SQL Server Pivot multiple columns based on one col

Question

I have following source and destination tables in sql server 2008R2. How can I do pivot(s) in TSQL to get to the destination from source.

SourceTbl

empId    empIndex    empState    empStDate    empEndDate
========================================================
10        1           AL          1/1/2012     12/1/2012
10        2           FL          2/1/2012     2/1/2013
15        1           FL          3/20/2012    1/1/2099

DestTbl

empId    empState1  empState1StDate    empState1EndDt    empState2  empState2StDate    empState2EndDt
=========================================================================================================
10        AL         1/1/2012           12/1/2012         FL         2/1/2012           2/1/2013
15        FL         3/20/2012          1/1/2099          NULL       NULL               NULL

Hoping that the empIndex will somehow help in the pivot.

Answer 1

Wow this was more complicated than i imagined, but I did get it to work great! thanks. This was my final version. The TextKey contains the data you want to turn into columns, and the TextValue is the value that ends up inside each cell.

DECLARE @cols AS NVARCHAR(MAX),
        @query AS NVARCHAR(MAX)


select @cols = STUFF((SELECT distinct ', ' + QUOTENAME(TextKey) 
                    from #SourceTbl
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

set @query = 'SELECT FromEntityID, DisplayName, ' + @cols + ' 
              FROM 
              (
                  select FromEntityID, DisplayName, TextKey, TextValue
                  from #SourceTbl
              ) x
              pivot 
              (
                  min(TextValue)
                  for TextKey in (' + @cols + ')
              ) p 
              ORDER BY FromEntityID
              '

execute(@query)

Answer 2

Since you are using SQL Server there are several different ways that you can convert the rows into columns. You can use an aggregate function with a CASE expression:

select empid,
  max(case when empindex = 1 then empstate end) empState1,
  max(case when empindex = 1 then empStDate end) empStDate1,
  max(case when empindex = 1 then empEndDate end) empEndDate1,
  max(case when empindex = 2 then empstate end) empState2,
  max(case when empindex = 2 then empStDate end) empStDate2,
  max(case when empindex = 2 then empEndDate end) empEndDate2
from sourcetbl
group by empid;

See SQL Fiddle with Demo.

If you want to use the PIVOT function to get the result, then I would recommend first unpivoting the columns empState, empStDate and empEndDate so you will have multiple rows first. You can use the UNPIVOT function or CROSS APPLY to convert the data the code will be:

select empid, col+cast(empindex as varchar(10)) col,  value
from sourcetbl
cross apply
(
  select 'empstate', empstate union all
  select 'empstdate', convert(varchar(10), empstdate, 120) union all
  select 'empenddate', convert(varchar(10), empenddate, 120)
) c (col, value);

See Demo. Once the data is unpivoted, then you can apply the PIVOT function so the final code will be:

select empid,
  empState1, empStDate1, empEndDate1,
  empState2, empStDate2, empEndDate2
from 
(
  select empid, col+cast(empindex as varchar(10)) col,  value
  from sourcetbl
  cross apply
  (
    select 'empstate', empstate union all
    select 'empstdate', convert(varchar(10), empstdate, 120) union all
    select 'empenddate', convert(varchar(10), empenddate, 120)
  ) c (col, value)
) d
pivot
(
  max(value)
  for col in (empState1, empStDate1, empEndDate1,
              empState2, empStDate2, empEndDate2)
) piv;

See SQL Fiddle with Demo.

Th above versions will work great if you have a limited number of empindex, but if not then you can use dynamic SQL:

DECLARE @cols AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT ',' + QUOTENAME(col+cast(empindex as varchar(10))) 
                    from SourceTbl
                    cross apply
                    (
                      select 'empstate', 1 union all
                      select 'empstdate', 2 union all
                      select 'empenddate', 3
                    ) c (col, so)
                    group by col, so, empindex
                    order by empindex, so
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

set @query = 'SELECT empid,' + @cols + ' 
            from 
            (
                select empid, col+cast(empindex as varchar(10)) col,  value
                from sourcetbl
                cross apply
                (
                  select ''empstate'', empstate union all
                  select ''empstdate'', convert(varchar(10), empstdate, 120) union all
                  select ''empenddate'', convert(varchar(10), empenddate, 120)
                ) c (col, value)
            ) x
            pivot 
            (
                max(value)
                for col in (' + @cols + ')
            ) p '

execute sp_executesql @query;

See SQL Fiddle with Demo

You can use these queries to INSERT INTO your DestTbl, or instead of storing the data in this format, you now have a query to get the desired result.

These queries place the data in the format:

| EMPID | EMPSTATE1 | EMPSTDATE1 | EMPENDDATE1 | EMPSTATE2 | EMPSTDATE2 | EMPENDDATE2 |
---------------------------------------------------------------------------------------
|    10 |        AL | 2012-01-01 |  2012-12-01 |        FL | 2012-02-01 |  2013-02-01 |
|    15 |        FL | 2012-03-20 |  2099-01-01 |    (null) |     (null) |      (null) |