From: Forest Wilkinson <lyris-pg@tibit.com>
To: pgsql-sql@postgresql.org
Subject: Empty queries guaranteed to work?
Date: Tue, 27 May 2003 15:53:26 -0700
Message-ID: <41r7dvgk50cs46thfsk7shdbg9tbjeninq@4ax.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: quoted-printable

Tom Lane mentioned in this post that an empty query can be sent to the
server to determine whether the connection is still good:

http://archives.postgresql.org/pgsql-hackers/2001-10/msg00643.php

Is a query of "" guaranteed to work as long as the connection is good?
What about ";" or " "?

Background:

I am maintaining some postgres client code (C++/libpq) that, during
exception handling for a failed query, must determine whether the
database connection is still good.  This check is currently done by
sending a "select version()" query and checking the result.  However,
even that simple query fails with PGRES_FATAL_ERROR when the
connection is still good, if executed in an aborted transaction.  (I
have no idea why in the world a fatal error would be reported, when
the connection is perfectly good and a rollback is all that's needed.)
I need to be able to distinguish this situation from a real fatal
error, and PQstatus() appears to be unreliable for this purpose.  A ""
query that returns PGRES_EMPTY_QUERY seems to be a good indicator that
the connection is good, even within aborted transactions.  Now I just
need to know whether this is documented and guaranteed to work.

Suggestions of alternative methods are welcome.