MERGE, UPDATE, INSERT T-SQL Cannot INSERT value NULL - sql

Question

Goal: Keep a running table of student class ranks each month of the year
Haves: I have code that provides me with columns
StudentID; '+#DateTXT+'
The DateTXT is dynamic variable, returns whatever month I'm running the code in.
Needs: I'm trying to use the MERGE, UPDATE, INSERT functions to where I can run the code once and establish a table:
| StudentID | Jan |
| 56789 | 2 |
| 12345 | 7 |
Then each month I add a new month column the permanent table:
EXEC('ALTER TABLE StudentRanking
ADD ' + #DateTXT + ' smallint NOT NULL DEFAULT(999)')
| StudentID | Jan | Feb |
| 56789 | 2 | 999 |
| 12345 | 7 | 999 |
I'll run the ranking code again for February and save it into a temporary table, which I will use to merge, update, insert with the StudentRanking table:
| StudentID | Feb |
| 56789 | 3 |
(note.. student 12345 doesn't come up)
So I'd like to end up with a running list:
EXEC('
MERGE StudentRanking AS TARGET
USING ##TEMPDB2 AS SOURCE ON (TARGET.StudentID = SOURCE.StudentID)
WHEN MATCHED AND TARGET.' + #DateTXT + ' <> SOURCE.' + #DateTXT + '
THEN UPDATE SET TARGET.' + #DateTXT + ' = SOURCE.' + #DateTXT + '
WHEN NOT MATCHED BY TARGET THEN
INSERT (StudentID, ' + #Rank_TXT + ')
VALUES (SOURCE.StudentID, SOURCE.' + #Rank_TXT + ') ')
| StudentID | Jan | Feb |
| 56789 | 2 | 3 |
| 12345 | 7 |null |
Problem: Some students leave the school, thereby creating a null ranking in proceeding months (e.g. 12345 has no rank in February), so when I try to INSERT the results from a temporary table, I get this ERROR:
SQL Server Database Error: Cannot insert the value NULL into column 'Feb', table 'tempdb.dbo.##TEMPDB'; column does not allow nulls. UPDATE fails.
I could do an ISNULL(ranking,0) but I'd rather have nulls than 0's

Answer 1

Always open for a different, better, approach #GarethD!
I actually got it to work by doing:
EXEC('ALTER TABLE StudentRanking
ADD ' + #Date_TXT + ' smallint DEFAULT(null)')
WHEN MATCHED AND TARGET.' + #Date_TXT + ' IS NULL
THEN UPDATE SET TARGET.' + #Date_TXT + ' = SOURCE.' + #Date_TXT + '

Answer 2

The quick fix is make the column not nullable. It is worth pointing out though that this solution does not scale well. A more scalable approach would be to use a properly normalised table where StudentID and Month make up your primary key.
You then have something like:
CREATE TABLE dbo.StudentRanking
(
Date DATE NOT NULL,
StudentID INT NOT NULL,
Score INT NOT NULL,
CONSTRAINT PK_StudentRanking__StudentID_Date PRIMARY KEY(Date, StudentID),
);
You can then create a view on top of this to get the table in the format you wanted:
CREATE VIEW dbo.StudentRankingByYear
WITH SCHEMABINDING
AS
SELECT StudentID,
Year = DATEPART(YEAR, Date),
Jan = SUM(CASE WHEN DATEPART(MONTH, Date) = 1 THEN Score END),
Feb = SUM(CASE WHEN DATEPART(MONTH, Date) = 2 THEN Score END),
Mar = SUM(CASE WHEN DATEPART(MONTH, Date) = 3 THEN Score END),
Apr = SUM(CASE WHEN DATEPART(MONTH, Date) = 4 THEN Score END),
May = SUM(CASE WHEN DATEPART(MONTH, Date) = 5 THEN Score END),
Jun = SUM(CASE WHEN DATEPART(MONTH, Date) = 6 THEN Score END),
Jul = SUM(CASE WHEN DATEPART(MONTH, Date) = 7 THEN Score END),
Aug = SUM(CASE WHEN DATEPART(MONTH, Date) = 8 THEN Score END),
Sep = SUM(CASE WHEN DATEPART(MONTH, Date) = 9 THEN Score END),
Oct = SUM(CASE WHEN DATEPART(MONTH, Date) = 10 THEN Score END),
Nov = SUM(CASE WHEN DATEPART(MONTH, Date) = 11 THEN Score END),
Dec = SUM(CASE WHEN DATEPART(MONTH, Date) = 12 THEN Score END)
FROM dbo.StudentRanking
GROUP BY StudentID, DATEPART(YEAR, Date);
GO
You can even create an indexed view on top of this to get the table in the format you wanted (if you really needed to, but the query should perform well enough without the need for a view), the only difference is you cannot have null columns, so the missing months would have to show as 0:
CREATE VIEW dbo.StudentRankingByYear
WITH SCHEMABINDING
AS
SELECT StudentID,
Year = DATEPART(YEAR, Date),
Jan = SUM(CASE WHEN DATEPART(MONTH, Date) = 1 THEN Score ELSE 0 END),
Feb = SUM(CASE WHEN DATEPART(MONTH, Date) = 2 THEN Score ELSE 0 END),
Mar = SUM(CASE WHEN DATEPART(MONTH, Date) = 3 THEN Score ELSE 0 END),
Apr = SUM(CASE WHEN DATEPART(MONTH, Date) = 4 THEN Score ELSE 0 END),
May = SUM(CASE WHEN DATEPART(MONTH, Date) = 5 THEN Score ELSE 0 END),
Jun = SUM(CASE WHEN DATEPART(MONTH, Date) = 6 THEN Score ELSE 0 END),
Jul = SUM(CASE WHEN DATEPART(MONTH, Date) = 7 THEN Score ELSE 0 END),
Aug = SUM(CASE WHEN DATEPART(MONTH, Date) = 8 THEN Score ELSE 0 END),
Sep = SUM(CASE WHEN DATEPART(MONTH, Date) = 9 THEN Score ELSE 0 END),
Oct = SUM(CASE WHEN DATEPART(MONTH, Date) = 10 THEN Score ELSE 0 END),
Nov = SUM(CASE WHEN DATEPART(MONTH, Date) = 11 THEN Score ELSE 0 END),
Dec = SUM(CASE WHEN DATEPART(MONTH, Date) = 12 THEN Score ELSE 0 END),
Records = COUNT_BIG(*)
FROM dbo.StudentRanking
GROUP BY StudentID, DATEPART(YEAR, Date);
GO
CREATE UNIQUE CLUSTERED INDEX UQ_StudentRankingByYear__StudentID_Year
ON dbo.StudentRankingByYear (StudentID, Year);

Answer 3

Change your ALTER TABLE to:
EXEC('ALTER TABLE StudentRanking
ADD ' + #DateTXT + ' smallint NULL')
Assuming the down votes are because I didn't offer an alternative that was normalized, I'd recommend using PIVOT for this type of problem.
Setup:
CREATE TABLE dbo.StudentRanking
(
MonthID CHAR(3) NOT NULL,
StudentID INT NOT NULL,
Score INT NOT NULL,
CONSTRAINT PK_StudentRanking__StudentID_Date PRIMARY KEY(MonthID, StudentID),
);
INSERT INTO dbo.StudentRanking VALUES ('JAN', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('FEB', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('MAR', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('APR', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('MAY', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('JUN', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('JUL', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('AUG', 56321, 3)
INSERT INTO dbo.StudentRanking VALUES ('SEP', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('OCT', 56321, 3)
INSERT INTO dbo.StudentRanking VALUES ('NOV', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('DEC', 56321, 2)
INSERT INTO dbo.StudentRanking VALUES ('JAN', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('FEB', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('MAR', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('APR', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('MAY', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('JUN', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('JUL', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('AUG', 56821, 2)
INSERT INTO dbo.StudentRanking VALUES ('SEP', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('OCT', 56821, 2)
INSERT INTO dbo.StudentRanking VALUES ('NOV', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('DEC', 56821, 1)
INSERT INTO dbo.StudentRanking VALUES ('JAN', 56021, 3)
INSERT INTO dbo.StudentRanking VALUES ('FEB', 56021, 3)
INSERT INTO dbo.StudentRanking VALUES ('MAR', 56021, 3)
INSERT INTO dbo.StudentRanking VALUES ('APR', 56021, 3)
INSERT INTO dbo.StudentRanking VALUES ('MAY', 56021, 3)
INSERT INTO dbo.StudentRanking VALUES ('JUN', 56021, 4)
INSERT INTO dbo.StudentRanking VALUES ('JUL', 56021, 5)
Query
SELECT * FROM StudentRanking
PIVOT (SUM(Score) FOR MonthID IN (JAN, FEB, MAR, APR,
MAY, JUN, JUL, AUG, SEP, OCT, NOV, DEC)) AS PVT
Results
The SUM(SCORE) is harmless in this instance since there is never more than one record per student per month. It's just there to allow PIVOT to know what to work around.