Horton
Menu

Re: speedup COPY TO for partitioned table.

jian he <jian.universality@gmail.com>

View thread
From: jian he <jian.universality@gmail.com>
To: Kirill Reshke <reshkekirill@gmail.com>
Cc: vignesh C <vignesh21@gmail.com>, David Rowley <dgrowleyml@gmail.com>, Melih Mutlu <m.melihmutlu@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2025-04-01T01:01:10Z
Lists: pgsql-hackers

Attachments

On Mon, Mar 31, 2025 at 4:05 PM Kirill Reshke <reshkekirill@gmail.com> wrote:
>
> Hi!
> I reviewed v7. Maybe we should add a multi-level partitioning case
> into copy2.sql regression test?
>

sure.

> I also did quick benchmarking for this patch:
>
> ==== DDL
>
>  create table ppp(i int) partition by range (i);
>
> genddl.sh:
>
> for i in `seq 0 200`; do echo "create table p$i partition of ppp for
> values from ( $((10 * i)) ) to ( $((10 * (i + 1))) ); "; done
>
> === insert data data:
> insert into ppp select i / 1000  from generate_series(0, 2000000)i;
>
> === results:
>
>
> for 2000001 rows speedup is 1.40 times : 902.604 ms (patches) vs
> 1270.648 ms (unpatched)
>
> for 4000002 rows speedup is 1.20 times : 1921.724 ms (patches) vs
> 2343.393 ms (unpatched)
>
> for 8000004 rows speedup is 1.10 times : 3932.361 ms (patches) vs
> 4358.489ms (unpatched)
>
> So, this patch indeed speeds up some cases, but with larger tables
> speedup becomes negligible.
>

Thanks for doing the benchmark.

Commits

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

  1. Support COPY TO for partitioned tables.

  2. Re-implement the ereport() macro using __VA_ARGS__.