Message-Id: <36025ECD.1DC5CE75@insightdist.com>
Date: Fri, 18 Sep 1998 09:23:26 -0400
From: Byron Nikolaidis <byronn@insightdist.com>
Mime-Version: 1.0
To: "Billy G. Allie" <Bill.Allie@mug.org>
Cc: pgsql-interfaces@postgreSQL.org, pgsql-hackers@postgreSQL.org
Subject: Re: [HACKERS] Re: [INTERFACES] Odbc parser error
References: <199809172337.TAA25897@bgalli.mug.org>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit


Billy G. Allie wrote:

> How about subsitution any occurance of /=[ \t]*?/ with "IS NULL".  This should
> work for both select and update statements.
> --

Yes, sure, it could be done, assuming I could get a regex library for VC++.  But I
think it would be more of a kludge than a solution.  Even the current substitution
of the ? with the parameter data is a kludge.  The frontend shouldn't be rewriting
the sql string to stick parameters in.

The real point of the whole thing is that if the backend recognized parameters and
there was a protocol for passing the data over, we would be able to do some cool
things we have never been able to do.  Not just passing NULL, but also passing
larger parameter data without worrying about running into the upper limit of the
statement string.  It would also be easier to send binary data.  And, assuming the
protocol allowed it, you could retrieve information from the backend about a
particular parameter, such as its precision, data type, nullablity, etc.

And its not just ODBC that would benefit.  JDBC has the same semantics.  And I
believe Embedded SQL does too.

Byron