How to Group on Pivoted Table with Dynamic Rows? - sql-server

Question

I have a dataset like this
[student][datetime][dictionary][value]
[1234 ][datetime][1 ][a ]
[1234 ][datetime][2 ][b ]
I want to create a dataset like this
[student][1 ][2 ]
[1234 ][a ][b ]
I understand that SQL Server pivots are a way to accomplish this. I've written the below code to attempt to pivot the data however, it's naturally not grouped.
So my results end up like this:
[student][datetime][1 ][2 ]
[1234 ][datetime][null ][b ]
[1234 ][datetime][a ][null ]
How can I best group these? I don't care about having the other source columns such as datetime in the final dataset, only the matrix of dictionaries and values
Thank you
DECLARE #columns AS VARCHAR(MAX);
DECLARE #sql AS VARCHAR(MAX);
SELECT #columns = substring((Select DISTINCT ',' + QUOTENAME(dictionary) FROM syuservalues FOR XML PATH ('')),2, 1000);
SELECT #sql =
'SELECT Pivoted.*
FROM syuservalues
PIVOT
( MAX(value)
FOR dictionary IN( ' + #columns + ' )) as Pivoted where student = 327392'; ' Will eventually want it for all students
EXECUTE(#sql);

Answer 1

a and b is brought to two rows because, you have added [datetime] column for pivot and the [datetime] value for a and b will be different.
SAMPLE TABLE
CREATE TABLE #TEMP([student] INT,[datetime] DATETIME,[dictionary] INT,[value] VARCHAR(30))
INSERT INTO #TEMP
SELECT 1234,'2015-02-01',1,'a'
UNION ALL
SELECT 1234, '2015-02-01',2,'b'
Declare a variable to get columns for pivot dynamically
DECLARE #cols NVARCHAR (MAX)
SELECT #cols = COALESCE (#cols + ',[' + CAST([dictionary] AS VARCHAR(10)) + ']',
'[' + CAST([dictionary] AS VARCHAR(10)) + ']')
FROM
(
SELECT DISTINCT [dictionary]
FROM #TEMP
) PV
ORDER BY CAST([dictionary] AS INT)
Now pivot it. I have written the logic inside.
DECLARE #query NVARCHAR(MAX)
SET #query = '-- Your pivoted result is here
SELECT *
FROM
(
-- Select data before pivot. We are not taking [datetime] column because
-- it will not bring ur expected result
SELECT student,[dictionary],value
FROM #TEMP
) x
PIVOT
(
-- Value in each dynamic column
MIN(value)
-- Tell the columns to pivot
FOR [dictionary] IN (' + #cols + ')
) p
ORDER BY student;'
EXEC SP_EXECUTESQL #query

Answer 2

For this kind of problem, you can just use an arbitrary aggregate function so that you get rid of the nulls. I usually use max because it can be used on both numeric and string data types.
select pivoted.student, [1] = max([1]), [2] = max([2])
from syuservalues
pivot(max(value) for dictionary in(...)) pivoted
group by pivoted.student