Help with GROUP BY with SUM

Question

Hello all:
I have always struggled with getting this right, so hopefully someone can assist with this. 
Basically I have two tables:
CREATE TABLE [dbo].[Placement](
[PlacementID] [int] IDENTITY(1,1) NOT NULL,
[TravelerID] [int] NOT NULL,
[ActivityTypeID] [int] NULL,
[ActivityStatusID] [int] NULL,
[StartDate] [datetime] NULL,
[EndDate] [datetime] NULL,
[NumberOfWeeks] [int] NULL,
[BookingCompanyID] [int] NOT NULL
CREATE TABLE [dbo].[TravelerCompanyData](
[TravelerID] [int] NOT NULL,
[CompanyID] [int] NOT NULL,
[StatusID] [int] NOT NULL
I am trying to join the placement table to the TravelerCompanyData table so that I only get one row per TravelerID. The point of this query is to sum up the number of weeks. So I want to sum the number of weeks of Placements that have been completed (if
any), and add them to however many week the person has been currently on assignment. I cannot use the NumberOfWeeks column if a placement is currently "On Assignment" because it counts the number of weeks from the start date to the end date, and
the Traveler will not be all the way to the end date yet.
However, I think with my SUMs in there and all the columns I am referencing in the CASE statement, I am not getting the grouping right. 
The query below will yield four columns, one for each sum because there are four separate rows in the placement table that fit the criteria of the JOIN/WHERE
Some notes:
ActivityStatusID = 23 means the placement status is "on assignment"
= 24 means the placement in "complete"
CompanyID 10 and 12 are the only companies that I want to fit this summing criteria. 
I am using only one travelerID to simplify the result set. 
SELECT
trav.TravelerID
,CASE
WHEN p2.EndDate IS NULL -- Not on assignment, can sum number of weeks
THEN SUM(p.NumberOfWeeks)
ELSE -- multiple assignments, count weeks
CASE
WHEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) % 7 < 3
THEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) / 7
+
ISNULL(p.NumberOfWeeks,0)
WHEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) % 7 >= 3
THEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) / 7 + 1
+
ISNULL(p.NumberOfWeeks,0)
END
END AS WksOnAssignment
FROM TravelerCompanyData trav
-- Completed placements ------------------
LEFT JOIN dbo.Placement p ON p.TravelerID = trav.TravelerID
AND p.ActivityStatusID IN (24)
AND trav.CompanyID = p.BookingCompanyID
-- Placement if on assignment ------------
LEFT JOIN dbo.Placement p2 ON p2.TravelerID = trav.TravelerID
AND p2.ActivityStatusID IN (23)
AND trav.CompanyID = p2.BookingCompanyID
WHERE ( trav.CompanyID IN (10,12))
AND
( p.PlacementID IS NOT NULL
OR
p2.PlacementID IS NOT NULL )
AND trav.TravelerID = 31788
GROUP BY trav.TravelerID
,p.NumberOfWeeks
,p2.StartDate
,p2.EndDate
If I add the NumberOfWeeks column in, it is adding that number to the number of weeks the current On Assignment placement currently has for each row (56 weeks), but I need to add that 56 weeks to the sum of the NumberOfWeeks column. Here is the result set:
TravelerID NumberOfWeeks WksOnAssignment
31788 8 64
31788 12 68
31788 25 81
31788 33 89
So am I joining wrong, grouping wrong, or summing wrong? 
Thanks so much for any pointers!!!

Answer 1

Thanks for the reply. I did some more experimenting and I discovered I don't need the SUM in this case at all. I get the same results when remove the SUM completely, but still GROUP BY the same columns. 
Alternatively, if I SUM and the only GROUP BY tcd.TravelerID, I will get one row, but it is summing each NumberOfWeeks value by the result of the expression. 
Query where I have included the NumberOfWeeks from the p Placement join, and the StartDate and EndDate of p2 Placement table (on assignment match) - NOTE: I have noticed that with this query, the SUM makes no difference:
SELECT
trav.TravelerID
,p.NumberOfWeeks
,p2.StartDate
,p2.EndDate
,CASE
WHEN p2.EndDate IS NULL -- Not on assignment, can sum number of weeks
THEN p.NumberOfWeeks
ELSE -- multiple assignments, count weeks
CASE
WHEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) % 7 < 3
THEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) / 7
+
ISNULL(p.NumberOfWeeks,0)
WHEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) % 7 >= 3
THEN (DATEDIFF(day, p2.StartDate, GETDATE()) + 1) / 7 + 1
+
ISNULL(p.NumberOfWeeks,0)
END
END AS WksOnAssignment
FROM TravelerCompanyData trav
-- Completed placements ------------------
LEFT JOIN dbo.Placement p ON p.TravelerID = trav.TravelerID
AND p.ActivityStatusID IN (24)
AND trav.CompanyID = p.BookingCompanyID
-- Placement if on assignment ------------
LEFT JOIN dbo.Placement p2 ON p2.TravelerID = trav.TravelerID
AND p2.ActivityStatusID IN (23)
AND trav.CompanyID = p2.BookingCompanyID
WHERE ( trav.CompanyID IN (10,12))
AND
( p.PlacementID IS NOT NULL
OR
p2.PlacementID IS NOT NULL )
AND trav.TravelerID = 31788
GROUP BY trav.TravelerID
,p.NumberOfWeeks
,p2.StartDate
,p2.EndDate
Gives:
TravelerID NumberOfWeeks StartDate EndDate WksOnAssignment
31788 8 2013-04-01 2014-10-04 64
31788 12 2013-04-01 2014-10-04 68
31788 25 2013-04-01 2014-10-04 81
31788 33 2013-04-01 2014-10-04 89
See it is summing correctly on the row, but how can I get it to sum the NumberOfWeeks column and add it to 56, which is what the calculation yields from the StartDate of the "On Assignment" placement from p2 until today's date?
Hope that makes sense!