) EXPLAIN Which takes about 7 seconds to run. Thank you. -- Dan Langille The FreeBSD Diary - http://freebsddiary.org/ - practical examples