From: Rod Taylor <pg@rbt.ca>
To: Richard Broersma Jr <rabroersma@yahoo.com>
Cc: Michael Glaesemann <grzm@seespotcode.net>,
	General Postgresql List <pgsql-sql@postgresql.org>
In-Reply-To: <20060701014108.79890.qmail@web31811.mail.mud.yahoo.com>
References: <20060701014108.79890.qmail@web31811.mail.mud.yahoo.com>
Date: Sat, 01 Jul 2006 13:06:31 -0400
Message-Id: <1151773591.673.87.camel@home>
Mime-Version: 1.0
Subject: Re: Alternative to Select in table check constraint
Content-Type: text/plain
Content-Transfer-Encoding: 7bit

On Fri, 2006-06-30 at 18:41 -0700, Richard Broersma Jr wrote:
> > >         CHECK   ( 1 = ALL (     SELECT COUNT(STATUS)
> > >                                 FROM BADGES
> > >                                 WHERE STATUS = 'A'
> > >                                 GROUP BY EMPNO))
> > 
> >  From the PostgreSQL docs (http://www.postgresql.org/docs/8.1/ 
> > interactive/sql-createtable.html)
> > 
> > CREATE UNIQUE INDEX one_a_badge_per_employee_idx
> > ON badges (empno)
> > WHERE status = 'A';
> > http://www.postgresql.org/docs/8.1/interactive/sql-createindex.html
> > http://www.postgresql.org/docs/8.1/interactive/indexes-partial.html
> 
> Michael,
> 
> Partial indexs seem to be "what the doctor ordered!"   And your suggest is right on, the idea of
> the constraint is to allow only one active badge status at a time.
> 
> But now that I think about it, using the authors suggestion (if it actually worked), how would
> would it be possible to change the active status from one badge to another?

Unset the status first then set on the new one. Same transaction of
course.

You may find this type of constraint is more workable with a TRIGGER
deferred until commit time than a unique constraint which cannot (at
this time in PostgreSQL) be deferred.
--