SQL SUMs in where clause with conditionals - sql

Question

I want to get a "totals" report for business XYZ. They want the season,term,distinct count of employees, and total employee's dropped hours, only when dropped hours of anemployee != any adds that equal the drops.
trying to do something like this:
select year,
season,
(select count(distinct empID)
from tableA
where a.season = season
and a.year = year) "Employees",
(select sum(hours)
from(
select distinct year,season,empID,hours
from tableA
where code like 'Drop%'
)
where a.season = season
and a.year = year) "Dropped"
from tableA a
-- need help below
where (select sum(hours)
from(
select distinct year,season,empID,hours
from tableA
where code like 'Drop%'
)
where a.season = season
and a.year = year
and a.emplID = emplID)
!=
(select sum(hours)
from(
select distinct year,season,empID,hours
from tableA
where code like 'Add%'
)
where a.season = season
and a.year = year
and a.emplID = emplID)
group by year,season
It appears I am not correctly doing my where clause correctly. I dont believe I am joining the emplID to each emplID correctly to exlude those whos "drops" <> "adds"
EDIT:
sample data:
year,season,EmplID,hours,code
2015, FALL, 001,10,Drop
20150 FALL, 001,10,Add
2015,FALL,002,5,Drop
2015,FALL,003,10,Drop
The total hours should be 15. EmplyID 001 should be removed from the totaling because he has drops that are exactly equal to adds.

Answer 1

I managed to work it out with a bit of analytics .. ;)
with tableA as (
select 2015 year, 1 season, 1234 empID, 2 hours , 'Add' code from dual union all
select 2015 year, 1 season, 1234 empID, 3 hours , 'Add' code from dual union all
select 2015 year, 1 season, 1234 empID, 4 hours , 'Add' code from dual union all
select 2015 year, 1 season, 1234 empID, 2 hours , 'Drop' code from dual union all
select 2015 year, 1 season, 2345 empID, 5 hours , 'Add' code from dual union all
select 2015 year, 1 season, 2345 empID, 3.5 hours, 'Add' code from dual union all
select 2015 year, 2 season, 1234 empID, 7 hours , 'Add' code from dual union all
select 2015 year, 2 season, 1234 empID, 5 hours , 'Add' code from dual union all
select 2015 year, 2 season, 2345 empID, 5 hours , 'Add' code from dual union all
select 2015 year, 2 season, 7890 empID, 3 hours , 'Add' code from dual union all
select 2014 year, 1 season, 1234 empID, 1 hours , 'Add' code from dual union all
select 2014 year, 1 season, 1234 empID, 2 hours , 'Add' code from dual union all
select 2014 year, 1 season, 1234 empID, 4 hours , 'Add' code from dual
),
w_group as (
select year, season, empID, hours, code,
lead(hours) over (partition by year, season, empID, hours
order by case when code like 'Drop%' then 'DROP'
when code like 'Add%' then 'ADD'
else NULL end ) new_hours
from tableA
)
select year, season, count(distinct empID),
sum(hours-nvl(new_hours,0)) total_hours
from w_group
where code like 'Add%'
group by year, season
/
YEAR SEASON COUNT(DISTINCTEMPID) TOTAL_HOURS
---------- ---------- -------------------- -----------
2015 1 2 15.5
2014 1 1 7
2015 2 3 20
(the first part "with tableA" is just faking some data, since you didn't provide any) :)
[edit]
corrected based on your data, and your explanation - in short, you're counting the DROPs, (minus the ADDs), I was doing the reverse
[edit2] replaced below query with minor tweak based on comment/feedback: don't count an empID if their DROP-ADD zero out)
with tableA as (
select 2015 year, 'FALL' season, '001' empID, 10 hours, 'Drop' code from dual union all
select 2015 year, 'FALL' season, '001' empID, 10 hours, 'Add' code from dual union all
select 2015 year, 'FALL' season, '002' empID, 5 hours, 'Drop' code from dual union all
select 2015 year, 'FALL' season, '003' empID, 10 hours, 'Drop' code from dual
),
w_group as (
select year, season, empID, hours, code,
lag(hours) over (partition by year, season, empID, hours
order by case when code like 'Drop%' then 'DROP'
when code like 'Add%' then 'ADD'
else NULL end ) new_hours
from tableA
)
select year, season, count(distinct empID),
sum(hours-nvl(new_hours,0)) total_hours
from w_group
where code like 'Drop%'
and hours - nvl(new_hours,0) > 0
group by year, season
/
YEAR SEAS COUNT(DISTINCTEMPID) TOTAL_HOURS
---------- ---- -------------------- -----------
2015 FALL 2 15
[/edit]

Answer 2

I think you can do what you want with just conditional aggregation. Something like this:
select year, season, count(distinct empID) as Employees,
sum(case when code like 'Drop%' then hours end) as Dropped
from tableA
group by year, season;
It is hard to tell exactly what you want, because you do not have sample data and desired results (or better yet, a SQL Fiddle). You might also want a having clause:
having (sum(case when code like 'Drop%' then hours end) <>
sum(case when code like 'Add%' then hours end)
)

Answer 3

Are you wanting the result of something like this?
SELECT
year
,season
,COUNT(DISTINCT empID) AS Employees
,SUM(CASE WHEN code LIKE 'Drop%' THEN hours ELSE 0 END) AS Dropped
FROM
TableA
GROUP BY
year
,season
HAVING
(
SUM(CASE WHEN code LIKE 'Drop%' THEN hours ELSE 0 END)
- SUM(CASE WHEN code LIKE 'Add%' THEN hours ELSE 0 END)
) <> 0