> On Thu, Dec 1, 2016 at 9:40 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>>
>> OK, then my vote is to do it that way for now.
Thanks for your opinion. That's fine with me.
> Am 02.12.2016 um 07:22 schrieb Amit Kapila <amit.kapila16@gmail.com>:
> Done that way in attached patch.
Did a quick review: The patch applies cleanly against current head. make installcheck with force_parallel_mode =
regresspasses all tests. My manual tests show that parallel query is working for prepared statements in SQL with
PREPAREand EXECUTE. CREATE TABLE AS EXECUTE is working, EXPLAIN on that shows a parallel plan, EXPLAIN ANALZE indicates
0launched workers for that. Looks fine so far!
You should however include a sentence in the documentation on that parallel plan w/o workers corner-case behaviour.
Feelfree to take that from my patch or phase a better wording.
And again my question regarding back patching to 9.6:
- 9.6 is currently broken as Laurenz showed in [1]
- 9.6 does not have documented that SQL PREPARE prepared statements cannot not use parallel query
The former could be fixed by back patching the full patch which would void the latter. Or it could be fixed by
disablinggeneration of parallel plans in extended query protocol prepare. Alternatively only the change in execMain.c
couldbe back patched. In these cases we would need to have the a separate wording for the 9.6 docs.
Best regards,
Tobias
[1] A737B7A37273E048B164557ADEF4A58B539990D0@ntex2010i.host.magwien.gv.at