Re: pg_dump versus materialized views

* Tom Lane (tgl(at)sss(dot)pgh(dot)pa(dot)us) wrote:
> I've thought for some time that, given that it can't reproduce the MV
> states exactly, pg_dump shouldn't even try. I think it would be more
> useful to have two operating modes selectable by command line switch:
> refresh all matviews, or refresh none of them.

I'm not convinced that adding a new switch for it is actually a great
idea, and it's pretty late to be doing so anyway, imv.. I'd vote for
just "don't refresh mat views" in pg_dump for this go-round.

> Or maybe there's some other better way to approach it.

My preference would be for pg_dump to actually reproduce what the
current state of the database is, possibly by dumping the exact contents
of the matview as if it was a regular relation, but any such discussion
should be post-9.3.

Thanks,

Stephen