interesting check constraint behavior

"Gauthier, Dave" <dave.gauthier@intel.com> writes:

thedb=# create table foo (col1 text, constraint chk check (col1 in ('a','b','c',null)));
CREATE TABLE
thedb=# insert into foo (col1) values ('xxx');
INSERT 0 1

Hmmmm... I would have thought that this would have violated the constraint because 'xxx' is not null and nit one of the allowed values.

Nulls are tricky. That constraint is equivalent to

col1 = 'a' or col1 = 'b' or col1 = 'c' or col1 = null

The last reduces to null (not false), so you get either TRUE or NULL out
of the OR condition. CHECK constraints are defined to not fail on a null
result (which is not terribly consistent, but it's what the spec says).
So basically that check constraint will never fail.

Is there a different way I can allow for a static set of values AND null too?

Plain old check (col1 in ('a','b','c')) would work that way. If you
actually want to force it to be non-null, you have to say that
explicitly; usually people use a separate NOT NULL constraint for that.

regards, tom lane

On 7 Jan 2010, at 24:12, Gauthier, Dave wrote:

thedb=# create table foo (col1 text, constraint chk check (col1 in ('a','b','c',null)));
CREATE TABLE
thedb=# insert into foo (col1) values ('xxx');
INSERT 0 1

Hmmmm... I would have thought that this would have violated the constraint because �xxx� is not null and nit one of the allowed values.

As Tom already explained, "value in ('yada', 'yada', null)" evaluates to null. You'll need to make sure your check expression evaluates to either true or false - not null.

I guess you need: check(col1 in ('a', 'b', 'c') or col1 is null)

Alban Hertroys

--
If you can't see the forest for the trees,
cut the trees and you'll see there is no forest.

!DSPAM:737,4b45bc2610731598743439!

Both suggestions (exclude the 'null' from the list, and include "or col1 is null") work.

And if I ever wanted to require that col1 is not null, just add that as a separate constraint.

Thanks guys for the advise and explanation !

-----Original Message-----
From: Alban Hertroys [mailto:dalroi@solfertje.student.utwente.nl]
Sent: Thursday, January 07, 2010 5:49 AM
To: Gauthier, Dave
Cc: pgsql-general@postgresql.org
Subject: Re: [GENERAL] interesting check constraint behavior

On 7 Jan 2010, at 24:12, Gauthier, Dave wrote:

thedb=# create table foo (col1 text, constraint chk check (col1 in ('a','b','c',null)));
CREATE TABLE
thedb=# insert into foo (col1) values ('xxx');
INSERT 0 1

Hmmmm... I would have thought that this would have violated the constraint because 'xxx' is not null and nit one of the allowed values.

As Tom already explained, "value in ('yada', 'yada', null)" evaluates to null. You'll need to make sure your check expression evaluates to either true or false - not null.

I guess you need: check(col1 in ('a', 'b', 'c') or col1 is null)

Alban Hertroys

--
If you can't see the forest for the trees,
cut the trees and you'll see there is no forest.

!DSPAM:1015,4b45bc2510731992717809!