Re: pgsql: Separate block sampling functions

On Fri, May 15, 2015 at 12:22 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Michael Paquier <michael(dot)paquier(at)gmail(dot)com> writes:
>> On Fri, May 15, 2015 at 12:03 PM, Simon Riggs <simon(at)2ndquadrant(dot)com> wrote:
>>> On 15 May 2015 at 03:50, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>>>> Simon Riggs <simon(at)2ndQuadrant(dot)com> writes:
>>>>> Separate block sampling functions
>
>>>> This patch broke buildfarm member crake.
>
>>> OK, thanks. I missed that amongst the other unrelated failures. Looking now.
>
>> This needs a patch to file_text_array_fdw which I think is available here:
>> https://github.com/adunstan/file_text_array_fdw
>> Simon, do you mind if I send a pull request?
>
> TBH, I think that this patch itself was a bad idea and should be reverted.
> I don't object to changing APIs used by external modules when there's a
> good reason to break them, but having looked at this patch all I see is
> change for the sake of change. What new functionality have you introduced?

If you look at the TABLESAMPLE patch, separating the block sampling
into a separate facility makes quite some sense.

> Or to put it more baldly: it's likely that you've broken quite a large
> number of third-party FDWs, not just this one. A lot of people have
> probably copied-and-pasted what was in the contrib FDWs.

make_foreignscan() has been changed as well by 1a8a4e5c..
--
Michael