Received: from sss.sss.pgh.pa.us (sss.pgh.pa.us [209.114.166.2]) by hub.org (8.9.3/8.9.3) with ESMTP id RAA34187 for ; Fri, 27 Aug 1999 17:20:25 -0400 (EDT) (envelope-from tgl@sss.pgh.pa.us) Received: from sss.sss.pgh.pa.us (localhost [127.0.0.1]) by sss.sss.pgh.pa.us (8.9.1/8.9.1) with ESMTP id RAA07650; Fri, 27 Aug 1999 17:19:18 -0400 (EDT) To: Tim Pizey cc: pgsql-sql@postgresql.org, bloomsbury_development@messageboards.paneris.org Subject: Re: [SQL] Newbie dbadmin out of his league In-reply-to: Your message of Fri, 27 Aug 1999 21:12:32 +0100 <3.0.3.32.19990827211232.006f4c40@popmail.dircon.co.uk> Date: Fri, 27 Aug 1999 17:19:18 -0400 Message-ID: <7648.935788758@sss.pgh.pa.us> From: Tom Lane Tim Pizey writes: > 2. How do I use COPY? > Presumably I again have to take control of the key values, > drop the index > copy from tab delimited file containing hard ids > create id sequence > create index > modify id definition Yup, that's about what you need to do. You can leave the "DEFAULT" clause where it is, since it won't be invoked during a COPY that's supplying non-default values for the ID column. (A good thing too, since I don't think we support ALTER TABLE ADD DEFAULT...) Just create the sequence with the right starting value (one past last ID being loaded). You can do that before or after the COPY, doesn't matter. A COPY will be way faster than a series of INSERT commands, especially if each INSERT is invoking a nextval(). nextval() is great for serializing live updates but it's pretty inefficient for a bulk-loading situation. Creating the index after the load is reputed to be faster than building it incrementally, as well. BTW, I believe pg_dump gets this right, so you could look at the script generated by pg_dump of a small sample table for details. regards, tom lane