Re: [PATCH v1] Avoid deforming unused trailing columns in COPY TO

From: Chao Li <li(dot)evan(dot)chao(at)gmail(dot)com>
To: Tomás Senart <ts(at)perfloop(dot)ai>
Cc: pgsql-hackers(at)lists(dot)postgresql(dot)org
Subject: Re: [PATCH v1] Avoid deforming unused trailing columns in COPY TO
Date: 2026-11-19 01:44:32
Message-ID: 4401F3C1-8330-4B2B-BD8D-C9A1F0E4316B@gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

> On Nov 19, 2026, at 04:52, Tomás Senart <ts(at)perfloop(dot)ai> wrote:
>
> Hi Chao,
>
> Thanks for the review. Both points are right, and v2 attached does both.
>
> 1. max_attnum is now an AttrNumber placed after the two bool fields,
> so it fits in the existing padding and CopyToStateData does not grow.
> 2. The slot_getsomeattrs() call in the else branch is gone.
> CopyOneRowTo() already deforms the needed prefix, so that branch now
> only sets copyslot = slot.
>
> With assertions enabled, the regress suite (243 tests) and the
> subscription TAP tests, including 031_column_list, pass. The code
> change is otherwise the same as v1.
>
> Regards,
> Tomás
> <v2-0001-Avoid-deforming-unused-trailing-columns-in-COPY-T.patch>

V2 looks good to me.

I just ran a simple performance test on my MacBook M4 with a 50-column table containing 1 million rows, copying the first 3 columns, middle 3 columns, and last 3 columns. The test shows some improvements:
```
Master (091b2926044) v2
First 3 columns 155ms 133ms
Middle 3 columns 156ms 140ms
Last 3 columns 157ms 159ms
```

The durations shown in the table are averages from multiple runs. The tests were run with builds that had assertions and debugging disabled.

We can see that copying the first 3 columns and middle 3 columns shows a noticeable performance improvement. When copying the last 3 columns, there is little difference before and after the patch.

I ran the tests with builds disabling asserts and debug.

Best regards,
--
Chao Li (Evan)
HighGo Software Co., Ltd.
https://www.highgo.com/

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Joao Rito 2026-11-19 01:45:13 [PATCH] pg_createsubscriber: add hint to wal_level error
Previous Message Bharath Rupireddy 2026-11-19 01:38:38 Re: pg_xmin_horizon: a system view of everything pinning the xmin horizon