Partial indexes - Mailing list pgsql-general

In 8.1.2.

Table contains columns:
    keyp1 not null, keyp2 not null, keyp3 nullable.

The queries will be separated into two kinds:
one on those rows where keyp3 is null and
the second on where keyp3 is not null.

I think I want to:
    create unique index pk on table tbl (keyp1, keyp2);
    create unique index range on table tbl (keyp1, keyp2, keyp3) where keyp3 is not null;

Are these indexes redundant given uniqueness requirement and the type of queries
that will be run against the table?  I don't think a non-unique index over
all three columns be adequate given the uniqueness constraints.

Opinions?

Thanks,

elein
elein@varlena.com