Message-ID: <0086c87f900172e4103a4563e7ffc05856e21412.camel@cybertec.at>
Subject: Re: Patch: Global Unique Index
From: Laurenz Albe <laurenz.albe@cybertec.at>
To: Vik Fearing <vik@postgresfriends.org>, Cary Huang
 <cary.huang@highgo.ca>,  Thomas Kellerer <shammat@gmx.net>
Cc: pgsql-hackers <pgsql-hackers@lists.postgresql.org>
Date: Wed, 30 Nov 2022 13:28:50 +0100
In-Reply-To: <922afef1-430c-e929-2589-ec3190bb11d7@postgresfriends.org>
References: <184879c5306.12490ea581628934.7312528450011769010@highgo.ca>
	 <3866851668762233@ja6lqgzbqb6tniny.sas.yp-c.yandex.net>
	 <2187643.1668783992@sss.pgh.pa.us>
	 <6cdcb7b1-2567-1373-79d1-96467a9b305b@gmx.net>
	 <CAFj8pRCXadc7aCdcYKG-CP-v8q5rpoXLEGbH8z13g-3X4Y6MKQ@mail.gmail.com>
	 <fa359051-7474-671b-ce36-675fe9b5622b@gmx.net>
	 <184aada3c76.1258dc67d3231678.3658037825540104775@highgo.ca>
	 <ae3b677b-7f62-6c6a-bd54-53579f057ae1@postgresfriends.org>
	 <26fcfb7700783faa397c712ea9721090958e0e51.camel@cybertec.at>
	 <922afef1-430c-e929-2589-ec3190bb11d7@postgresfriends.org>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.42.4 (3.42.4-2.fc35) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Archived-At: 
 <https://www.postgresql.org/message-id/0086c87f900172e4103a4563e7ffc05856e21412.camel%40cybertec.at>
Precedence: bulk

On Wed, 2022-11-30 at 10:09 +0100, Vik Fearing wrote:
> On 11/29/22 17:29, Laurenz Albe wrote:
> > On Tue, 2022-11-29 at 13:58 +0100, Vik Fearing wrote:
> > > I disagree.  A user does not need to know that a table is partitionned,
> > > and if the user wants a unique constraint on the table then making them
> > > type an extra word to get it is just annoying.
> > 
> > Hmm.  But if I created a primary key without thinking too hard about it,
> > only to discover later that dropping old partitions has become a problem,
> > I would not be too happy either.
> 
> I have not looked at this patch, but my understanding of its design is 
> the "global" part of the index just makes sure to check a unique index 
> on each partition.  I don't see from that how dropping old partitions 
> would be a problem.

Right, I should have looked closer.  But, according to the parallel discussion,
ATTACH PARTITION might be a problem.  A global index is likely to be a footgun
one way or the other, so I think it should at least have a safety on
(CREATE PARTITIONED GLOBAL INDEX or something).

Yours,
Laurenz Albe