On Tue, Jan 30, 2018 at 1:21 AM, Thomas Munro
<thomas.munro@enterprisedb.com> wrote:
> About the idea: it makes some kind of sense to me that we should lock
> the underlying table, in all the same cases that you could do DML on
> the view automatically. I wonder if this is a problem for the
> soundness: "Tables appearing in a subquery are ignored and not
> locked."
Yeah, that seems like a pretty bad idea. It's exposing what is
basically an implementation detail to users. I think that if we
change the rules for which subqueries get flattened in a future
release, then the behavior will also change. That seems bad.
I also think that this is a bad idea for another reason, which is that
it leaves us with no syntax to say that you want to lock the view
itself, and pg_dump wants do that if only we had syntax for it.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company