Time overlaping in sql code - sql

Question

I have got a table:
---------------------------------------------------
***************************************************
worker | job | starttime | endtime
---------------------------------------------------
w1 | j1 | 2014-01-02 08:00 | 2014-01-02 10:00
---------------------------------------------------
w1 | j2 | 2014-01-02 08:00 | 2014-01-02 11:00
---------------------------------------------------
w1 | j3 | 2014-01-02 09:00 | 2014-01-02 12:00
---------------------------------------------------
w1 | j4 | 2014-01-02 12:00 | 2014-01-02 13:00
---------------------------------------------------
w1 | j1 | 2014-01-02 12:00 | 2014-01-02 13:00
I want the result as
w1's j1 is 2 hours
w1's j2 is 1 hours
w1's j3 is 1 hours its like extra time
w1's j4 is 1 hours
For example, When preparing the cake, coat flour and eggs at the same time but coat eggs takes longer time.
finding total time is easy but how can I separate them ?
Is it possible code it in sql?

Answer 1

Have a look at DATEDIFF
Returns the count (signed integer) of the specified datepart
boundaries crossed between the specified startdate and enddate.
(This is also true for timeparts)
select worker,job, DATEDIFF(HOUR,endtime,starttime) as diff from mytable

Answer 2

I edited my answer to handle the case where the end time is equal to a previous end time.
So here's something like what you want. We take the max end time where
the job is not the current one and
the end time is less than the current row's end time
and compare that the current row's end time to determine how long the current row extends the previous job.
If that response is null, the row is the first row in the job, and in that case we just use the difference between start and end.
If the end time is equal to a previous end time, determined because the row_number is greater than one when partitioned over end time, we output zero as the time since that is the differential between this row and the previous one.
Please note that the "time" data type is limited to 24 hours so it is probably not the appropriate data type for the output. I just used it because it produces nice looking results for the times in the example.
declare #times table (
rowId int identity,
worker varchar(10),
job varchar(10),
starttime datetime,
endtime datetime
)
insert into #times (worker,job,starttime,endtime)
select 'w1','j1','2014-01-02 08:00','2014-01-02 9:25'
union select 'w1','j2','2014-01-02 08:00','2014-01-02 11:00'
union select 'w1','j3','2014-01-02 09:00','2014-01-02 12:00'
union select 'w1','j4','2014-01-02 12:00','2014-01-02 14:22'
union select 'w1','j5','2014-01-02 12:00','2014-01-02 14:22'
select *,
case when
ROW_NUMBER() over (partition by endtime order by starttime,endtime) > 1 then cast('0:00' as time)
else
coalesce(
cast((endtime - (select max(endtime) from #times where endtime < t.endtime)) as time),
cast((endtime - starttime) as time)
)
end as differential3
from #times t