From: | Mike Nolan <nolan(at)gw(dot)tssi(dot)com> |
---|---|
To: | pgsql-general(at)postgresql(dot)org (pgsql general list) |
Subject: | Re: guaranteeing that a sequence never skips (fwd) |
Date: | 2004-10-03 17:48:48 |
Message-ID: | 200410031748.i93HmmOX019531@gw.tssi.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
> On Sun, 2004-10-03 at 08:58, David Garamond wrote:
> > Am I correct to assume that SERIAL does not guarantee that a sequence
> > won't skip (e.g. one successful INSERT gets 32 and the next might be 34)?
> >
> > Sometimes a business requirement is that a serial sequence never skips,
> > e.g. when generating invoice/ticket/formal letter numbers. Would an
> > INSERT INTO t (id, ...) VALUES (SELECT MAX(col)+1 FROM t, ...) suffice,
> > or must I install a trigger too to do additional checking?
>
> You will have to lock the whole table and your parallel performance will
> be poor.
Locking the table isn't sufficient to guarantee that a sequence value
never skips. What if a transaction fails and has to be rolled back?
I've written database systems that used pre-numbered checks, what's usually
necessary is to postpone the check-numbering phase until the number of
checks is finalized, so that there's not much chance of anything else
causing a rollback.
--
Mike Nolan
From | Date | Subject | |
---|---|---|---|
Next Message | Uwe C. Schroeder | 2004-10-03 18:01:46 | Re: guaranteeing that a sequence never skips |
Previous Message | Scott Marlowe | 2004-10-03 17:24:28 | Re: [HACKERS] OT moving from MS SQL to PostgreSQL |