"""krislioe_at_gmail.com ΠΙΣΑΜ(Α):
"""
> Hi,
>
> >1. Generally the same as having index on nullable columns.
> This is FTS issue right ? I try not use two different tables because
> all other attributes is the same except for this nullable column, what
> is other solution to this condition ?
>
> > 2. Why? Why do you use unique constraint in place of PK?
> Because I use Surrogate Key as PK and natural Key as Unique constraint,
> and when doing data entry user is using natural Key.
>
> Thank you,
> xtanto
- Considering what null values are not stored in B*Tree indexes. As
long as you gonna ask "what invoices is without orders" - you're going
ahead to FTS. This can be avoided with bitmap indexes (however it would
not be very appropriate for unique column unless the majority of your
values are null). Another solution is to use function based index (for
example nvl(order_id, -1) ) or some surrogate order id (for example -1
mean "no order"). Another thing to consider - when joining invoices
with orders your probably would need to use outer join - this can
eliminate access paths from the optimizer.
- Still can't see the point. If you have PK - you have PK. Why do you
hiding the fact what you have PK under unique constraint?
Received on Thu Nov 09 2006 - 17:45:42 CST
