Neil Conway <nconway@klamath.dyndns.org> writes:
> Currently, PostgreSQL allows this -- when the session ends and the temp
> table is dropped, an subsequent queries on the view fail. Is this the
> optimal behavior?
Well, I think it's better than refusing views on temp tables, as the
spec would have us do.
The "correct" behavior is probably to drop such views on backend exit.
Possibly we should invent the notion of temp views, and disallow
references from non-temp views to temp tables. That seems like it
might be less likely to cause unpleasant surprises than just silently
dropping views that reference temp tables.
In any case I'd say this is something best tackled in the context of
generalized reference tracking ... which is something we know we need,
but no one's stepped up to make it happen yet. I don't think this
particular problem is bad enough to warrant a special-purpose
implementation mechanism.
regards, tom lane