Received: from sss.sss.pgh.pa.us (sss.pgh.pa.us [209.114.166.2])
	by hub.org (8.9.3/8.9.3) with ESMTP id UAA97691
	for <pgsql-interfaces@postgreSQL.org>;
	Mon, 16 Aug 1999 20:24:15 -0400 (EDT)
	(envelope-from tgl@sss.pgh.pa.us)
Received: from sss.sss.pgh.pa.us (localhost [127.0.0.1])
	by sss.sss.pgh.pa.us (8.9.1/8.9.1) with ESMTP id UAA06513;
	Mon, 16 Aug 1999 20:23:31 -0400 (EDT)
To: Nicolas Aragon <nyco@jet.es>
cc: pgsql-interfaces@postgreSQL.org
Subject: Re: [INTERFACES] types oids 
In-reply-to: Your message of Tue, 17 Aug 1999 01:18:35 +0200
	<19990817011835.A165@atlante>
Date: Mon, 16 Aug 1999 20:23:31 -0400
Message-ID: <6511.934849411@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Nicolas Aragon <nyco@jet.es> writes:
> But this leads me to new questions because it seems that standard types
> aren't hardcoded anywhere. pg_type tells me the name, size and some other
> information about the type. But what about the "kind" of the type (numeric,
> text, date...)?

You pretty much have to know from the name what the datatype is.
Postgres doesn't have any type categorization like that.  (Maybe it
should ... I think there are some hacks in the parser that amount to
categorizing the standard datatypes, but that knowledge is not visible
outside the parser.)

> I can only think of comparing the name with a list of names for standard
> types. Is there another simpler method? A standard relation between standard
> types and their OIDs?

In reality, the OIDs assigned to standard types like "bool" and "int4"
are unlikely ever to change; they certainly will not change faster than
once per Postgres major release.  You could probably get away with
hard-wiring the knowledge that OID 16 == bool, say.  If that seems a
little too ugly, reading the OIDs for known types during startup is
plenty good enough.

If you are dealing with user-defined types, looking up the OIDs at
startup is the way to go.  (User type OIDs could change across a
database dump/reload, so hardwiring those would be a bad idea.)

> I can even think that.... is the user (admin) allowed to change the names
> of standard types?

Not if he wants things to keep working ;-)

> Another, more trivial question: do you know where to find nice logos
> for postgres?

See the website.

			regards, tom lane