Re: Should we add GUCs to allow partition pruning to be disabled?

Hi David.

On 2018/04/21 14:09, David Rowley wrote:
> On 20 April 2018 at 20:51, Amit Langote <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp> wrote:
>> set constraint_exclusion to off;
>>
>> -- not ok!
>
> It needed a bit more effort than I put in the first time around to
> make this work properly. constraint_exclusion = 'off' becomes a bit of
> a special case for partitioned tables now. To make this work I had to
> get rid of hasInheritedTarget and make a new enum that tracks if we're
> inheritance planning for an inheritance parent or a partitioned table.
> We can't simply only set hasInheritedTarget to true when planning with
> inheritance parents as constraint_exclusion = 'partition' must still
> know that we're planning using the inheritance planner.
>
> v2 patch attached.

Thanks for the updated patch.

Your proposed changes to inheritance_planner() look fine to me. In the
comment added by the patch in relation_excluded_by_constraints():

+ /*
+ * When constraint_exclusion is set to 'partition' we only handle
+ * OTHER_MEMBER_RELs, or BASERELs in cases where the result target is an
+ * inheritance parent or a partitioned table.
+ */

Just to clarify this a bit, would it be a good idea to be specific by
appending " (see inheritance_planner() where this is determined)" or some
such to this sentence?

BTW, while we're at it, would it also be a good idea to consider the patch
you had proposed, which I then posted an updated version of, to adjust the
documentation in ddl.sgml (in the section 5.10. Table Partitioning)
regarding the relationship between constraint exclusion and declarative
partitioning?

https://www.postgresql.org/message-id/c2838545-0e77-3c08-cd14-1c3bbf9eb62d%40lab.ntt.co.jp

Thanks,
Amit