Search code examples
sqlsql-serversubquerypivotcorrelated-subquery

SQL Server : pivot functionality, need to pivot a table


I have data in SQL Server in the format below.

-ID          ID2  status  time
-1384904453  417  stop    2013-11-19 23:40:43.000
-1384900211  417  start   2013-11-19 22:30:06.000
-1384822614  417  stop    2013-11-19 00:56:36.000
-1384813810  417  start   2013-11-18 22:30:06.000
-1384561199  417  stop    2013-11-16 00:19:45.000
-1384554623  417  start   2013-11-15 22:30:06.000
-1384475231  417  stop    2013-11-15 00:26:58.000
-1384468224  417  start   2013-11-14 22:30:06.000
-1384388181  417  stop    2013-11-14 00:16:20.000
-1384381807  417  start   2013-11-13 22:30:06.000
-1384300222  417  stop    2013-11-12 23:50:11.000
-1384295414  417  start   2013-11-12 22:30:06.000
-1384218209  417  stop    2013-11-12 01:03:17.000
-1384209015  417  start   2013-11-11 22:30:06.000

What I need is to be able to show data in the following format.

-ID2  start                      stop
-417  2013-11-19 22:30:06.000    2013-11-19 23:40:43.000
-417  2013-11-18 22:30:06.000    2013-11-19 00:56:36.000

Is it possible to do this? I tried pivot in SQL Server but it only returns one record. Can someone help please?


Solution

  • You can use the PIVOT function to get the result, I would just apply the row_number() windowing function to the data so you can return multiple rows for each ID2:

    select id2, start, stop
    from
    (
      select id2, status, time,
        row_number() over(partition by status
                          order by time) seq
      from yourtable
    ) d
    pivot
    (
      max(time)
      for status in (start, stop)
    ) piv
    order by start desc;
    

    See SQL Fiddle with Demo.

    You could also use an aggregate function with a CASE expression to get the final result:

    select
      id2,
      max(case when status = 'start' then time end) start,
      max(case when status = 'start' then time end) stop
    from 
    (
      select id2, status, time,
        row_number() over(partition by status
                          order by time) seq
      from yourtable
    ) d
    group by id2, seq;
    

    See SQL Fiddle with Demo