SQL Transpose rows to columns without aggregate

问题

I know there is lot of answers but none of my use. My table looked like

where we store meter on and off status on given time. After every Off status there is an On status. We want to transpose table like this

    MeterNo |       [power off]       |          [power on]
____________________________________________________________________
  x0039938  | 2016-10-08 14:14:37.610 | 2016-10-08 14:17:15.047
____________________________________________________________________
  x0039938  | 2016-10-08 14:20:50.257 | NULL
____________________________________________________________________
  x0039938  | 2016-10-08 14:23:07.610 | 2016-10-08 14:23:17.920
____________________________________________________________________
  x0039940  | 2016-10-08 15:45:38.250 | 2016-10-08 15:52:40.080
____________________________________________________________________

and so on

I tried this query

    SELECT *
  FROM
(
  SELECT MeterNo, Status, CreatedOn
    FROM [BreakdownShutdowDetails]
) s
PIVOT
(
  MAX(CreatedOn) FOR Status IN ( [power off], [power on])
) p

which gives only on result per meterno like this

Basic idea to get time duration between Off on On status of each meter. Please suggest for right query. Thanks.

Update 1 Sample data

1   x0039938    power off   2016-10-08 14:14:37.610
2   x0039938    power on    2016-10-08 14:17:15.047
3   x0039938    power off   2016-10-08 14:20:50.257
4   x0039938    power off   2016-10-08 14:23:07.610
5   x0039938    power on    2016-10-08 14:23:17.920
6   x0039940    power off   2016-10-08 15:45:38.250
7   x0039940    power on    2016-10-08 15:52:40.080
8   x0040281    power off   2016-10-08 15:59:26.513
9   x0040281    power on    2016-10-08 16:20:23.323
10  x0039940    power off   2016-10-08 16:26:29.133
19  x0040281    power off   2016-10-08 17:17:48.900
22  x0039937    power off   2016-10-08 17:24:24.617
23  x0039937    power on    2016-10-08 17:24:38.590
24  x0039937    power off   2016-10-08 17:33:31.843
25  x0039937    power on    2016-10-08 17:35:47.470
27  x0039940    power off   2016-10-08 17:37:18.360
28  x0040281    power on    2016-10-08 17:40:08.093
30  x0043637    power off   2016-10-09 14:32:23.130
31  x0039937    power off   2016-10-09 14:32:24.893
32  x0040281    power off   2016-10-09 14:32:27.387
33  x0039940    power off   2016-10-09 14:32:29.407
34  x0040281    power on    2016-10-09 15:01:42.480

回答1:

It called commonly called as Gaps and islands problem. Try something like this (Not tested)

SELECT MeterNo,
       [power off] = Min(CASE
                           WHEN status = 'power off' THEN CreatedOn
                         END),
       [power on] = Min(CASE
                           WHEN status = 'power on' THEN CreatedOn
                         END)
FROM   (SELECT *,
                Row_number() OVER (partition BY MeterNo ORDER BY CreatedOn) - 
                Dense_rank() OVER (partition BY MeterNo ORDER BY status)  AS seq_grp
        FROM   Yourtable) t
GROUP  BY MeterNo,
          seq_grp

来源：https://stackoverflow.com/questions/41462218/sql-transpose-rows-to-columns-without-aggregate

标签

sql-server

gaps-and-islands