Message-ID: <3cc4f166-0b58-47e2-a03c-7c2873d62b14@oss.nttdata.com>
Date: Wed, 25 Jun 2025 08:18:28 +0900
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: pg_dump --with-* options
To: Robert Haas <robertmhaas@gmail.com>,
 Nathan Bossart <nathandbossart@gmail.com>
Cc: Jeff Davis <pgsql@j-davis.com>, Peter Eisentraut <peter@eisentraut.org>,
 pgsql-hackers <pgsql-hackers@postgresql.org>
References: <7cc52488-f876-4ad3-affd-6e4b0ef0cb09@eisentraut.org>
 <aEMLzPQtjRbj2R7X@nathan>
 <CA+TgmoactJEUHeXuxHLKNXHzm6-XrQFigPur-4nHTMFSiSRkiQ@mail.gmail.com>
 <aErpuM6RXSlK1FBW@nathan>
 <7bf0246542713d80f12c738c1d365d1365750c18.camel@j-davis.com>
 <5e398522803da9bf3f9fc33ca43084ec21ab0fcd.camel@j-davis.com>
 <aFLenfMRS9jFINxq@nathan>
 <031558c60e84362898922caa6a90587e7fdf2a57.camel@j-davis.com>
 <CA+TgmoYW4ycbzZehdS9nOw60ZdtwzWZ-qoVpBe3+QCpeBxvKug@mail.gmail.com>
 <aFrWvlsWwE92Mcpr@nathan>
 <CA+TgmoYuP7TvwThJERO2X7CHJCrTMpiv=Fd=EPOcyWwL=8J4Hg@mail.gmail.com>
Content-Language: en-US
From: Fujii Masao <masao.fujii@oss.nttdata.com>
In-Reply-To: 
 <CA+TgmoYuP7TvwThJERO2X7CHJCrTMpiv=Fd=EPOcyWwL=8J4Hg@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: 
 <https://www.postgresql.org/message-id/3cc4f166-0b58-47e2-a03c-7c2873d62b14%40oss.nttdata.com>
Precedence: bulk


On 2025/06/25 5:07, Robert Haas wrote:
> On Tue, Jun 24, 2025 at 12:48 PM Nathan Bossart
> <nathandbossart@gmail.com> wrote:
>> On Mon, Jun 23, 2025 at 01:38:10PM -0400, Robert Haas wrote:
>>> I had thought we had a consensus that pg_upgrade should preserve stats
>>> but regularly pg_dump shouldn't include them; perhaps I misunderstood
>>> or that changed.
>>
>> I think it's a bit of both.  I skimmed through the past discussions and
>> found that not only was there a rough consensus in 2024 that stats _should_
>> be on by default [0], but also that an updated vote tally didn't show much
>> of a consensus at all [1].  Like you, I thought we had pretty much closed
>> that door, but the aforementioned analysis along with further discussion
>> has me convinced that we might want to reconsider [2].
> 
> Well, I don't know: I still think that's the right answer, so I don't
> really want to reconsider, but I understand that I'm not in charge
> here.

For the record, my vote is: default "off" for pg_dump and pg_dumpall,
and "on" for pg_restore.

For pg_dump and pg_dumpall, I agree with Jeff's idea in [1],
but if the statistics is skipped by default, I don't think
we need a --no-statistics option. So, here's how I think
the options should work:

     * Keep: --schema-only, --data-only, --statistics-only, --no-schema, --no-data, -and -statistics
     * Remove: --no-statistics, --with-schema, and --with-data
     * Combinations:
         Schema + Data + Stats     : --statistics
         Schema + Data             : (default)
         Schema + Stats            : --no-data --statistics
         Data + Stats              : --no-schema --statistics
         Schema only               : --schema-only   (or --no-data)
         Data only                 : --data-only     (or --no-schema)
         Stats only                : --statistics-only (or --no-schema --no-data --statistics)

As I mentioned in [2], if we treat --statistics in the similar way to
--sequence-data, i.e., allow --statistics to be used with --schema-only
or --data-only, we could simplify further:

     * Keep: --schema-only, --data-only, --statistics-only, and --statistics
     * Remove: --no-schema, --no-data, --no-statistics, --with-schema, and --with-data
     * Combinations:
         Schema + Data + Stats     : --statistics
         Schema + Data             : (default)
         Schema + Stats            : --schema-only --statistics
         Data + Stats              : --data-only --statistics
         Schema only               : --schema-only
         Data only                 : --data-only
         Stats only                : --statistics-only

Some may find this confusing due to mixing --statistics with --schema-only
or --data-only, so I understand if there's hesitation.

For pg_restore, I believe there's agreement to restore statistics
by default if they exist in the archive. So:

     * Keep: --schema-only, --data-only, --statistics-only, --no-schema, --no-data, and --no-statistics
     * Remove: --with-schema, --with-data, and --statistics
     * Combinations:
         Schema + Data + Stats     : (default)
         Schema + Data             : --no-statistics
         Schema + Stats            : --no-data
         Data + Stats              : --no-schema
         Schema only               : --schema-only   (or --no-data --no-statistics)
         Data only                 : --data-only     (or --no-schema --no-statistics)
         Stats only                : --statistics-only (or --no-schema --no-data)

Thought?

Regards,

[1] https://postgr.es/m/031558c60e84362898922caa6a90587e7fdf2a57.camel@j-davis.com
[2] https://postgr.es/m/94f89b0a-5d83-4a67-9092-50ba3913441c@oss.nttdata.com

-- 
Fujii Masao
NTT DATA Japan Corporation