Adding GROUP BY to PIVOT operator (how?)

  • I am trying to show how many employees were hired in each department by startyear (2001, 2002, 2003) and then to rollup these year totals by department. So, for example, Engineering hired one employee in 2001, another in 2002, and another in 2003 with 3 total in the Engineering department.

    I am able to get the first part of this problem solved with a PIVOT (showing 1 employee under each year) but I'm unable to add a GROUP BY clause after the PIVOT that will rollup each years total by department to show total number (3) of employees hired by department (eg. Enginneering).

    Is GROUP BY the solution or is there another way to roll up by DeptName?

    ---this works:

    select DeptName, [2001], [2002], [2003]

    from

    (select DeptName, EmpID, year(StartDate) as startyear

    from StaffHistory

    ) as p

    PIVOT (count(empid) for Startyear in ([2001], [2002], [2003])) as pvt

    ---this doesn't

    select DeptName, [2001], [2002], [2003]

    from

    (select DeptName, EmpID, year(StartDate) as startyear

    from StaffHistory

    ) as p

    PIVOT (count(empid) for Startyear in ([2001], [2002], [2003])) as pvt

    GROUP BY DeptName

    CREATE TABLE StaffHistory

    (DeptName varchar(40)NOT NULL, EmpID int NOT NULL, StartDate datetime NOT NULL, EndDate datetime NULL);

    INSERT INTO StaffHistory

    VALUES

    ('Engineering', 1, '2003-12-16 00:00:00.000', null),

    ('Tool Design', 2, '2003-07-01 00:00:00.000', null),

    ('Sales', 3, '2003-11-01 00:00:00.000', null),

    ('Engineering', 4, '2002-07-01 00:00:00.000', null),

    ('Tool Design', 5, '2002-07-01 00:00:00.000', null),

    ('Sales', 6, '2002-11-01 00:00:00.000', null),

    ('Engineering', 7, '2001-07-01 00:00:00.000', null),

    ('Tool Design', 8, '2001-07-01 00:00:00.000', null),

    ('Sales', 9, '2001-11-01 00:00:00.000', null);

    --Quote me

  • Try this, it will give you a start.

    SELECT DeptName

    , [2001]

    , [2002]

    , [2003]

    , [All]

    FROM (

    SELECT DeptName

    , startyear = isnull(cast(year(StartDate) AS CHAR(4)), 'All')

    , Cnt = Count(empid)

    FROM StaffHistory

    GROUP BY DeptName

    , year(StartDate)

    WITH rollup

    ) src

    pivot(sum(Cnt) FOR startyear IN (

    [2001]

    , [2002]

    , [2003]

    , [All]

    )) pvt

    WHERE DeptName IS NOT NULL

  • This solution works, Thanks Sam.

    I have a question about this line:

    , startyear = isnull(cast(year(StartDate) AS CHAR(4)), 'All')

    I understand ISNULL works because a NULL for startdate is created by the creation of an All column. But, can you explain why it is that you have to use the CAST function, as in , "cast(year(StartDate) AS CHAR(4)"? Is it because datetime is not convertible to a string value?

    --Quote me

  • Look at the remarks section here: http://msdn.microsoft.com/en-us/library/ms184325.aspx

    When using ISNULL, check and replacement expressions should be of compatible data types.

  • For general knowledge and maybe performance (you'll have to test your case), I leave you with my personal preference:

    SELECT DeptName,

    SUM( CASE WHEN YEAR( StartDate) = 2001 THEN 1 ELSE 0 END) [2001],

    SUM( CASE WHEN YEAR( StartDate) = 2002 THEN 1 ELSE 0 END) [2002],

    SUM( CASE WHEN YEAR( StartDate) = 2003 THEN 1 ELSE 0 END) [2003],

    COUNT(*) [All]

    FROM StaffHistory

    GROUP BY DeptName

    You can find more information here:

    Cross Tabs and Pivots, Part 1 – Converting Rows to Columns [/url]

    Cross Tabs and Pivots, Part 2 - Dynamic Cross Tabs [/url]

    Luis C.
    General Disclaimer:
    Are you seriously taking the advice and code from someone from the internet without testing it? Do you at least understand it? Or can it easily kill your server?

    How to post data/code on a forum to get the best help: Option 1 / Option 2

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply