Re: proposal - patch: psql - sort_by_size - Mailing list pgsql-hackers

From Pavel Stehule
Subject Re: proposal - patch: psql - sort_by_size
Date
Msg-id CAFj8pRAb1pNPorOqz8NdNSaTx5aO6Frx4T0ZqugT8aABdKUvGA@mail.gmail.com
Whole thread Raw
In response to Re: proposal - patch: psql - sort_by_size  (Pavel Stehule <pavel.stehule@gmail.com>)
List pgsql-hackers


pá 13. 9. 2019 v 9:35 odesílatel Pavel Stehule <pavel.stehule@gmail.com> napsal:


čt 12. 9. 2019 v 0:01 odesílatel Alvaro Herrera from 2ndQuadrant <alvherre@alvh.no-ip.org> napsal:
On 2019-Jul-31, Rafia Sabih wrote:

> I had a look at this patch, seems like a useful thing to have.

So the two initial questions for this patch are

1. Is this a feature we want?
2. Is the user interface correct?

I think the feature is useful, and Rafia also stated as much.  Therefore
ISTM we're okay on that front.

As for the UI, Fabien thinks the patch adopts one that's far too
simplistic, and I agree.  Fabien has proposed a number of different UIs,
but doesn't seem convinced of any of them.  One of them was to have
"options" in the command,
 \dt+ [-o 1d,2a]

Another idea is to use variables in a more general form.  So instead of
Pavel's proposal of SORT_BY_SIZE=on we could do something like
SORT_BY=[list]
where the list after the equal sign consists of predetermined elements
(say SIZE, NAME, SCHEMA and so on) and indicates a specific column to
sort by.  This is less succint than Fabien's idea, and in particular you
can't specify it in the command itself but have to set the variable
beforehand instead.

for more generic design probably you need redesign psql report systems. You cannot to use just ORDER BY 1,2 on some columns, but you need to produce (and later hide) some content (for size).

So it can be unfunny complex patch. I finished sort inside pspg and I it looks to be better solution, than increase complexity (and less maintainability (due support old releases)).

I changed status for this patch to withdrawn

I like a idea with enhancing \dt about some clauses like " \dt+ [-o 1d,2a]". But it needs probably significant redesign of describe.c module. Maybe implementation of some simple query generator for queries to system catalogue can good.

Surely - this should be implemented from scratch - I am not a volunteer for that.

Pavel


Regards

Pavel


--
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

pgsql-hackers by date:

Previous
From: Fabien COELHO
Date:
Subject: Re: pgbench - allow to create partitioned tables
Next
From: Dilip Kumar
Date:
Subject: Re: pgbench - allow to create partitioned tables