From: "Josh Berkus" <josh@agliodbs.com>
Subject: Re: Constraint problem
To: Stephan Szabo <sszabo@megazone23.bigpanda.com>,
	Josh Berkus <josh@agliodbs.com>
Cc: <pgsql-sql@postgresql.org>
Date: Thu, 16 May 2002 21:40:59 -0700
Message-ID: <web-1462528@davinci.ethosmedia.com>
In-Reply-To: <20020516175335.N1443-100000@megazone23.bigpanda.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit

Stephan,

> AFAIK we don't actually have statement triggers right now, anyway,
> but...
> Actually, I guess an after trigger that just checked the new values
> would
> work as well, right?  See if there exists more than one row with the
> new
> key value.  The after trigger should fire after all the changes are
> done
> and in the worst case you could use constraint triggers and
> deferment.
> The hard part about this really is making it work with concurrent
> modifications, and if you're willing to forgo that, it shouldn't be
> too
> bad.

Thanks.  I'll give your solution a try and see what the performance hit
is.  

Is the Unique Constraint issue on the To-Do list?

Chris:  The "holding value" solution won't work for me.  I'm usually
changing dozens of values at a time, and they need to stay in the same
sequence while blocks of them move on the scale.

FYI, I'm implementing Joe Celko's "Linear Nested Set Model" of tree
structures.

-Josh Berkus