> In doing some more 7.0 testing, I ran across a difference in functionality
> concerning unique indexes and errors that are reported when you try to 
> violate the index.  I'm not sure if this change is intentional, so I'm 
> bringing it up here.  In 6.5.3, if you try to update a row that violates 
> a unique index, the query fails and said error is reported to the 
> application.  However, in 7.0 the query succeeds, but updates 0 rows.  Hence, 
> no errors are reported back to the application.    This is not normally 
> a problem because I typically check the constrait before updating.  
> 
> 
> in 7.0/beta3
> basement=>  update foobar set unique_colum = '2000-04-09' where foobar_id = 32;
> UPDATE 0
> basement=> 
> 
> in 6.5.3
> basement=> update foobar set unique_colum = '2000-04-09' where foobar_id = 32;
> ERROR:  Cannot insert a duplicate key into a unique index
> basement=> 

I'm not sure how your table looks like, but seems following test with
current (not b3) works here:

test=# create table foobar (unique_column date unique, foobar_id int primary key);
NOTICE:  CREATE TABLE/UNIQUE will create implicit index 'foobar_unique_column_key' for table 'foobar'
NOTICE:  CREATE TABLE/PRIMARY KEY will create implicit index 'foobar_pkey' for table 'foobar'
CREATE
test=# insert into foobar values('2000-4-8', 32);
INSERT 2231126 1
test=# insert into foobar values('2000-4-9', 33);
INSERT 2231127 1
test=# update foobar set unique_column = '2000-04-09' where foobar_id = 32;
ERROR:  Cannot insert a duplicate key into unique index foobar_unique_column_key
--
Tatsuo Ishii

From pgsql-hackers-owner(at)hub(dot)org  Thu Apr  6 21:26:03 2000

In response to

• Unique Key Violation 7.0 vs. 6.5.3 at 2000-04-07 00:30:37 from Brian Hirt

pgsql-hackers by date

Next:	From: Tom Lane	Date: 2000-04-08 03:48:30
	Subject: Re: [HACKERS] Doc updates
Previous:	From: Tom Lane	Date: 2000-04-07 01:12:09
	Subject: Re: 7.0 like selectivity