To: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
From: Vik Fearing <vik@postgresfriends.org>
Subject: GROUP BY DISTINCT
Message-ID: <bf3805a8-d7d1-ae61-fece-761b7ff41ecc@postgresfriends.org>
Date: Sun, 21 Feb 2021 13:52:24 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
Content-Type: multipart/mixed;
 boundary="------------D258B3365BDD8AD18F46455C"
Content-Language: en-US
Precedence: bulk

This is a multi-part message in MIME format.
--------------D258B3365BDD8AD18F46455C
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

When combining multiple grouping items, such as rollups and cubes, the
resulting flattened grouping sets can contain duplicate items.  The
standard provides for this by allowing GROUP BY DISTINCT to deduplicate
them prior to doing the actual work.

For example:
GROUP BY ROLLUP (a,b), ROLLUP (a,c)

expands to the sets:
(a,b,c), (a,b), (a,b), (a,c), (a), (a), (a,c), (a), ()

but:
GROUP BY DISTINCT ROLLUP (a,b), ROLLUP (a,c)

expands to just the sets:
(a,b,c), (a,b), (a,c), (a), ()

Attached is a patch to implement this for PostgreSQL.
-- 
Vik Fearing

--------------D258B3365BDD8AD18F46455C
Content-Type: text/x-patch; charset=UTF-8;
 name="0001-implement-GROUP-BY-DISTINCT.v01.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
 filename="0001-implement-GROUP-BY-DISTINCT.v01.patch"