Re: counting algorithm for incremental matview maintenance

On Tue, May 14, 2013 at 3:52 PM, Kevin Grittner <kgrittn(at)ymail(dot)com> wrote:
> We could drive the triggering of incremental maintenance off of the
> dependency information which is already stored, but for performance
> we probably want to add a new pg_class flag to indicate that the
> relation is referenced by a matview definition which specifies
> incremental update. That would allow a fast path for skipping
> other tests for DML on non-referenced relations, at the expense of
> some additional catalog updates on some DDL.

I'm afraid this might require creating a matview or updating the
definition of a matview to refer to different relations to take
AccessExclusiveLock on those relations, in order to avoid SnapshotNow
problems while updating this flag for those relations, and I think
that's probably unacceptable. Some thought may be needed here to come
up with a good solution.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company