Content-Type: text/plain;
  charset="iso-8859-2"
From: Richard Huxton <dev@archonet.com>
Organization: Archonet Ltd
To: Tomasz Myrta <jasiek@klaster.net>
Subject: Re: virtual table
Date: Mon, 23 Jun 2003 09:54:54 +0100
User-Agent: KMail/1.4.3
Cc: PGSQL-SQL <pgsql-sql@postgresql.org>
References: <3EF61E44.2020601@klaster.net>
	<200306230929.24372.dev@archonet.com>
	<3EF6BCB4.2040209@klaster.net>
In-Reply-To: <3EF6BCB4.2040209@klaster.net>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Message-Id: <200306230954.54185.dev@archonet.com>

On Monday 23 Jun 2003 9:39 am, Tomasz Myrta wrote:
> Dnia 2003-06-23 10:29, U=BFytkownik Richard Huxton napisa=B3:
> > Look into views - you'll need to provide triggers to handle the
> > update/inserts.
>
> I think view won't change too much - there is not too much difference for
> this case between creating view and empty table with trigger returning
> null. I was thinking about something more comfortably - table without
> pre-declared columns so I don't have to change table definition when
> pl/pgsql function changes.
>
> Can anyone tell me how triggers work? Do they need to look into table
> column definition to retrieve data?

Hmm - I think your problem is going to come before that. Any time you do an=
=20
INSERT, PostgreSQL is going to need to know the types of all the columns=20
involved.

For this sort of thing, I try to keep all the related bits (initial functio=
n,=20
views, triggers) in the same text-file to encourage me to remember to updat=
e=20
them all together.

It sounds like this table1 is only being used during transfer. Is there any=
=20
reason why you aren't just inserting the required rows into table2.

--=20
  Richard Huxton