From: Bruce Momjian <maillist@candle.pha.pa.us>
Message-Id: <199808261638.MAA15502@candle.pha.pa.us>
Subject: Re: [INTERFACES] new odbc fails
In-Reply-To: <35E40FC4.A0F5A3B3@insightdist.com> from Byron Nikolaidis at "Aug
	26, 1998  9:38:13 am"
To: byronn@insightdist.com (Byron Nikolaidis)
Date: Wed, 26 Aug 1998 12:38:46 -0400 (EDT)
Cc: john.boris@onlinesvc.com, pgsql-interfaces@postgreSQL.org,
	daveh@insightdist.com
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

> 
> 
> John J. Boris, Sr. wrote:
> 
> > I downloaded and installed the new odbc driver. I am running WIndows NT 4.0
> > and my PostreSQL is on a LINUX box. I can connect to the database using MS
> > Access 97, link the tables, it allows me to select unique keys for each
> > table but when I try to view a table I get an ODBC failed message and the
> > table is filled with NAME#.
> >
> 
> This is most likely because of the backend canonifier problem.  Check your
> psqlodbc.log (enabled with the CommLog driver option) file to see if there is
> an error message from the backend concerning "palloc failure, memory
> exhausted".
> 
> The unique key you are specifying in Access can result in "keyset" queries of
> the form:
> 
> select from table where (k1 = 1 AND k2 = 1 AND k3 = 1) OR (k1 = 1 AND k2 = 1
> AND k3 = 2) ......... (k1 = 1 and k2 = 1 and k3 = 10)
> 
> Access usually uses a rowset size of 10, thus you will have 10 groups of ORs.
> Depending on how many parts your key is made up of, your AND groupings may look
> differently than I have shown it.
> 
> Bottom line is, a fix for the backend is being worked on in two different
> capacities.  One fix, which we use here, is a patch, written by Dave Hartwig,
> which rewrites the statements into UNIONS in the backend.  This is called the
> KSQO patch (KeySet Query Optimization, enabled with the set ksqo to 'ON'
> command, which you can put in the Connect Settings driver/datasource options).
> I'm not sure if we have this patch on our website or if it will be included in
> Postgres 6.4.  You may want to check the website and inquire about it.
> 
> Also, Bruce Momjian  is trying to fix the problem in the backend.  I'm not sure
> of its current status or if it will be complete in Postgres 6.4.

I am ready to address this.  I will send an e-mail today.

-- 
Bruce Momjian                          |  830 Blythe Avenue
maillist@candle.pha.pa.us              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)