Tom Lane wrote:
> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> > Yes, I realize that, but when I create an index on a temp table, I can
> > create it even though someone else tries to do the same in another
> > session. If these views on temp tables go away on session exit, and
> > can't be reliably accessed by other users, they should be in the temp
> > schema and therefore invisible to other users and to prevent name
> > conflicts.
>
> I think we should provide the *facility* for temp views; that doesn't
> equate to feeling that we must have an enforcement mechanism to prevent
> you from using a non-temp view on a temp table. The enforcement
> mechanism would be notably more work to write and would slow down the
> creation of views (at least non-temp ones), in order to achieve what?
> Not much that I can see. Admittedly, it's a bit silly to use a
> non-temp view with a temp table, but I don't think the system needs to
> go out of its way to prevent silliness.
My feeling is that either the view is temporary, fully, or it isn't. I
don't see having it in the public namespace _and_ removing it on session
exit as defensible. I would like to update the TODO list to say:
Place views on temporary tables in temporary namespaces
It may be difficult, but I clearly think it is a bug if we don't do it.
-- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610)
359-1001+ If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square,
Pennsylvania19073