Re: Converting SetOp to read its two inputs separately
От | Richard Guo |
---|---|
Тема | Re: Converting SetOp to read its two inputs separately |
Дата | |
Msg-id | CAMbWs485qhU__TBtJ5K0f2MBuQq2VzUc0djN0wpycf0+n_eOQw@mail.gmail.com обсуждение исходный текст |
Ответ на | Converting SetOp to read its two inputs separately (Tom Lane <tgl@sss.pgh.pa.us>) |
Ответы |
Re: Converting SetOp to read its two inputs separately
Re: Converting SetOp to read its two inputs separately |
Список | pgsql-hackers |
On Thu, Nov 14, 2024 at 11:00 AM Tom Lane <tgl@sss.pgh.pa.us> wrote: > Aside from that minor TODO, the main thing that's left undone in this > patch series is to persuade the thing to exploit presorted input > paths. Right now it fails to do so, as can be seen in some of the > regression test cases, eg > > regression=# set enable_hashagg = 0; > SET > regression=# explain (costs off) select unique1 from tenk1 except select unique2 from tenk1 where unique2 != 10; > QUERY PLAN > ------------------------------------------------------------------ > SetOp Except > -> Sort > Sort Key: tenk1.unique1 > -> Index Only Scan using tenk1_unique1 on tenk1 > -> Sort > Sort Key: tenk1_1.unique2 > -> Index Only Scan using tenk1_unique2 on tenk1 tenk1_1 > Filter: (unique2 <> 10) > (8 rows) > > Obviously the sorts are unnecessary, but it's not seeing that. > I suppose this requires integrating generate_nonunion_paths with > the logic from commit 66c0185a3. I tried to make sense of that, > but failed --- either I don't understand it, or there are a > number of things wrong with it. I'd welcome some help with > getting that done. I think we may need to do the following to make this work: 1. We need to teach set_operation_ordered_results_useful() that sorted input paths are also useful for INTERSECT/EXCEPT, so that we can have setop_pathkeys set for the subqueries. 2. In generate_nonunion_paths(), we need to provide a valid "interesting_pathkeys" when calling build_setop_child_paths(). Thanks Richard
В списке pgsql-hackers по дате отправления: