From: Bruce Momjian <pgman@candle.pha.pa.us>
Message-Id: <200303201804.h2KI43b23863@candle.pha.pa.us>
Subject: Re: I want to send comments to the backend!
In-Reply-To: <24626.1048181454@sss.pgh.pa.us>
To: Tom Lane <tgl@sss.pgh.pa.us>
Date: Thu, 20 Mar 2003 13:04:03 -0500 (EST)
Cc: =?ISO-8859-1?Q?Gerhard_H=E4ring?= <lists@ghaering.de>,
	pgsql-interfaces@postgresql.org
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=US-ASCII

Tom Lane wrote:
> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> > Yes, I see that now.  res is NULL and PQresultStatus() is returning
> > PGRES_NONFATAL_ERROR.  What is strange is that "" and "\n" both return
> > an OK result, so I am trying to figure out why comments are different.
> 
> The backend special-cases an input string that contains only whitespace
> (look in postgres.c).  The special case doesn't apply here, though.
> 
> >> Perhaps PGRES_EMPTY_QUERY would be more appropriate?  Not sure where in
> >> the chain would be best to change the behavior, though.
> 
> > Yes.
> 
> Well, we could change the backend --- but that would be a nontrivial
> change, and to tell you the truth I think the special response for empty
> query is a wart on the protocol anyhow.  In the long run I'd rather get
> rid of it.

Yes, I see that now:

	if (strspn(parser_input->data, " \t\r\n") == parser_input->len)

> Or we could change PQexec to return a PGRES_EMPTY_QUERY result if it
> gets nothing back from the backend except ReadyForQuery.  Question is,
> does that create the possibility of masking error conditions?
> 
> Or we could just change PQresultStatus to return PGRES_EMPTY_QUERY for
> a null input.  Attractive 'cause it's a one-line change, but I think it
> really does create the possibility of masking errors --- application
> programmer errors, mostly.

Could we change the backend to return NullCommand() if the query string
was empty, rather than checking for whitespace directly?  I thought the
whitespace test was there only as a performance hack.  In fact, the
comment in postgres.c says that:
 
                     * if there is nothing in the input buffer, don't
                     * bother trying to parse and execute anything; just
                     * send back a quick NullCommand response.

In pg_exec_query_string(), if nothing is processed, we can do the call
there. or course assuming destination is correct?

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073