Kevin Grittner escribió:
> I'm OK with that approach, and in the absence of anyone pushing for
> another direction, will make that change to pg_dump. I'm thinking
> I would only do this for materialized views which were not
> scannable, but which cause REFRESH failures on other materialized
> views if not refreshed first (recursively evaluated), rather than
> just automatically refreshing all MVs on restore. The reason this
> seems important is that some MVs may take a long time to refresh,
> and a user might want a dump/restore to get to a point where they
> can use the rest of the database while building the contents of
> some MVs in the background or during off hours.
Maybe it would be a good idea to try to put such commands at the very
end of the dump, if possible.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services