Re: disabling OIDs?

I don't know why they use OID's for cursors. But I do know that if you
run a trace the SQL that creates the cursor uses OID's,
so it doesn't work if the table is created without OID's. Also, if you
want to have updateable cursors against views (i.e., a view with rules
for INSERT, UPDATE, and DELETE), you must name the OID and CTID as
columns in the view. Again, we learned this the hard way. Mark Dexter

-----Original Message-----
From: gsstark(at)mit(dot)edu [mailto:gsstark(at)mit(dot)edu]
Sent: Monday, December 13, 2004 12:06 PM
To: Mark Dexter
Cc: pgsql-general(at)postgresql(dot)org; gsstark(at)mit(dot)edu
Subject: Re: disabling OIDs?

"Mark Dexter" <MDEXTER(at)dexterchaney(dot)com> writes:

> > For what it's worth, OIDs are required if you ever want to use
> > updateable cursors with the ODBC driver. We discovered this the
> > hard way. Mark Dexter

That's unfortunate. Is it because it's difficult to track down the
primary key of the table? Is it any easier to track down the primary key
of the table in 8.0? It would be much better if it checked the primary
key and used that instead of OIDs.

Though I'm unclear implementing "updateable cursors" in the client-end
is really a good idea. I suppose it's nice if you understand the
limitations inherent.

--
greg