Re: Confusing docs about GetForeignUpperPaths in fdwhandler.sgml

On 2016/11/03 23:39, Robert Haas wrote:
> On Wed, Oct 26, 2016 at 3:18 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> A larger issue is that I think the API itself is poorly designed, as
>> I stated awhile ago (<31706(dot)1457547166(at)sss(dot)pgh(dot)pa(dot)us>)

I agree on that point. I plan to rewrite direct modify using upper
planner path-ification; I think we would no longer need the planner API
PlanDirectModify, but I expect the executor/explain APIs (ie,
BeginDirectModify, IterateDirectModify, EndDirectModify, and
ExplainDirectModify) would be still useful after that rewrite. Before
that, however, I'd like to work on extend postgres_fdw so as to handle
more cases such as UPDATE/DELETE on a join and INSERT, with the existing
API.

> I'd really be quite happy to see you take a more active hand
> in the FDW discussions; clearly, you've got a lot of understanding of
> that area that is pretty much unique to you. I'd even support an
> effort to rewrite the work that has already been done in a form more
> to your liking, but I think you'd need to actually pay attention to
> the threads on a somewhat regular basis in order that to be practical.

+1

Best regards,
Etsuro Fujita