SQL pivot with variable lines - sql

Question

I have to get a variable amount of lines into columns. I have something like:
EMP EMP_ID DIV_ID ALLOCATION
Smith, Tom 3605 11300 20.00
Smith, Tom 13605 11310 80.00
Benetar, Pat 7460 11012 25.00
Benetar, Pat 7460 11015 75.00
Walkin, Chris 13892 11012 90.00
Walkin, Chris 13892 11015 10.00
Kent, Clark 12262 10015 50.00
Kent, Clark 12262 11210 25.00
Kent, Clark 12262 11220 25.00
What I am looking for is something like:
EMP EMP_ID DIV_ID_01 DIV_01_ALOC DIV_ID_02 DIV_02_ALOC DIV_ID_03 DIV_03_ALOC
Smith, Tom 3605 11300 20.00 11310 80.00
Benetar, Pat 13605 11012 25.00 11015 75.00
Walkin, Chris 13892 11012 90.00 11015 10.00
Kent, Clark 12262 11015 50.00 11210 25.00 11220 25.00
I would like to avoid using a large amount of CASE statements. I am trying now to work with pivots, but am having a tough time with headers.

Answer 1

UPDATED:
After a couple of attempts, I came up with the solution below which utilizes two PIVOT functions and a GROUP BY in order to match your expected result.
Here is the code below. NOTE: This is meant for SQL Server 2005+
with testdata(Emp, EMP_ID, DIV_ID, ALLOCATION)
as
(
select 'Smith, Tom',3605, 11300,20.00
union all
select 'Smith, Tom',3605, 11310, 80.00
union all
select 'Benetar, Pat',7460, 11012,25.00
union all
select 'Benetar, Pat',7460, 11015,75.00
union all
select 'Walkin, Chris',13892, 11012, 90.00
union all
select 'Walkin, Chris', 13892, 11015, 10.00
union all
select 'Kent, Clark', 12262, 10015, 50.00
union all
select 'Kent, Clark', 12262, 11210, 25.00
union all
select 'Kent, Clark', 12262, 11220, 25.00
)
SELECT Emp
,EMP_ID
,MAX([Div1]) AS DIV_ID_01
,MAX([Alloc1]) AS DIV_01_ALOC
,MAX([Div2]) AS DIV_ID_02
,MAX([Alloc2]) AS DIV_02_ALOC
,MAX([Div3]) AS DIV_ID_03
,MAX([Alloc3]) AS DIV_03_ALOC
FROM (
SELECT *
,cast(dense_rank() OVER (PARTITION BY emp_id
ORDER BY div_id asc) AS nvarchar) AS [emp_rnk]
,'Alloc' + cast(dense_rank() OVER (PARTITION BY emp_id
ORDER BY div_id asc) AS nvarchar) AS [piv_Alloc_rnk]
,'Div' + cast(dense_rank() OVER (PARTITION BY emp_id
ORDER BY div_id asc) AS nvarchar) AS [piv_Div_rnk]
FROM testdata td
) query
/* After both PIVOT functions are compplete, it still returns a single row for each EMP_ID.
So further aggregation is needed to 'flatten' the result. */
PIVOT (Max(Div_id) FOR [piv_Div_rnk] IN ([Div1],[Div2],[Div3])) AS pivot1
PIVOT (Max(Allocation) FOR [piv_Alloc_rnk] in([Alloc1],[Alloc2],[Alloc3])) AS pivot2
/* Since there is only one value in each of the columns created by the PIVOTS for each EMP_ID taking
the MAX() value and grouping by EMP and EMP_ID flattens the result down to the desired output. */
GROUP BY emp, emp_id
ORDER BY DIV_ID_01 DESC

Answer 2

Just for reference, if you try to use dynamic PIVOT here, you would end up needing a query that looks something like this.
SELECT [EMP],
[EMP_ID],
MIN(DIV_ID_1) [DIV_ID_1],
SUM(DIV_1_ALOC) [DIV_1_ALOC],
MIN(DIV_ID_2) [DIV_ID_2],
SUM(DIV_2_ALOC) [DIV_2_ALOC],
MIN(DIV_ID_3) [DIV_ID_3],
SUM(DIV_3_ALOC) [DIV_3_ALOC]
FROM (SELECT [EMP],
[EMP_ID],
[DIV_ID],
[ALLOCATION],
CONCAT('DIV_ID_',DENSE_RANK () OVER (PARTITION BY [EMP] ORDER BY [DIV_ID])) ID_RN,
CONCAT('DIV_',DENSE_RANK () OVER (PARTITION BY [EMP] ORDER BY [DIV_ID]),'_ALOC') ALLOC_RN
FROM EmpTable
) t
PIVOT (
MIN([DIV_ID])
FOR ID_RN IN ([DIV_ID_1],[DIV_ID_2],[DIV_ID_3]) ) p1
PIVOT (
SUM([ALLOCATION])
FOR ALLOC_RN IN ([DIV_1_ALOC],[DIV_2_ALOC],[DIV_3_ALOC]) ) p2
GROUP BY [EMP],
[EMP_ID]
You would need to dynamically create the SELECT and also the PIVOT columns because of the double pivot.
On the other hand, if you use CASE statements, you'll only need to dynamically create the SELECT, since that query would look like
SELECT [EMP],
[EMP_ID],
MIN(CASE WHEN RN = 1 THEN [DIV_ID] END) [DIV_ID_1],
SUM(CASE WHEN RN = 1 THEN [ALLOCATION] END) [DIV_1_ALOC],
MIN(CASE WHEN RN = 2 THEN [DIV_ID] END) [DIV_ID_2],
SUM(CASE WHEN RN = 2 THEN [ALLOCATION] END) [DIV_2_ALOC],
MIN(CASE WHEN RN = 3 THEN [DIV_ID] END) [DIV_ID_3],
SUM(CASE WHEN RN = 3 THEN [ALLOCATION] END) [DIV_3_ALOC]
FROM (
SELECT *,
DENSE_RANK () OVER (PARTITION BY [EMP] ORDER BY [DIV_ID]) RN
FROM EmpTable
) t
GROUP BY [EMP],
[EMP_ID]
Your dynamic statment would look something like
DECLARE #CaseSelect VARCHAR(MAX)
SELECT #CaseSelect = COALESCE(#CaseSelect + ',','')
+ 'MIN(CASE WHEN RN = ' + RN + ' THEN [DIV_ID] END) [DIV_ID_' + RN + '],'
+ 'SUM(CASE WHEN RN = ' + RN + ' THEN [ALLOCATION] END) [DIV_' + RN + '_ALOC]'
FROM (
SELECT DISTINCT CONVERT(VARCHAR(2),DENSE_RANK () OVER (PARTITION BY [EMP] ORDER BY [DIV_ID])) RN
FROM EmpTable
) t
ORDER BY RN
DECLARE #SQL VARCHAR(MAX)
SET #SQL = '
SELECT [EMP],
[EMP_ID], '
+ #CaseSelect + '
FROM (
SELECT *,
DENSE_RANK () OVER (PARTITION BY [EMP] ORDER BY [DIV_ID]) RN
FROM EmpTable
) t
GROUP BY [EMP], [EMP_ID]
'
EXEC(#SQL)
Replacing EmpTable with your actual table name ofcourse.