Dropping OID column

On Thu, 2003-01-16 at 13:41, Rod Taylor wrote:

ALTER TABLE .. SET WITHOUT OIDS;

I'd prefer this, as it's more similar to the CREATE TABLE syntax.

Cheers,

Neil
--
Neil Conway <neilc@samurai.com> || PGP Key ID: DB3C29FC

Rod Taylor <rbt@rbt.ca> writes:

What is the preferred syntax:
ALTER TABLE .. DROP COLUMN oid;
or
ALTER TABLE .. SET WITHOUT OIDS;

If we ever got around to supporting the inverse function (add oids),
I do not think we'd want to spell it like "ADD COLUMN oid" --- that
would interfere with making a plain user column named "oid", which was
one of the reasons why people wanted to be able to drop OIDs in the
first place.

So I lean towards the SET WITH/WITHOUT OIDS syntax. That keeps it
clearly out of the domain of user columns.

regards, tom lane

Tom Lane writes:

If we ever got around to supporting the inverse function (add oids),
I do not think we'd want to spell it like "ADD COLUMN oid" --- that
would interfere with making a plain user column named "oid", which was
one of the reasons why people wanted to be able to drop OIDs in the
first place.

Is there really a reason why OIDs can't be a pure user-space column?

--
Peter Eisentraut peter_e@gmx.net

Peter Eisentraut <peter_e@gmx.net> writes:

Is there really a reason why OIDs can't be a pure user-space column?

We'd need some kluge to keep "SELECT *" from including OID. Also I'd be
a bit worried about the impact on the cost of HeapTupleGetOid --- it
might not matter, or it might.

regards, tom lane

I guess I'd prefer the DROP COLUMN syntax. It means we don't have to add
another non-standard command, and people can figure out how to drop the oid
column just by trying...

Chris