Sender: root@sonnet1.sonnet.com
Message-ID: <37A0E659.5F423AA8@sonnet.com>
Date: Thu, 29 Jul 1999 16:40:09 -0700
From: Brian Curnow <bcurnow@sonnet.com>
MIME-Version: 1.0
To: pgsql-interfaces@postgresql.org
Subject: perl5 Should consumeInput obliterate unretrieved Notifys?
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Hello,

I've got a Perl script that makes the following series of calls:

conn->exec(SELECT
...
conn->exec(UPDATE
...
$conn->socket
select (..) [with short timeout]
conn->consumeInput
conn->notifies

What I am seeing is that if a Notify arrives between the SELECT and
UPDATE, the exec(UPDATE) call consumes the notify, and when the next
consumeInput is eventually called, the notify disappears.  The easy fix
is obviously to call notifies() before even starting the select(), but I
wonder if the loss of notify is correct.

The docs for libpq state that you should call notifies() after each
PQexec or PQgetResult, but I figured that was because a future select()
would not catch the notify since PQexec already had read it, and you
might wait (possibly forever) without checking.  Further up in the
documentation, it states

"Once a notification is returned from PQnotifies, it is considered
handled and will be removed from the list of notifications."

which would imply the loss of the notify is incorrect.  perldoc Pg
doesn't say anything contradictory.

Does anyone have thoughts on this?

Brian Curnow