Re: bug: virtual generated column can be partition key

From: Ashutosh Bapat <ashutosh(dot)bapat(dot)oss(at)gmail(dot)com>
To: jian he <jian(dot)universality(at)gmail(dot)com>
Cc: Fujii Masao <masao(dot)fujii(at)oss(dot)nttdata(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>
Subject: Re: bug: virtual generated column can be partition key
Date: 2025-04-22 08:55:27
Message-ID: CAExHW5vYwW1OLTt1-+qEOWA9othNhytrxLnLUoT=JL9jk-3vxQ@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Sorry I missed this email while sending the patches - our emails crossed in
the air.

On Tue, Apr 22, 2025 at 2:15 PM jian he <jian(dot)universality(at)gmail(dot)com> wrote:

> On Tue, Apr 22, 2025 at 3:02 PM jian he <jian(dot)universality(at)gmail(dot)com>
> wrote:
> > Other than that, it looks good to me for fixing this bug.
>
> The error message seems not that intuitive.
>
> +CREATE TABLE gtest_part_key (f1 date NOT NULL, f2 bigint, f3 bigint
> GENERATED ALWAYS AS (f2 * 2) VIRTUAL) PARTITION BY RANGE
> ((gtest_part_key));
> +ERROR: cannot use generated column in partition key
> +LINE 1: ...D ALWAYS AS (f2 * 2) VIRTUAL) PARTITION BY RANGE ((gtest_par...
> + ^
> +DETAIL: Column "f3" is a generated column.
>

Yes. It's not apparent where does f3 appear in the partition key, to a lay
users. Users who explicitly use whole row expression in a partition key,
would know that a whole row expression contains all columns. And the error
location points to the whole-row reference. So the current state isn't that
bad.

>
> with the attached patch. now,
> +CREATE TABLE gtest_part_key (f1 date NOT NULL, f2 bigint, f3 bigint
> GENERATED ALWAYS AS (f2 * 2) VIRTUAL) PARTITION BY RANGE
> ((gtest_part_key));
> ERROR: cannot use generated column in partition key
> LINE 1: ...D ALWAYS AS (f2 * 2) VIRTUAL) PARTITION BY RANGE ((gtest_par...
> ^
> DETAIL: Generated column "f3" is part of the partition key of
> relation "gtest_part_key"
>
>
The DETAIL here is just mentioning what's already known from the command,
so the change in the DETAIL may not be useful.

If I understand your intention correctly, we have to mention something to
the effect "partition key expression contains a whole-row reference which
in turn contains a generated column." But that might be too verbose.

--
Best Wishes,
Ashutosh Bapat

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Arseniy Mukhin 2025-04-22 09:05:46 amcheck support for BRIN indexes
Previous Message Ashutosh Bapat 2025-04-22 08:47:28 Re: bug: virtual generated column can be partition key