To: pgsql-sql@postgresql.org
From: "Stuart" <smcg2297@frii.com>
Subject: Re: data dependent sequences?
Date: Tue, 17 Jul 2007 07:34:26 -0600
Lines: 31
Message-ID: <f7iggu$1nb$1@sea.gmane.org>
References: <AKEFJAHAPDBEDKIICNJKMEJPCDAA.smcg2297@frii.com>
	<1184515497.5778.141.camel@localhost.localdomain>
Mime-Version: 1.0
Content-Type: text/plain;
	charset="iso-2022-jp"
Content-Transfer-Encoding: 7bit


"Ragnar" <gnari@hive.is> wrote in message news:1184515497.5778.141.camel@localhost.localdomain...
> On sun, 2007-07-15 at 09:28 -0600, Stuart McGraw wrote:
> > Advice requested :-)  I have a table like:
> > 
> > CREATE TABLE items (
> >     id INT,
> >     typ INT    ...
> >     PRIMAY KEY (seq,typ));
> > 
> > I would like 'id' to be like a SERIAL except that I
> > want independent sequences for each value of 'typ'.
> 
> what possible advantage could there be to that?
> 
> if you need gapless series, then sequences (and serial)
> are not adequate anyways.
> 
> just use one sequence.

I am not looking for gapless sequences.  The reason I
want to do this is the "typ" column is actually an indicator
of the source of the rest of the infomation in the row.
The "rules" for assigning the id number vary depending
on the source -- in some cases they start at 1 and increment
by one, in other cases they start at, say, 1000000 and 
increment by 10.  There are a lot existing data using these 
rules and I cannot change that.  I can of course have the 
application do the assignments, but in general eould prefer 
to push this down into the database if posible.