Re: "could not find pathkey item to sort" for TPC-DS queries 94-96

Tomas Vondra <tomas.vondra@enterprisedb.com>

From: Tomas Vondra <tomas.vondra@enterprisedb.com>
To: Robert Haas <robertmhaas@gmail.com>, James Coleman <jtc331@gmail.com>
Cc: Luc Vlaming <luc@swarm64.com>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2021-04-15T20:18:34Z
Lists: pgsql-hackers

On 4/15/21 2:21 AM, Robert Haas wrote:
> On Wed, Apr 14, 2021 at 8:20 PM James Coleman <jtc331@gmail.com> wrote:
>>> Hmm, could be. Although, the stack trace at issue doesn't seem to show
>>> a call to create_incrementalsort_plan().
>>
>> The changes to gather merge path generation made it possible to use
>> those paths in more cases for both incremental sort and regular sort,
>> so by "incremental sort" I read Tomas as saying "the patches that
>> brought in incremental sort" not specifically "incremental sort
>> itself".
> 
> I agree. That's why I said "hmm, could be" even though the plan
> doesn't involve one.
> 

Yeah, that's what I meant. The difference to pre-13 behavior is that we
now call generate_useful_gather_paths, which also considers adding extra
sort (unlike plain generate_gather_paths).

So now we can end up with "Gather Merge -> Sort" paths that would not be
considered before.


regards

-- 
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company



Commits

  1. Rename find_em_expr_usable_for_sorting_rel.

  2. Fix planner failure in some cases of sorting by an aggregate.