Re: Reducing memory consumed by RestrictInfo list translations in partitionwise join planning

Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>

From: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>

To: pgsql-hackers <pgsql-hackers@postgresql.org>, Richard Guo <guofenglinux@gmail.com>, David Rowley <dgrowleyml@gmail.com>

Date: 2023-10-31T05:17:44Z

Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

Add assertion to verify derived clause has constant RHS
- 887160d1beae 18.0 landed
Make derived clause lookup in EquivalenceClass more efficient
- 88f55bc97622 18.0 landed
Doc: improve documentation for jsonpath behavior.
- 7014c9a4bba2 17.0 cited
Work around implementation restriction in adjust_appendrel_attrs.
- 767c598954bb 16.0 cited

Attachments

0002-Free-intermediate-Relids-created-in-adjust_-20231031.patch (text/x-patch) patch 0002
0003-Save-commuted-RestrictInfo-for-later-use-20231031.patch (text/x-patch) patch 0003
0001-Report-memory-used-for-planning-a-query-in--20231031.patch (text/x-patch) patch 0001
0004-Avoid-translating-RestrictInfo-repeatedly-20231031.patch (text/x-patch) patch 0004

On Thu, Sep 14, 2023 at 4:39 PM Ashutosh Bapat
<ashutosh.bapat.oss@gmail.com> wrote:
>
> The patch set is thus
> 0001 - patch used to measure memory used during planning
>
> 0002 - Patch to free intermediate Relids computed by
> adjust_child_relids_multilevel(). I didn't test memory consumption for
> multi-level partitioning. But this is clear improvement. In that
> function we free the AppendInfos array which as many pointers long as
> the number of relids. So it doesn't make sense not to free the Relids
> which can be {largest relid}/8 bytes long at least.
>
> 0003 - patch to save and reuse commuted RestrictInfo. This patch by
> itself shows a small memory saving (3%) in the query below where the
> same clause is commuted twice. The query does not contain any
> partitioned tables.
> create table t2 (a int primary key, b int, c int);
> create index t2_a_b on t2(a, b);
> select * from t2 where 10 = a
> Memory used without patch: 13696 bytes
> Memory used with patch: 13264 bytes
>
> 0004 - Patch which implements the hash table of hash table described
> above and also code to avoid repeated RestrictInfo list translations.
>
> I will add this patchset to next commitfest.
>
> --
> Best Wishes,
> Ashutosh Bapat

PFA rebased patches. Nothing changes in 0002, 0003 and 0004. 0001 is
the squashed version of the latest patch set at
https://www.postgresql.org/message-id/CAExHW5sCJX7696sF-OnugAiaXS=Ag95=-m1cSrjcmyYj8Pduuw@mail.gmail.com.

-- 
Best Wishes,
Ashutosh Bapat