Reply-To: jesper.pedersen@redhat.com
Subject: Re: [HACKERS] Proposal: Local indexes for partitioned table
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: Simon Riggs <simon@2ndquadrant.com>, Robert Haas <robertmhaas@gmail.com>,
 Pg Hackers <pgsql-hackers@postgresql.org>,
 Maksim Milyutin <milyutinma@gmail.com>
References: <20171114174946.m7ihrns7jxzrsavz@alvherre.pgsql>
From: Jesper Pedersen <jesper.pedersen@redhat.com>
Organization: Red Hat
Message-ID: <3d57d8ad-e76d-f0cf-0c42-d1202c643dba@redhat.com>
Date: Tue, 14 Nov 2017 14:07:20 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.4.0
MIME-Version: 1.0
In-Reply-To: <20171114174946.m7ihrns7jxzrsavz@alvherre.pgsql>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi,

On 11/14/2017 12:49 PM, Alvaro Herrera wrote:
>> Thanks, pushed.
> 
> Here's the remaining bits, rebased.
> 

First of all, thanks for working on this.

I have been looking at the "CREATE INDEX ... ONLY" syntax, and I think 
it could cause some confusion due to the "Note" described in 
create_index.sgml.

An alternative, maybe crazy, could be to treat a partitioned index as 
one; e.g. all operations are on the definition. That way ONLY, ATTACH 
and DETACH could be eliminated. Once a partition is detached any 
partitioned indexes would be marked as a normal index, and a partition 
could only be attached if it had indexes satisfying all partition index 
definitions. Bloat could be handled by swapping the index as suggested 
by David [1]. It would be less flexible, but people always have the 
option to define indexes directly on the partitions.

Some comments.

0004's alter_index.sgml are missing the "PARTITION" keyword for both the 
ATTACH and DETACH commands.

A small thing, for

-- test.sql --
CREATE TABLE test (a integer NOT NULL) PARTITION BY HASH(a);
CREATE TABLE test_p00 PARTITION OF test FOR VALUES WITH (MODULUS 2, 
REMAINDER 0);
CREATE TABLE test_p01 PARTITION OF test FOR VALUES WITH (MODULUS 2, 
REMAINDER 1);
CREATE INDEX idx_test_a ON test (a);
-- test.sql --

I would expect that the index names were 'test_p00_idx_test_a' and 
'test_p01_idx_test_a'.

psql completion for "CREATE INDEX blah ON tes<TAB>" only lists the child 
tables. Completion for the ALTER INDEX commands is missing.

[1] 
https://www.postgresql.org/message-id/CAKJS1f_Wf%3DM06o8iQi72SxN%2BZpLV4c0CwYoN5xYVY4aWWX-jBQ%40mail.gmail.com

Best regards,
  Jesper