Date: Sat, 14 Oct 2006 18:11:26 -0600
From: Michael Fuhr <mike@fuhr.org>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: PostgreSQL SQL List <pgsql-sql@postgresql.org>,
	Joost Kraaijeveld <J.Kraaijeveld@Askesis.nl>
Subject: Re: Foreign key reference counting strategy?
Message-ID: <20061015001126.GA20713@winnie.fuhr.org>
References: <1160828741.28838.9.camel@panoramix>
	<4530EB84.4050904@logix-tt.com>
	<20061014234844.GA20342@winnie.fuhr.org>
	<13525.1160870286@sss.pgh.pa.us>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <13525.1160870286@sss.pgh.pa.us>

On Sat, Oct 14, 2006 at 07:58:06PM -0400, Tom Lane wrote:
> Michael Fuhr <mike@fuhr.org> writes:
> > On Sat, Oct 14, 2006 at 03:52:04PM +0200, Markus Schaber wrote:
> >> Create an "after delete" trigger on the referencing table that checks
> >> whether there still are records with the same key (IF EXISTS()), and
> >> deletes the referenced row otherwise.
> 
> > In a concurrent environment that delete can fail with a foreign key
> > constraint violation because IF EXISTS won't see uncommitted changes
> > in other transactions.
> 
> No, I don't think so, because the DELETE will already be holding
> exclusive lock on the doomed PK row, which any would-be inserters of
> matching FK rows will be blocked on.  AFAICS the DELETE should go
> through and then the inserters will fail.

Unless the inserters got there first.  I just tested both ways; if
the insert acquires the lock first then the delete fails, but if the
delete acquires the lock first then the insert fails.

-- 
Michael Fuhr