Tom Lane wrote:
> If we were willing to abuse the ALTER TABLE syntax some more, it would
> be possible to support changing the datatypes of f1 and f2
> simultaneously, thereby allowing the above to work. The
infrastructure
> for hacking multiple tables in parallel is already there in CVS tip,
> but it only gets exercised in cases involving inheritance.
Just a clarification: isn't ALTER DOMAIN the best place to do this?
IMHO, this is one of those rare cases were extending PostgreSQL beyond
the SQL spec is justified. Right now, as I understand it, the only way
to do these types of things is to bring down the RI rules for a short
time until the table manipulation is completed, which seems inelegant,
not to mention the convenience factor.
Plus, I see anything that encourages usage of domains as good thing, as
domains themselves are very good things (and quite underutilized by the
unwashed masses, I expect).
Merlin