Reply-To: <jppelletier@e-djuster.com>
From: Jean-Pierre Pelletier <jppelletier@e-djuster.ca>
MIME-Version: 1.0
Thread-Index: AdFv41QPz7mKk0YFQt26fcWzK1PeHA==
Date: Thu, 25 Feb 2016 10:43:53 -0500
Message-ID: <6fecdc5455863ad2585e3922fdd314e5@mail.gmail.com>
Subject: Re: Declarative partitioning
To: pgsql-hackers@postgresql.org
Content-Type: text/plain; charset=UTF-8
Precedence: bulk
Sender: pgsql-hackers-owner@postgresql.org

Why not based it on "Exclusion Constraint" ?

Most discussions as of late seems to focus on Range overlaps which appeal
(I would think) is that it supports both "equality" and "overlaps", two
popular partitioning schemes.

"Equality" as in "value1 = value2" can be implemented with "range
overlaps"
as "range(value1,value) = range(value,value2)".

I would think that Partitioning schemes can be Declarative, Efficient and
not restricted to Equality and Overlaps as long as all partitions (of a
partitioned table) are using a single partitioning definition expressed
as:
- An Immutable Expression on tuple columns, in the simplest case a single
column
- An Operator, in the simplest case, "equality"

That seems very close to the semantic of "Constraint Exclusion" as
described here:
http://thoughts.davisjeff.com/2010/09/25/exclusion-constraints-are-general
ized-sql-unique/

If partitioning could be based on EC, it would bring these additional
benefits:
- The choice of operator as long as it is boolean. commutative and
Indexable
- The use of Expression/Function and not just bare columns

Jean-Pierre Pelletier


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers