SQL Double Pivot With Sum - sql

Question

I have a dataset in SQL server management studio 17 that requires some pivoting to collapse multiple rows into one row for ease of reading and saving space. Per unique id there are a set of codes which act as operators on the values, with the intention of displaying the results in a total column at the end. I've performed a primary pivot with one set of codes, but it looks like I need to do a second one to do further collapsing/amalgamation.
Here is some unpivoted data
PO_NUM u_id CODE_1 CODE_2 VALUE
---------------------------------------
0M274316 1 3 9;13 150
0N274316 1 9 9 200000
0O274316 1 6 9 210000
0P274316 1 6 13 21000
0Q274316 1 6 9 50000.5
0R274316 1 15 9 0
0M274317 2 3 9;13 150
0N274317 2 9 9 300000
0O274317 2 6 9 220000
0P274317 2 6 13 22000
0Q274317 2 6 8 90000.5
0R274317 2 15 8 0
With a sample pivot with the first unique id as an example I get:
PO_NUM uid CODE_1 CODE_2 6 9 15
-----------------------------------------------------
0M274316 1 3 9;13 210000 200000 0
0M274316 1 9 21000 NULL NULL
0M274316 1 13 50000.5 NULL NULL
What I need to do is another pivot to have '9;13' as columns with sums being used with the values. I have this initial pivot command as:
SELECT PO_NUM, CODE_1, CODE_2, VALUE from table
Pivot (max(VALUE) for CODE_2 in ([6],[9],[15])) pvt1
Which merges the data as expected when I don't include the CODE_1 vals. But trying to add the second pivot with CODE_2 just further messes the data. How do I pivot a second time using the pivoted data from the first command?
Expected output would be:
PO_NUM uid 6 9 15 9;13 9(2) 13 total
------------------------------------------------------------------------
0M274316 1 281000.5 200000 0 150 null null 481150.5
Basically they rows have been collapsed into one with the codes being broken out into columns. In reality the columns will be renamed to avoid conflicts with names, therefore I used 9(2) just to show the data is still there.
Sorry about the edits.

Answer 1

If I understand correctly, just use conditional aggregation:
select po_num, u_id, code_2,
max(case when code_1 = 3 then value end) as [3],
max(case when code_1 = 6 then value end) as [6],
max(case when code_1 = 9 then value end) as [9],
max(case when code_1 = 15 then value end) as [15]
from t
group by po_num, u_id, code_2;
I think this would be expressed using pivot as:
SELECT PO_NUM, u_id, CODE_2, [3], [6], [9], [15]
from table
Pivot (max(VALUE) for CODE_2 in ([3], [6], [9], [15])) pvt1;

Answer 2

You can try this.
DECLARE #MyTable TABLE(PO_NUM VARCHAR(10), u_id int, CODE_1 varchar(10), CODE_2 varchar(10), VALUE DECIMAL(18,2))
INSERT INTO #MyTable VALUES
('0M274316', 1, '3', '9;13', 150),
('0N274316', 1, '9', '9', 200000),
('0O274316', 1, '6', '9', 210000),
('0P274316', 1, '6', '13', 21000),
('0Q274316', 1, '6', '9', 50000.5),
('0R274316', 1, '15', '9', 0),
('0M274317', 2, '3', '9;13', 150),
('0N274317', 2, '9', '9', 300000),
('0O274317', 2, '6', '9', 220000),
('0P274317', 2, '6', '13', 22000),
('0Q274317', 2, '6', '8', 90000.5),
('0R274317', 2, '15', '8', 0)
SELECT *, [6] + [9] + [15] + [9;13] + [9(2)] + [13] as total FROM (
SELECT PO_NUM,u_id, CODE + ( CASE WHEN DENSE_RANK() OVER( PARTITION BY CODE ORDER BY C ) > 1 THEN '(' + CAST ( C AS varchar(50)) + ')' ELSE '' END ) CODE, VALUE
FROM (
SELECT MIN(PO_NUM) OVER(PARTITION BY u_id) PO_NUM, u_id, 1 C, CODE_1 CODE, VALUE FROM #MyTable
UNION ALL
SELECT MIN(PO_NUM) OVER(PARTITION BY u_id) PO_NUM, u_id, 2 C, CODE_2 CODE, VALUE from #MyTable
) T
) SRC
PIVOT(SUM(VALUE) FOR CODE IN ( [6], [9], [15], [9;13], [9(2)], [13])) PVT
Result:
PO_NUM u_id 6 9 15 9;13 9(2) 13 total
---------- ----------- ------------ ----------- --------- --------- ----------- ----------- -------------
0M274316 1 281000.50 200000.00 0.00 150.00 460000.50 21000.00 962151.00
0M274317 2 332000.50 300000.00 0.00 150.00 520000.00 22000.00 1174150.50