Question

http://msdn.microsoft.com/en-us/library/ms181765.aspx

I see the sql below from above link:

SELECT BusinessEntityID, SalariedFlag
FROM HumanResources.Employee
ORDER BY CASE SalariedFlag WHEN 1 THEN BusinessEntityID END DESC
    ,CASE WHEN SalariedFlag = 0 THEN BusinessEntityID END;
GO

Here is one result I get:

BusinessEntityID,SalariedFlag

7,1

5,1

3,1

1,1

2,0

4,0

6,0

8,0

Could anyone explain why the records with same salariedFlag are next to each other and why salariedFlag=1 chunk is above the salariedFlag=0 chunk?

Answer 1

The sort clause is equivalent to the following, which may be slightly more obvious:

ORDER BY CASE SalariedFlag WHEN 1 THEN BusinessEntityID ELSE null END DESC
        ,CASE WHEN SalariedFlag = 0 THEN BusinessEntityID ELSE null END;

So the first sort field is the BusinessEntityID when SalariedFlag = 1, or null.
That will group all the rows where SalariedFlag = 0 together as they all have a null first sort field.
The rows that SalariedFlag = 1 wil be sorted by BusinessEntityID. It looks like nulls get sorted last in a descending sort so all the SalariedFlag != 1 go last.

That’s the major sort, for the secondary sort, much the same thing happens:
All the rows where SalariedFlag = 0 will be sorted by BusinessEntityID. Since their primary sort fields were all null, they will end up ordered by BusinessEntityID.

And all the rows where SalariedFlag != 0 will be grouped together with a null secondary ordering. If those rows had SalariedFlag = 1, then they would already have been sorted by the primary ordering.

If SalariedFlag can only be 0 or 1 then this sort can be (slightly) simplified to:

ORDER BY CASE SalariedFlag WHEN 1 THEN BusinessEntityID END DESC
        , BusinessEntityID;