I think the important point in your comment is that it emphasises the need to think carefully about the impact on NULL on such things.

Whether your suggestion was appropriate or not (as Olekander suggests) is completely down to the business requirements – but it’s the technicians who have to remember that NULL can turn a simple business request into something a little more complex than first thought.

Your code treats NULL state exactly the same way as ‘CLOSED’ state, which unlikely to be a desired behavior.

Regards,
Oleksandr

If state is allowed to be null, I would have slightly changed the index syntax creation (and the query accordingly) as follows:

create index t1_f3 on t1(case when nvl(state,’CLOSED’) = ‘CLOSED’ then to_number(null) else n1 end);

select
/*+ index(t1) t1_f3 */
v1
from
t1
where
case when nvl(state,’CLOSED’) = ‘CLOSED’ then to_number(null) else n1 end = 100;

Even if in this case it seems to be of no consequences in the query result

What do you think?

Best Regards

Mohamed Houri

Nice explanation and thanks, in particular, for mentioning the caveat about maintaining statistics on hidden columns of FBI.

p.s. you should have posted this as another quiz… :)

Valid point, given the specific example I’ve shown; but doing it in the way I’ve shown it addresses the more generic problem of indexing “everything that isn’t closed”.