Thread: Unique functional index and FK constraints

Tom,

When you restructured the unique index location and validation for
foreign key constraints around 9/22 you added the restriction that the
supporting unique index may not be functional.  I believe that this
restriction is not necessary.  Suppose I had a unique index on
LOWER(login).  That would imply that login is unique as well.  Any
function which returns different results given the same input is no good
for a functional index anyway.

http://archives.postgresql.org/pgsql-committers/2002-09/msg00293.php

Kris Jurka

Kris Jurka <jurka@ejurka.com> writes:
> When you restructured the unique index location and validation for
> foreign key constraints around 9/22 you added the restriction that the
> supporting unique index may not be functional.  I believe that this
> restriction is not necessary.

Hmm ... I'm not convinced.  What about functions that return NULL for
some inputs?  A unique index will not reject multiple NULL entries.
        regards, tom lane