On 23/04/10 15:50, Adrian von Bidder wrote:
On Friday 23 April 2010 03.27:29 Craig Ringer wrote:
insert into test (a,b) values ('fred',NULL);
insert into test (a,b) values ('fred',NULL);
... and will succeed:
Hmm. Perhaps not as ugly as "none" placeholders:
create unique index on test (b)
On Friday 23 April 2010 03.27:29 Craig Ringer wrote:
> insert into test (a,b) values ('fred',NULL);
> insert into test (a,b) values ('fred',NULL);
>
>
> ... and will succeed:
Hmm. Perhaps not as ugly as "none" placeholders:
create unique index on test (b) where a is null;
create unique index o
On 23/04/2010 1:42 AM, Said Ramirez wrote:
Primary keys are defined as 'unique not null' even if they are
composite. So I believe postgres would not let you do that
You can, however, add a UNIQUE constraint on the column set as a whole.
PostgreSQL does *not* enforce non-null in this case, so s
Primary keys are defined as 'unique not null' even if they are
composite. So I believe postgres would not let you do that:
5.3.4. Primary Keys
Technically, a primary key constraint is simply a combination of a
unique constraint and a not-null constraint.
A primary key indicates that a c
Does any SQL standard allows for a multicolumn primary key where in one
record there is a null in on of the primary key columns?
regards
Szymon Guz
We've got a table that has a definition as follows:
CREATE TABLE linking_table (
fk int8 REFERENCES source_table( pk1 ),
value int8,
PRIMARY KEY( fk1, value )
);
I would've thought that the multicolumn primary key would behave as a multicolumn index is supposed to behave per
http://www.postgres