Question

There must be a better way to write this, but I’m just not sure what that is. Basically I’m trying to count the distinct values from one column where a condition is met in a separate column. I found this link, but am not sure how to apply it here.

Here is the query, I am using SQL Server 2008R2

SELECT lot,
    (SELECT COUNT(DISTINCT d.pid) FROM invdet d WHERE upk = 0 and d.lot = [invdet].lot) as noUpk,
    (SELECT COUNT(DISTINCT d.pid) FROM invdet d WHERE upk = 1 and d.lot = [invdet].lot) as isUpk
FROM invdet
WHERE ([status] in ('PQ','P2','FA','F2','BH','RL','SC','LD','PS'))
GROUP BY lot
HAVING COUNT(CASE WHEN invdet.upk = 1 THEN 1 ELSE null END) > 0

Answer 1

You can use the fact that COUNT DISTINCT does not count NULLs to your advantage. In order to examine all the rows, create a WHERE EXISTS clause instead of limiting your result set to those statuses:

SELECT lot,
    COUNT(DISTINCT (CASE WHEN upk = 0 THEN pid ELSE NULL END)) as noUpk
    COUNT(DISTINCT (CASE WHEN upk = 1 THEN pid ELSE NULL END)) as isUpk
FROM invdet q
WHERE EXISTS
(
    SELECT 1 
    FROM invdet i 
    WHERE i.[status] in ('PQ','P2','FA','F2','BH','RL','SC','LD','PS')
    AND i.lot = q.lot
)
GROUP BY lot
HAVING COUNT(CASE WHEN upk = 1 THEN 1 ELSE null END) > 0

---

You can also effectively move the HAVING clause into a WHERE EXISTS clause as well, which may be faster, resulting in:

SELECT lot,
    COUNT(DISTINCT (CASE WHEN upk = 0 THEN pid ELSE NULL END)) as noUpk
    COUNT(DISTINCT (CASE WHEN upk = 1 THEN pid ELSE NULL END)) as isUpk
FROM invdet q
WHERE EXISTS
(
    SELECT 1 
    FROM invdet i 
    WHERE i.[status] in ('PQ','P2','FA','F2','BH','RL','SC','LD','PS')
    AND i.lot = q.lot
)
WHERE EXISTS
(
    SELECT 1 
    FROM invdet i 
    WHERE i.upk = 1 
    AND i.lot = q.lot
)
GROUP BY lot