On Tue, 2011-02-08 at 09:10 -0800, Jeff Davis wrote:
> On Mon, 2011-02-07 at 18:23 +0100, Dimitri Fontaine wrote:
> > I would think
> > CREATE TYPE foo AS RANGE (bar) USING (btree_ops);
> > The USING clause is optional, because you generally have a default btree
> > opclass for the datatype.
> There are other options, like "CANONICAL", so where do those fit?
> If CREATE TYPE already has an options list, it seems a little strange to
> add grammar to support this feature. "USING" doesn't seem to mean a lot,
> except that we happen to use it in other contexts to mean "operator
For the user-facing part, how about just passing it as a parameter
called "SUBTYPE_OPCLASS"? It sounds a little on the "internal detail"
side, but so do some other type definition parameters.
As for the catalog, I'm inclined to leave the compare function in there
directly and just add a dependency on the opclass. That way, it's only
one syscache lookup rather than two, to get the compare function oid.
Then again, perhaps that doesn't matter anyway. Thoughts?
In response to
pgsql-hackers by date
|Next:||From: Jeff Davis||Date: 2011-02-09 05:50:50|
|Subject: Range Type constructors|
|Previous:||From: Dan Ports||Date: 2011-02-09 05:24:19|
|Subject: Re: SSI patch version 14|