Date: Wed, 20 Oct 2004 09:56:03 -0700 (PDT)
From: Stephan Szabo <sszabo@megazone.bigpanda.com>
To: Markus Bertheau <twanger@bluetwanger.de>
Cc: pgsql-sql@postgresql.org
Subject: Re: time of constraint checking
In-Reply-To: <1098287892.2649.15.camel@dicaprio.akademie1.de>
Message-ID: <20041020095039.N4154@megazone.bigpanda.com>
References: <1098287892.2649.15.camel@dicaprio.akademie1.de>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Wed, 20 Oct 2004, Markus Bertheau wrote:

> http://www.postgresql.org/docs/current/static/sql-createtable.html says,
> down at the explanation of DEFERRABLE, that constraints are checked
> after every command. Why does the following not work then:
>
> CREATE TABLE foo (
>     pos INT UNIQUE
> );
>
> INSERT INTO foo (pos) VALUES (1);
> INSERT INTO foo (pos) VALUES (2);
>
> UPDATE foo SET pos = CASE WHEN pos = 2 THEN 1 ELSE 2 END;
> ERROR:  duplicate key violates unique constraint "foo_pos_key"

Unique constraints are currently non-compliant in that it checks on each
changed row rather than at the end of the statement. I thought this was
mentioned somewhere, but it might make sense to mention it in the
compatibility section.

> Also, are deferrable constraints other that FK constraints in the works?

I don't think anyone's looking at it right now.