Re: Expression errors with "FOR UPDATE" and postgres_fdw with partition wise join enabled.

From: Etsuro Fujita <fujita(dot)etsuro(at)lab(dot)ntt(dot)co(dot)jp>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: Robert Haas <robertmhaas(at)gmail(dot)com>, Andres Freund <andres(at)anarazel(dot)de>, Ashutosh Bapat <ashutosh(dot)bapat(at)enterprisedb(dot)com>, Rajkumar Raghuwanshi <rajkumar(dot)raghuwanshi(at)enterprisedb(dot)com>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: Expression errors with "FOR UPDATE" and postgres_fdw with partition wise join enabled.
Date: 2018-08-06 12:30:24
Message-ID: 5B683F60.2070806@lab.ntt.co.jp
Views: Raw Message | Whole Thread | Download mbox
Thread:
Lists: pgsql-hackers

(2018/08/03 22:28), Etsuro Fujita wrote:
> (2018/08/03 22:18), Etsuro Fujita wrote:
>> Here is a patch for refusing to generate PWJ paths when WRVs are
>> involved:
>>
>> * We no longer need to handle WRVs, so I've simplified
>> build_joinrel_tlist() and setrefs.c to what they were before
>> partition-wise join went in, as in the previous patch.
>>
>> * attr_needed data for each child is used for building child-joins'
>> tlists, but I think we can build those by applying
>> adjust_appendrel_attrs to the parent's tlists, without attr_needed. So,
>> I've also removed that as in the previous patch.
>
> One thing to add: as for the latter, we don't need the changes to
> placeholder.c either, so I've also simplified that to what they were
> before partition-wise join went in.

*** a/src/backend/optimizer/plan/planner.c
--- b/src/backend/optimizer/plan/planner.c
***************
*** 3960,3965 **** create_ordinary_grouping_paths(PlannerInfo *root,
RelOptInfo *input_rel,
--- 3960,3967 ----
*/
if (extra->patype != PARTITIONWISE_AGGREGATE_NONE &&
input_rel->part_scheme && input_rel->part_rels &&
+ (input_rel->reloptkind == RELOPT_BASEREL ||
+ input_rel->consider_partitionwise_join) &&
!IS_DUMMY_REL(input_rel))
{
/*
***************
*** 6913,6919 **** apply_scanjoin_target_to_paths(PlannerInfo *root,
* projection-capable, that might save a separate Result node, and
it also
* is important for partitionwise aggregate.
*/
! if (rel->part_scheme && rel->part_rels)
{
int partition_idx;
List *live_children = NIL;
--- 6915,6923 ----
* projection-capable, that might save a separate Result node, and
it also
* is important for partitionwise aggregate.
*/
! if (rel->part_scheme && rel->part_rels &&
! (rel->reloptkind == RELOPT_BASEREL ||
! rel->consider_partitionwise_join))
{
int partition_idx;
List *live_children = NIL;

In the above I used the test whether the relation's reloptkind is
RELOPT_BASEREL or not, but I noticed that I had overlooked the case of a
multi-level partitioned table. So I fixed that and added regression
test cases for that. I also revised comments a bit. Attached is an
updated version of the patch.

Best regards,
Etsuro Fujita

Attachment Content-Type Size
refuse-pwj-when-wrvs-involved-2.patch text/x-diff 50.6 KB

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Laurenz Albe 2018-08-06 13:45:24 Re: Loaded footgun open_datasync on Windows
Previous Message Ashutosh Bapat 2018-08-06 10:23:33 Re: TupleTableSlot abstraction