Re: Hide 'Execution time' in EXPLAIN (COSTS OFF)

Le jeudi 16 octobre 2014 10:43:25 Robert Haas a écrit :
> On Thu, Oct 16, 2014 at 10:06 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> > Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> >> On Tue, Oct 14, 2014 at 3:03 AM, David Rowley <dgrowleyml(at)gmail(dot)com>
wrote:
> >>> Hmm, was my case above not compelling enough?
> >>
> >> Apparently not to Tom, but it made sense to me.
> >
> > No, it wasn't. I'm not convinced either that that patch will get in at
> > all, or that it has to have regression tests of that particular form,
> > or that such a switch would be sufficient to make such tests platform
> > independent.
>
> People clearly want to be able to run EXPLAIN (ANALYZE) and get stable
> output. If the proposed change isn't enough to make that happen, we
> need to do more, not give up. Regardless of what happens to inner
> join removal.

From my point of view as a FDW implementor, the feature I need is to have
EXPLAIN (COSTS ON) with stable output for foreign scan nodes.

In the Multicorn FDW (Python API on top of the C-API), we introduced this
commit to make the tests pass on 9.4:

https://github.com/Kozea/Multicorn/commit/76decb360b822b57bf322892ed6c504ba44a8b28

Clearly, we've lost the ability to test that the costs as set from the Python
API are indeed used.

But I agree that it would be better to have more flexibility in the regression
framework itself.

If this use case is too marginal to warrant such a change, I'll keep the tests
as they are now.

--
Ronan Dunklau
http://dalibo.com - http://dalibo.org