Using distinct with an Order by

Question

Hi all,
I'm not an expert SQL coder and I've got stuck with the following issue.
I have a table named Workorder which contains a set of orders. Orders with status_id = 2 contain an error and this need to be re-submitted:
[ ID ] [WO_ID] [STATUS_ID]
1 10 2
2 10 1
3 20 2
So apparentely it seems very simple, however when orders are re-submitted things get complicated.
In the above query I need to pickup -just- the following row:
[ ID ] [WO_ID] [STATUS_ID]
3 20 2
The following row needs to be excluded from the list:
[ ID ] [WO_ID] [STATUS_ID]
1 10 2
Because the WO_ID has been later re-submitted successufully:
[ ID ] [WO_ID] [STATUS_ID]
2 10 1
So I need to pickup all workorders with a distinct on WO_ID ordered by ID desc, so that the distinct row contains just the highest ID (the latest order re-submitted).
I've tried in several ways, however I wasn't able to mix distinct + order by.
Maybe it can be done also in some other simpler ways ?
Thanks a lot
Frank

Answer 1

So, if the last ID for each WO_ID has STATUS_ID=2, that row needs to in the output.
But if the last ID has a different STATUS_ID, then that WO_ID is not to be in the output at all?
Perhaps this could work for you:
select
max(id) as id,
wo_id,
max(status_id) keep (dense_rank last order by id) as status_id
from workorder
group by wo_id
having max(status_id) keep (dense_rank last order by id) = 2
order by wo_idEdit:
Or an alternative using analytic functions:
select
id,
wo_id,
status_id
from (
   select
   id,
   wo_id,
   status_id,
   last_value(status_id) over (
      partition by wo_id
      order by id
      rows between unbounded preceding and unbounded following
   ) as last_status
   from workorder
where last_status = 2If your workorder table actually contains several other columns you wish to select along with the id and status, then the analytic way can be much easier than a bunch of "keep" expressions ;-)
Edited by: Kim Berg Hansen on Nov 29, 2011 12:22 PM