On Wed, 2011-08-03 at 13:07 -0400, Robert Haas wrote:
> A little OT here, but (as I think Simon said elsewhere) I think we
> really ought to be considering the table statistics when deciding
> whether or not to replan. It seems to me that the overwhelmingly
> common case where this is going to come up is when (some subset of)
> the MCVs require a different plan than run-of-the-mill values. It
> would be nice to somehow work that out.
That blurs the line a little bit. It sounds like this might be described
as "incremental planning", and perhaps that's a good way to think about
it.
Regards,Jeff Davis