Folks,

I just stumbled across a little problem that looks like this:

CREATE TABLE foo (
    id SERIAL PRIMARY KEY,
    f_text TEXT NOT NULL
);

CREATE UNIQUE INDEX uniq_foo_f_text ON foo(LOWER(TRIM(f_text)));

COPY foo(f_text) FROM stdin;
a
b
c
\.

Oh, shucks.  I'd like to permute these.  Lemme start a transaction:

test(at)[local]=> BEGIN;
BEGIN
test(at)[local]=> UPDATE foo set f_text = 'a' where id=3;
ERROR:  duplicate key violates unique constraint "uniq_foo_f_text"

Dang!  Everything was going to be kosher at the end of the
transaction, but I never got a chance.

Is there some way to make the index check INITIALLY DEFERRABLE the way
a regular column/table constraint could be?

I'd much appreciate any insights into this :)

Cheers,
D
-- 
David Fetter david(at)fetter(dot)org http://fetter.org/
phone: +1 415 235 3778

Remember to vote!

Responses

• Re: Deferrable UNIQUE INDEX? at 2006-01-04 02:20:43 from Michael Glaesemann
• Re: Deferrable UNIQUE INDEX? at 2006-01-04 02:33:41 from Andrew Dunstan

pgsql-hackers by date

Next:	From: Euler Taveira de Oliveira	Date: 2006-01-04 02:18:06
	Subject: Re: Inconsistent syntax in GRANT
Previous:	From: Josh Berkus	Date: 2006-01-04 01:19:41
	Subject: Inconsistent syntax in GRANT