On Wed, 2010-08-11 at 15:53 +0300, Heikki Linnakangas wrote:
> On 11/08/10 14:44, Simon Riggs wrote:
> > On Wed, 2010-08-11 at 13:25 +0300, Heikki Linnakangas wrote:
> >
> >> I concur that Boxuan's suggested "difficult" approach seems like the
> >> right one.
> >
> > Right, but you've completely ignored my proposal: lets do this in two
> > pieces. Get what we have now ready to commit, then add support for
> > partitioning later, as a second project.
>
> It seems like a pretty serious omission. What would you do, thrown a
> "MERGE to inherited tables not implemented" error?
It's not a "serious omission" to do work in multiple phases. I have not
proposed that we neglect that work, only that it happens afterwards.
Phasing work often allows the whole to be delivered quicker and it
reduces the risk that we end up with nothing at all or spaghetti code
through rushing things.
We have already split MERGE into two phases from its original scope,
where the majority thought for many years that MERGE without concurrent
locking was unacceptable. Splitting MERGE into 3 phases now is hardly an
earth shaking proposal.
-- Simon Riggs www.2ndQuadrant.comPostgreSQL Development, 24x7 Support, Training and Services