Re: ToDo: show size of partitioned table

From: Amit Langote <Langote_Amit_f8(at)lab(dot)ntt(dot)co(dot)jp>
To: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>
Cc: Ashutosh Bapat <ashutosh(dot)bapat(at)enterprisedb(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: ToDo: show size of partitioned table
Date: 2018-11-19 02:37:46
Message-ID: b5e9d6de-615b-47e2-c65f-49da3a568274@lab.ntt.co.jp
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On 2018/11/19 11:17, Alvaro Herrera wrote:
> On 2018-Jul-23, Pavel Stehule wrote:
>
>> p.s. Another patch can be replacement of relation type from "table" to
>> "partitioned table"
>>
>> postgres=# \dt+
>> List of relations
>> +--------+------------+-------------------+-------+---------+-------------+
>> | Schema | Name | Type | Owner | Size | Description |
>> +--------+------------+-------------------+-------+---------+-------------+
>> | public | data | partitioned table | pavel | 0 bytes | |
>> | public | data_2016 | table | pavel | 15 MB | |
>> | public | data_2017 | table | pavel | 15 MB | |
>> | public | data_other | table | pavel | 11 MB | |
>> +--------+------------+-------------------+-------+---------+-------------+
>> (4 rows)
>
> I think this is a clear improvement. The term "table" was introduced
> for this case by f0e44751d7 ("Implement table partitioning.") and now
> the author of that commit supports this change. I used the term "index"
> for partitioned indexes originally because I was copying the existing
> term, but now I too think they should say "partitioned indexes" instead,
> because they are different enough objects from plain indexes.
>
> To be certain I'm not going against some old decision, I digged up
> Amit's old patches. Turns out he submitted psql's describe.c using the
> term "partitioned table" on August 10th [1] and then based on a
> discussion where Robert suggested calling these new objects "partition
> roots" instead to avoid confusion, it was changed to "table" in the next
> submission on August 26th [2]. It seems the right call to have used the
> term "table" in many places (rather than "partition roots"), but at
> least in psql's \dt it seems extremely useful to show the type as
> "partitioned table" instead, because it is one place where the
> distinction is clearly useful.
>
> In this thread there have been no contrary votes, so I'm pushing this
> part soon.
>
> [1] https://postgr.es/m/ad16e2f5-fc7c-cc2d-333a-88d4aa446f96@lab.ntt.co.jp
> [2] https://postgr.es/m/169708f6-6e5a-18d1-707b-1b323e4a6baf@lab.ntt.co.jp

Yeah, I agree that showing "partitioned table" for partitioned tables in
this case is helpful.

Earlier on this thread [1], I had expressed a slight concern about the
consistency of mentioning "partitioned" in various outputs, because many
error messages say "table" even if the table is partitioned. But now I
think that it's orthogonal. We should show "partitioned" where it is helpful.

Thanks,
Amit

[1]
https://www.postgresql.org/message-id/5474c8b6-04e7-1afc-97b6-adb7471c2c71%40lab.ntt.co.jp

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Alvaro Herrera 2018-11-19 02:38:24 Re: Psql patch to show access methods info
Previous Message Alvaro Herrera 2018-11-19 02:37:12 Re: New function pg_stat_statements_reset_query() to reset statistics of a specific query