Select Count(Distinct Value) returns 1

Question

I'm designing a query in SSMS 2005 which looks something like this:

SELECT COUNT(DISTINCT ColumnName) FROM Table WHERE ColumnName IS NOT NULL

When I run the query with COUNT() it returns the value 1. When I run it without COUNT(), SSMS reports the correct value eg 212 records.

The column in question is of datatype numeric(16, 0).

For those who might ask, the query in full is:

SELECT COUNT(DISTINCT O_ID) FROM vEmployers
INNER JOIN vEnrolment ON O_ID = E_EnrolmentEmployer
WHERE E_START >= '01-AUG-2008' AND E_START < '01-AUG-2009'
AND O_ID IS NOT NULL AND O_ID IN (
    SELECT O_ID FROM vEmployers
    INNER JOIN vEnrolment ON O_ID = E_EnrolmentEmployer
    WHERE E_Start < '01-AUG-2008' and E_Start >= '01-AUG-2007'
)

This query basically gives a repeat business figure between two 12month periods.

So I'm wondering why "COUNT(DISTINCT ColumnName)" is returning 1 when "ColumnName IS NOT NULL" has been specified?

Here is a sample of the data when SELECT TOP 10 DISTINCT ColumnName FROM... etc is run:

1346116
1346131
1346425
1346923
1349935
1350115
1350153
2594787
2821944
2879631

Answer 1

I'm guessing it's because all rows returned share the same value for O_ID. You can do a COUNT(*) or COUNT() on a key that is unique to each row to get the row count.

Answer 2

Remove the DISTINCT and you'll get a count on all rows.

Answer 3

Could you please run these queries:

SELECT  COUNT(DISTINCT O_ID)
FROM    vEmployers
INNER JOIN
        vEnrolment
ON      O_ID = E_EnrolmentEmployer
WHERE   E_START >= '01-AUG-2008' AND
        E_START < '01-AUG-2009'
        AND O_ID IN
        (
        SELECT  O_ID
        FROM    vEmployers
        INNER JOIN
                vEnrolment
        ON      O_ID = E_EnrolmentEmployer
        WHERE   E_Start < '01-AUG-2008'
                AND E_Start >= '01-AUG-2007'
        )

and

SELECT  DISTINCT TOP 5 O_ID
FROM    vEmployers
INNER JOIN
        vEnrolment
ON      O_ID = E_EnrolmentEmployer
WHERE   E_START >= '01-AUG-2008' AND
        E_START < '01-AUG-2009'
        AND O_ID IN
        (
        SELECT  O_ID
        FROM    vEmployers
        INNER JOIN
                vEnrolment
        ON      O_ID = E_EnrolmentEmployer
        WHERE   E_Start < '01-AUG-2008'
                AND E_Start >= '01-AUG-2007'
        )
ORDER BY
        O_ID

verbatim, without changing anything?

Answer 4

SELECT   
  COUNT(*)  
FROM    vEmployers  
INNER JOIN vEnrolment ON O_ID = E_EnrolmentEmployer
WHERE 
        E_START >= '01-AUG-2008' 
        AND E_START < '01-AUG-2009'
        AND O_ID IS NOT NULL AND O_ID IN (
          SELECT O_ID FROM vEmployers
          INNER JOIN vEnrolment ON O_ID = E_EnrolmentEmployer
          WHERE E_Start < '01-AUG-2008' and E_Start >= '01-AUG-2007'
        )
GROUP BY
  O_Id

Answer 5

The use of the numeric(16, 0) made me suspect that it was data type related. Add a CAST in the COUNT clause to cast it to an INT type:

Count(Distinct Cast(O_ID as Int))