Re: ALTER TABLE .... make constraint DEFERRABLE

From: Simon Riggs <simon(at)2ndQuadrant(dot)com>
To: Bruce Momjian <bruce(at)momjian(dot)us>
Cc: pgsql-hackers(at)postgresql(dot)org
Subject: Re: ALTER TABLE .... make constraint DEFERRABLE
Date: 2010-06-03 08:53:20
Message-ID: 1275555200.21465.2845.camel@ebony
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Wed, 2010-06-02 at 21:06 -0400, Bruce Momjian wrote:
> Simon Riggs wrote:
> >
> > Deferrable unique constraints seem an interesting feature, though I have
> > either some questions or some issues, not sure which.
> >
> > I don't seem to be able to find any way to do an ALTER TABLE that adds
> > this new capability to an existing table.
>
> I was able to do it:
>
> test=> create table test (x int unique DEFERRABLE INITIALLY DEFERRED);
> NOTICE: CREATE TABLE / UNIQUE will create implicit index "test_x_key"
> for table "test"
> CREATE TABLE
>
> test=> alter table test add column y int;
> ALTER TABLE
>
> test=> alter table test add unique (y) DEFERRABLE INITIALLY DEFERRED;
> NOTICE: ALTER TABLE / ADD UNIQUE will create implicit index
> "test_y_key" for table "test"
> ALTER TABLE
>
> Is that what you were asking?

No. I wanted to defer an existing UNIQUE constraint. That doesn't seem
to be possible. You *can* add a whole new constraint and then drop the
old one, though that's not quite as cool.

> > There is no way to add a constraint via a CREATE TABLE AS SELECT, so
> > that means there is no way to use the feature at all in that case.
>
> Uh, CREATE TABLE AS SELECT seems to be very limited, but I have not
> heard any complaints about it before.
>
> > Also, foreign keys can't be defined that refer to a deferrable primary
> > key. That isn't mentioned at all in the manual with regard to the
> > DEFERRABLE clause, though it is mentioned in the FK section. You get
> > this error message
> > ERROR: cannot use a deferrable unique constraint for referenced table
> >
> > The use case for this feature looks a little narrow at present. Can we
> > do something about usability?
>
> Not sure why that was a limitation.

Regrettably it makes it an unusable limitation for many people.

All large tables are referenced in a typical database that uses PKs/FKs.

--
Simon Riggs www.2ndQuadrant.com

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Fujii Masao 2010-06-03 08:56:49 Re: Keepalive for max_standby_delay
Previous Message KaiGai Kohei 2010-06-03 08:45:10 [PATCH] Fix leaky VIEWs for RLS