Sub query inside a decode statement in oracle - sql

Question

I have data in a table like below:
EMPLOYEE_NUM START_DTM END_DTM
47567567446 5/9/2019 12:00:00 PM
76475676575675756 5/10/2019 12:00:00 PM 5/10/2019 11:59:59 PM
456756765767 5/3/2019 12:00:00 PM 5/8/2019 11:59:59 PM
74676576764565 5/2/2019 12:00:00 PM 5/8/2019 11:59:59 PM
98695689576 5/1/2019 12:00:00 PM
I want retrieve employee_num with below conditions:
if there are any entries with NULL as end_dtm,
then print employee_num with max(start_dtm) and whose end_dtm is null
if there are no entries with NULL as end_dtm,
then print employee_num with end_dtm=max(end_dtm)
I tried with
select
decode(select count(1) from employee where end_dtm is null,
0,
select employee_num where end_dtm=(select max(end_dtm) from employee),
select employee_num where start_dtm=(select max(start_dtm) from employee where end_dtm is null),
) from dual
Looks like this is not a valid query.
Could anybody please help.

Answer 1

So, you must distinguish between END_DTM being null vs. non-null first, and only then consider either END_DTM or START_DTM for further ordering.
The function NVL2 is exactly what you need for the first part. It takes three arguments; it returns the second argument when the first is non-null, and the third argument when it's null. So: NVL2(END_DTM, 1, 0) will return 1 when the date is non-null and 0 when it is null. And then, you can use NVL2 again, to choose END_DTM when it is not null, but START_DTM when it is null.
You can then use this in defining ROW_NUMBER(), in the ORDER BY clause, like so:
select [whatever]
from
( select t.*,
row_number() over (order by nvl2(END_DTM, 1, 0),
nvl2(END_DTM, END_DTM, START_DTM) desc) rn
from your_table t
)
where rn = 1
;
If END_DTM is always non-null (or if it is always null) then the first NVL2 value is constant, so the ordering is only by END_DTM (if END_DTM is always non-null), resp. by START_DTM (if END_DTM is always null). But if there is a mix of null and non-null END_DTM, then only those with null END_DTM are tied by the first ordering criterion, and then only those rows are considered for selecting the max START_DTM.

Answer 2

I would have used analytical function for this as following:
Select employee_num from
(Select t.employee_num
Sum(case when end_dtm is null then 1 end) as cnt,
Row_number()
over (order by end_dtm desc nulls last) as end_dtm_rn,
Row_number()
over (order by case when end_dtm is null then start_dtm end desc nulls last) as start_dtm_rn
From employee t)
Where case when cnt = 0
then end_dtm_rn
Else start_dtm_rn end = 1;
Cheers!!

Answer 3

Just use aggregation and case:
select employee_num,
(case when count(*) = count(end_dtm) -- no nulls
then max(end_dtm)
else max(start_dtm)
end)
from t
group by employee_num;
If you want this per row in the original data, use analytic functions:
select t.*,
(case when count(*) over (partition by employee_num) = count(end_dtm) over (partition by employee_num) -- no nulls
then max(end_dtm) over (partition by employee_num)
else max(start_dtm) over (partition by employee_num)
end)
from t;
EDIT:
An alternative reading of what you want suggests that you want one employee_num subquery to your conditions. If that is the case, then this is even simpler:
select t.*
from (select t.*
from t
order by (case when end_dtm is null then 1 else 2 end), -- nulls first
start_dtm desc
) t
where rownum = 1;