SQL Server: Turning rows into columns - sql

Question

I'm trying to make the following structure
ROW | GROUP | ORDER | COL_NAME | VALUE
---------------------------------------------
1 | 10 | 10 | FIRST_COL | Value1
2 | 10 | 10 | FIRST_COL | Value2
3 | 10 | 10 | FIRST_COL | Value3
4 | 10 | 20 | SECOND_COL| Val1
5 | 10 | 20 | SECOND_COL| Val2
6 | 20 | 10 | THIRD_COL | Opt3
...
into
FIRST_COL | SECOND_COL | THIRD_COL
-----------------------------------------------------
Value1 | Val1 | Opt3
Value2 | Val2 |
Value3 | |
What I currently have:
declare #cols varchar(max),
#query varchar(max)
select #cols = stuff((select ',' + quotename([COL_NAME])
from mt
group by [COL_NAME]
for xml path(''), type).value('.', 'varchar(max)'), 1, 1, '');
set #query = 'select ' + #cols + ' from (
select [COL_NAME], [VALUE]
from mt
) x
pivot (
min([VALUE])
for [COL_NAME] in (' + #cols + ')
) p
';
execute(#query);
The current code only shows the minimum values (since it's set to min([VALUES])), so only Value1, Val1 and Opt3 would be shown, but my question here is, how can I modify the code so I get the appropriate table/view?
Thanks!

Answer 1

I believe you need this:
set #query = 'select ' + #cols + ' from (
select [COL_NAME], [VALUE], ROW_NUMBER() OVER (PARTITION BY [COL_NAME] ORDER BY (SELECT NULL)) as seqnum
from mt
) x
pivot (
min([VALUE])
for [COL_NAME] in (' + #cols + ')
) p
';
The pivot uses all the columns in the subquery. You just need one to distinguish among the rows with the same column name.

Answer 2

Something like this?
Declare #YourTable table ([ROW] int,[GROUP] int,[ORDER] int,[COL_NAME] varchar(50),[VALUE] varchar(50))
Insert Into #YourTable Values
(1,10,10,'FIRST_COL' ,'Value1'),
(2,10,10,'FIRST_COL' ,'Value2'),
(3,10,10,'FIRST_COL' ,'Value3'),
(4,10,20,'SECOND_COL','Val1'),
(5,10,20,'SECOND_COL','Val2'),
(6,20,10,'THIRD_COL' ,'Opt3')
;with cteBase as (
Select *
,RowNr = Row_Number() over (Partition By [COL_NAME] Order by [ROW])
From #YourTable
)
Select First_Col = max(case when [COL_NAME]='FIRST_COL' then Value else '' end)
,Second_Col = max(case when [COL_NAME]='SECOND_COL' then Value else '' end)
,Third_Col = max(case when [COL_NAME]='THIRD_COL' then Value else '' end)
From cteBase
Group By RowNr
Returns
First_Col Second_Col Third_Col
Value1 Val1 Opt3
Value2 Val2
Value3

Answer 3

Try it like this:
DECLARE #tbl TABLE([ROW] INT,[GROUP] INT,[ORDER] INT,[COL_NAME] VARCHAR(100),VALUE VARCHAR(100));
INSERT INTO #tbl VALUES
(1,10,10,'FIRST_COL','Value1')
,(2,10,10,'FIRST_COL','Value2')
,(3,10,10,'FIRST_COL','Value3')
,(4,10,20,'SECOND_COL','Val1')
,(5,10,20,'SECOND_COL','Val2')
,(6,20,10,'THIRD_COL','Opt3');
WITH Sorted AS
(
SELECT tbl.*
,ROW_NUMBER() OVER(PARTITION BY tbl.[COL_NAME] ORDER BY tbl.[ROW]) AS SortInx
FROM #tbl AS tbl
)
SELECT MAX(CASE WHEN Sorted.[COL_NAME]='FIRST_COL' THEN Sorted.[VALUE] END) AS FIRST_COL
,MAX(CASE WHEN Sorted.[COL_NAME]='SECOND_COL' THEN Sorted.[VALUE] END) AS SECOND_COL
,MAX(CASE WHEN Sorted.[COL_NAME]='THIRD_COL' THEN Sorted.[VALUE] END) AS THIRD_COL
FROM Sorted
GROUP BY SortInx

Answer 4

As he stated, you need to use RowNumber() to populate your Row column, on your COL_NAME, so your data will look like this.
ROW | GROUP | ORDER | COL_NAME | VALUE
---------------------------------------------
1 | 10 | 10 | FIRST_COL | Value1
2 | 10 | 10 | FIRST_COL | Value2
3 | 10 | 10 | FIRST_COL | Value3
1 | 10 | 20 | SECOND_COL| Val1
2 | 10 | 20 | SECOND_COL| Val2
1 | 20 | 10 | THIRD_COL | Opt3
...
Then, just include the column in your SQL
set #query = 'select [Row], ' + #cols + ' from (
select [Row],[COL_NAME], [VALUE]
from mt
) x
pivot (
min([VALUE])
for [COL_NAME] in (' + #cols + ')
) p
';
execute(#query);
Then, you have
ROW FIRST_COL SECOND_COL THIRD_COL
1 Value1 Val1 Opt3
2 Value2 Val2 NULL
3 Value3 NULL NULL