On Thu, Feb 28, 2019 at 10:13:17AM -0500, Tom Lane wrote:
> Yeah, I think it's just because we won't search the pg_temp schema
> for function or operator names, unless the calling SQL command
> explicitly writes "pg_temp.foo(...)" or equivalent. That's an
> ancient security decision, which we're unlikely to undo. It
> certainly puts a crimp in the usefulness of putting extensions into
> pg_temp, but I don't think it totally destroys the usefulness.
> You could still use an extension to package, say, the definitions
> of a bunch of temp tables and views that you need to create often.
Even with that, it should still be possible to enforce search_path
within the extension script to allow such objects to be created
correctly, no? That would be a bit hacky, still for the purpose of
temp object handling that looks kind of enough to live with when
creating an extension.
--
Michael