Re: set-level update fails with unique constraint violation

On Sat, Jan 2, 2010 at 1:40 AM, Roman Neuhauser <neuhauser@sigpipe.cz> wrote:
> # david@fetter.org / 2009-12-31 08:04:58 -0800:
>> On Thu, Dec 31, 2009 at 10:52:20AM +0100, neuhauser+pgsql-general#postgresql.org@sigpipe.cz wrote:
>> > Hello,
>> >
>> > this fails with "duplicate key value":
>> >
>> >     CREATE TABLE x (
>> >       i INT NOT NULL UNIQUE
>> >     );
>> >     INSERT INTO x (i) VALUES (1), (2), (3);
>> >     UPDATE x SET i = i + 1;
>> >
>> > are there any plans to make this work?
>>
>> This will work in 8.5:
>>
>> CREATE TABLE x (
>>     i int NOT NULL UNIQUE DEFERRABLE INITIALLY DEFERRED
>> );
>> INSERT INTO x (i) VALUES (1), (2), (3);
>> UPDATE x SET i = i + 1;
>
> thanks, this might be a bearable workaround in some cases
> provided there's also SET CONSTRAINTS ... DEFERRED / IMMEDIATE.
> what I really want is a mode that fires the constraint check
> at the end of the statement.

What advantage would there be to a constraint that fires right after
to one that fires at the end of the transaction?