T-SQL Start and end date times from a single column

Question

I have a single table of on/off events for different assets. I need to get a list of start & stop event times without using a cursor.

Source:

Item          


        

Answer 1

declare @YourTable table (Item varchar(10), EventDate datetime, Event varchar(10)) 

insert into @YourTable values 
('A','2011-10-03 00:01:00''On'), 
('B','2011-10-03 00:01:00','On'), 
('A','2011-10-03 00:02:00','Off'), 
('C','2011-10-03 00:01:00','On'), 
('B','2011-10-03 00:02:00','Off'), 
('A','2011-10-03 00:02:02','On'),   
('C','2011-10-03 00:02:05','Off'), 
('A','2011-10-03 00:02:07','Off'), 
('D','2011-10-03 00:02:02','On')   

;with a as
(
select item, eventdate, event, ROW_NUMBER() Over(Partition by Item order by EventDate) RN 
from @YourTable
),
b as (
select item, eventdate, 0 e, rn, event from a where rn = 1
union all
select b.item, a.eventdate, case when b.event = 'Off' then b.e+1 else b.e end, a.rn, a.event
from b 
join a on b.rn + 1 = a.rn and a.item = b.item
)
select item, min(eventdate) Start, max(eventdate) [End]
 from b
group by item, e
order by item, e

Answer 2

Untested...

;WITH myCTE AS
(
   SELECT *,
         ROW_NUMBER() OVER (PARTITION BY Item ORDER BY EventDate) AS rn
   FROM MyTable
)
SELECT
   M1.Item,
   M1.EventDate AS Start,
   M2.EventDate AS End
FROM
   myCTE M1
   JOIN
   myCTE M2 ON M1.Item = M2.Item AND M1.rn+1 = M2.rn
WHERE
   M1.Event = 'On'
   AND
   M2.Event = 'Off';

Answer 3

Here's the solution, with proof of concept for everyone to verify.

I noticed that the Off event for C was marked as a new On. I fixed that, but that also led me to coding a solution that would allow having an event that had started but not finished, so I included an open ended event D.

Additionally, my solution works with overlapping periods.

declare @YourTable table (Item varchar(10), 
                          EventDate datetime, 
                          Event varchar(10))

insert into @YourTable values
('A',         '2011-10-03 00:01:00',     'On'),
('B',         '2011-10-03 00:01:00',     'On'),
('A',         '2011-10-03 00:02:00',     'Off'),
('C',         '2011-10-03 00:01:00',     'On'),
('B',         '2011-10-03 00:02:00',     'Off'),
('A',         '2011-10-03 00:02:02',     'On'),  
('C',         '2011-10-03 00:02:05',     'Off'),
('A',         '2011-10-03 00:02:07',     'Off'),
('D',         '2011-10-03 00:02:02',     'On')  

select tOn.Item, tOn.EventDate Start, tOff.EventDate [End]
from (
select Item, EventDate, 
       ROW_NUMBER() Over(Partition by Item order by EventDate) EventID
from @YourTable where Event = 'On'
) tOn
LEFT JOIN (
select Item, EventDate, 
       ROW_NUMBER() Over(Partition by Item order by EventDate) EventID
from @YourTable where Event = 'Off'
) tOff
on (tOn.Item = tOff.Item and tOn.EventID = tOff.EventID)

Explained

We divide the data set in 2: On events and Off events. Each containing a numbered row which restarts when Item changes.

Basically, we have a first in first out: The first On will be closed by the first Off, so overlapping periods will be supported by this query given this approach. So each On Event for A will have its Event ID, which will be linked to the correspondent Off Event ID.

Open ended periods will be supported by the LEFT JOIN.