On Fri, Dec 23, 2022 at 05:51:28PM +0900, Michael Paquier wrote:
> On Thu, Dec 22, 2022 at 09:27:24PM -0600, Justin Pryzby wrote:
> > This would do a single seqscan:
> > SELECT format('ALTER TABLE %I ALTER COLUMN %I TYPE TEXT',
> > attrelid::regclass, attname) FROM pg_attribute WHERE
> > atttypid='aclitem'::regtype; -- AND ...
> > \gexec
>
> FWIW, I find the use of a FOR loop with a DO block much cleaner to
> follow in this context, so something like the attached would be able
> to group the two queries and address your point on O(N^2). Do you
> like that?
LGTM. Thanks.
--
Justin