Craig Ringer wrote:
> On 11/05/2014 05:01 AM, Petr Jelinek wrote:
> > I guess I could port BDR sequences to this if it would help (once we
> > have bit more solid agreement that the proposed API at least
> > theoretically seems ok so that I don't have to rewrite it 10 times if at
> > all possible).
>
> Because the BDR sequences rely on all the other BDR machinery I suspect
> that'd be a pretty big thing to review and follow for someone who
> doesn't know the BDR code.
>
> Do you think it'd be simple to provide a blocking, transactional
> sequence allocator via this API? i.e. gapless sequences, much the same
> as typically implemented with SELECT ... FOR UPDATE on a counter table.
I think that would be a useful contrib module too; gapless allocation is
a frequent user need. If we can build it in some reasonable way with the
seqam API, it'd show the API is good.
But on the other hand, since BDR sequences are the ultimate goal of this
patch, we need to make sure that they can be made to work with whatever
the seqam API ends up being. It's not necessary for reviewers here to
review that code, but the BDR developers need to say "yes, this API
works for us" -- just like the slony developers (well, ssinger) did for
logical decoding.
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services