From: Tom Lane <tgl@sss.pgh.pa.us>
To: Sebastien FLAESCH <sf@4js.com>
cc: pgsql-interfaces@lists.postgresql.org
Subject: Re: Type scale returned by PQfmod() 65531 for time/timestamp output
 parameter?
In-reply-to: <093e1214-7057-f80c-5c3c-74076b7bb555@4js.com>
References: <093e1214-7057-f80c-5c3c-74076b7bb555@4js.com>
Comments: In-reply-to Sebastien FLAESCH <sf@4js.com>
	message dated "Thu, 15 Feb 2018 16:03:22 +0100"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <7536.1518711940.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 15 Feb 2018 11:25:40 -0500
Message-ID: <7537.1518711940@sss.pgh.pa.us>
Precedence: bulk

Sebastien FLAESCH <sf@4js.com> writes:
> When using a stored function returning output parameters defined with as=
 time[(n)]
> or timestamp[(n)], PQfmod() always returns the same SCALE (65531/0xFFFB)=
, no matter
> what time/timestamp precision is used.

Your misunderstanding is in assuming that typmod decoration on function
parameters means anything.  It doesn't; the function is effectively just
declared as taking or returning plain time or timestamp.

Perhaps someday that will change, but it'd be a big task with a lot of
hard decisions to make.  For instance, do we allow creation of both
f(time(2)) and f(time(4)), and if so what's the rule for choosing which
one to call?

			regards, tom lane