Splitting time range based on activities in sql - sql

Question

I need a Oracle sql to show the following output given the sample input.
Basically, an employee is schedule for a 9 hour shift.
I need to split up the activities during the day to separate records.
Especially the general activity of Cash. I need to create new records.
Activity start time end time
Shift 2010-01-01 8:00:00 2010-01-01 17:00:00
Open 2010-01-01 8:00:00 2010-01-01 9:00:00
Cash 2010-01-01 9:00:00 2010-01-01 16:00:00
Break 2010-01-01 10:00:00 2010-01-01 10:15:00
Lunch 2010-01-01 12:00:00 2010-01-01 13:00:00
Break 2010-01-01 14:30:00 2010-01-01 14:45:00
Close 2010-01-01 16:00:00 2010-01-01 17:00:00
OUTPUT:
Activity start time end time
Open 2010-01-01 8:00:00 2010-01-01 9:00:00
Cash 2010-01-01 9:00:00 2010-01-01 10:00:00
Break 2010-01-01 10:00:00 2010-01-01 10:15:00
Cash 2010-01-01 10:15:00 2010-01-01 12:00:00
Lunch 2010-01-01 12:00:00 2010-01-01 13:00:00
Cash 2010-01-01 13:00:00 2010-01-01 14:30:00
Break 2010-01-01 14:30:00 2010-01-01 14:45:00
Cash 2010-01-01 14:45:00 2010-01-01 16:00:00
Close 2010-01-01 16:00:00 2010-01-01 17:00:00
Any help is greatly appreciated.

Answer 1

This is a kind of gaps-and-islands problem. Assuming there is always an 'Open' record whose start matches the 'Shift' start, and a 'Close' record whose end matches the 'Shift' end; and that the general activity is always 'Cash' and its start matches the 'Open' end and its end matches the 'Close' start; then some of those records are redundant when filling in the gaps.
You can use the lead and lag functions to generate dummy 'Cash' records that sit between all the other activities, looking both forward and behind:
select activity orig_activity, start_time orig_start, end_time orig_end,
'Cash' as activity, lag(end_time) over (order by end_time) as start_time, start_time as end_time
from table1
where activity not in ('Shift', 'Cash')
union all
select activity orig_activity, start_time orig_start, end_time orig_end,
'Cash' as activity, end_time as start_time, lead(start_time) over (order by start_time) as end_time
from table1
where activity not in ('Shift', 'Cash')
order by orig_start;
ORIG_ ORIG_START ORIG_END ACTI START_TIME END_TIME
----- ------------------- ------------------- ---- ------------------- -------------------
Open 2010-01-01 08:00:00 2010-01-01 09:00:00 Cash 2010-01-01 08:00:00
Open 2010-01-01 08:00:00 2010-01-01 09:00:00 Cash 2010-01-01 09:00:00 2010-01-01 10:00:00
Break 2010-01-01 10:00:00 2010-01-01 10:15:00 Cash 2010-01-01 09:00:00 2010-01-01 10:00:00
Break 2010-01-01 10:00:00 2010-01-01 10:15:00 Cash 2010-01-01 10:15:00 2010-01-01 12:00:00
Lunch 2010-01-01 12:00:00 2010-01-01 13:00:00 Cash 2010-01-01 10:15:00 2010-01-01 12:00:00
Lunch 2010-01-01 12:00:00 2010-01-01 13:00:00 Cash 2010-01-01 13:00:00 2010-01-01 14:30:00
Break 2010-01-01 14:30:00 2010-01-01 14:45:00 Cash 2010-01-01 14:45:00 2010-01-01 16:00:00
Break 2010-01-01 14:30:00 2010-01-01 14:45:00 Cash 2010-01-01 13:00:00 2010-01-01 14:30:00
Close 2010-01-01 16:00:00 2010-01-01 17:00:00 Cash 2010-01-01 17:00:00
Close 2010-01-01 16:00:00 2010-01-01 17:00:00 Cash 2010-01-01 14:45:00 2010-01-01 16:00:00
That has duplicates from the same gap being seen, for instance, after the break and before lunch. By ignoring the original values you can remove those with distinct, or with union instead of union all. You can also exclude any generated rows with null start or end times, and any that overlap with other records - which could happen if two other activities were contiguous:
select activity, start_time, end_time from (
select 'Cash' as activity,
lag(end_time) over (order by end_time) as start_time,
start_time as end_time
from table1
where activity not in ('Shift', 'Cash')
union
select 'Cash' as activity,
end_time as start_time,
lead(start_time) over (order by start_time) as end_time
from table1
where activity not in ('Shift', 'Cash')
) tmp
where start_time is not null
and end_time is not null
and not exists (
select null from table1 where activity not in ('Shift', 'Cash') and (start_time = tmp.start_time or end_time = tmp.end_time)
)
order by start_time;
ACTI START_TIME END_TIME
---- ------------------- -------------------
Cash 2010-01-01 09:00:00 2010-01-01 10:00:00
Cash 2010-01-01 10:15:00 2010-01-01 12:00:00
Cash 2010-01-01 13:00:00 2010-01-01 14:30:00
Cash 2010-01-01 14:45:00 2010-01-01 16:00:00
You can then union that with all the original table rows, except the 'Cash' record:
...
union all
select activity, start_time, end_time
from table1
where activity not in ('Shift', 'Cash')
order by start_time;
ACTIV START_TIME END_TIME
----- ------------------- -------------------
Open 2010-01-01 08:00:00 2010-01-01 09:00:00
Cash 2010-01-01 09:00:00 2010-01-01 10:00:00
Break 2010-01-01 10:00:00 2010-01-01 10:15:00
Cash 2010-01-01 10:15:00 2010-01-01 12:00:00
Lunch 2010-01-01 12:00:00 2010-01-01 13:00:00
Cash 2010-01-01 13:00:00 2010-01-01 14:30:00
Break 2010-01-01 14:30:00 2010-01-01 14:45:00
Cash 2010-01-01 14:45:00 2010-01-01 16:00:00
Close 2010-01-01 16:00:00 2010-01-01 17:00:00
This also assumes that activities never overlap, but non-'Cash' activities could be adjacent.
There are probably other gaps-and-islands approaches that would work too.

Answer 2

I'll second Alex's answer. But, just for something completely different, you could figure out the distinct seconds in the shift, figure out what the person was doing each second, then group those into ranges for your results.
I think this would be less efficient than Alex's approach, but might be more flexible: it doesn't assume as much about how the input data will look.
with shift_data ( activity, start_time, end_time ) AS
-- This is just test data that would be in your database table
(
SELECT 'Shift',to_date('2010-01-01 8:00:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 17:00:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL UNION ALL
SELECT 'Open',to_date('2010-01-01 8:00:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 9:00:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL UNION ALL
SELECT 'Cash',to_date('2010-01-01 9:00:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 16:00:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL UNION ALL
SELECT 'Break',to_date('2010-01-01 10:00:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 10:15:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL UNION ALL
SELECT 'Lunch',to_date('2010-01-01 12:00:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 13:00:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL UNION ALL
SELECT 'Break',to_date('2010-01-01 14:30:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 14:45:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL UNION ALL
SELECT 'Close',to_date('2010-01-01 16:00:00','YYYY-MM-DD HH24:MI:SS'),to_date('2010-01-01 17:00:00','YYYY-MM-DD HH24:MI:SS') FROM DUAL
),
seconds_in_shift as (
-- Step 1: get a list of every second that falls in the shift
SELECT start_time + (ROWNUM - 1) / 86400 second
FROM shift_data
WHERE activity = 'Shift'
CONNECT BY ROWNUM <= ( (end_time - start_time) * 86400) + 1),
activity_each_second as (
-- Step 2: figure out what the person was doing every second. If multiple
-- activities overlap, choose whichever one had the shortest duration
-- Also, mark which seconds represent a transition from one activity to
-- another ("marker" column)
SELECT second,
MAX (activity) KEEP (DENSE_RANK FIRST ORDER BY end_time - start_time) activity,
CASE WHEN MAX (activity) KEEP (DENSE_RANK FIRST ORDER BY end_time - start_time)
!= NVL(LAG(MAX (activity) KEEP (DENSE_RANK FIRST ORDER BY end_time - start_time))
OVER ( PARTITION BY NULL ORDER BY SECOND),'#NULL#') THEN 'Y' ELSE NULL END marker
FROM seconds_in_shift ss
INNER JOIN shift_data sd ON ss.second BETWEEN sd.start_time AND sd.end_time
GROUP BY second),
ranges as (
-- Step 3: count the number of marker columns from the beginning of the shift
-- to the current second. Call this "activity_number".
select aes.*,
count(marker) OVER ( PARTITION BY NULL ORDER BY second) activity_number
from activity_each_second aes )
-- Finally, show the activity, start, and end time for each activity_number
SELECT activity,
round(min(second),'MI') start_time,
round(max(second),'MI') end_time
FROM ranges
GROUP BY activity, activity_number
ORDER BY activity_number;
Results:
Open 1/1/2010 8:00:00 A 1/1/2010 9:00:00 AM
Cash 1/1/2010 9:00:00 A 1/1/2010 10:00:00 AM
Break 1/1/2010 10:00:00 1/1/2010 10:15:00 AM
Cash 1/1/2010 10:15:00 1/1/2010 12:00:00 PM
Lunch 1/1/2010 12:00:00 1/1/2010 1:00:00 PM
Cash 1/1/2010 1:00:00 P 1/1/2010 2:30:00 PM
Break 1/1/2010 2:30:00 P 1/1/2010 2:45:00 PM
Cash 1/1/2010 2:45:00 P 1/1/2010 4:00:00 PM
Close 1/1/2010 4:00:00 P 1/1/2010 5:00:00 PM
NOTE: I cheated a bit by rounding the times to the nearest minute. Without rounding, there would be overlap in the ranges. E.g., 4PM on-the-dot would either be "Cash" or "Close", it wouldn't be both.

Answer 3

Assumptions: All activity intervals fall within the 'Shift' interval, and two activity intervals may have at most an endpoint in common (they may be adjacent - but they can't overlap in any way).
Also, I assumed you may have more than one employee in your table (so that must be addressed), and that the computation must be done separately for each calendar day. You will see this in the input data, and handled in the query.
Here is a way to get the desired result using only the analytic lag() function. It first collects only the activities different from 'Shift' and 'Cash', then it fills the gaps with 'Cash' (including at the beginning and/or end of the 'Shift', if no specific activity, like 'Open' or 'Close', starts or ends at the beginning or the end of a 'Shift'). The 'Shift' interval in particular, as presented in the inputs, is not particularly helpful in this solution; you will see how I handle that in the CTE I called prep below.
So I don't need to enter nls_date_format everywhere, I first ran
alter session set nls_date_format = 'yyyy-mm-dd hh24:mi:ss'
Then:
with
table1 ( empno, activity, start_time, end_time ) as (
select 101, 'Shift', to_date('2010-01-01 8:00:00') , to_date('2010-01-01 17:00:00') from dual union all
select 101, 'Open' , to_date('2010-01-01 8:00:00') , to_date('2010-01-01 9:00:00') from dual union all
select 101, 'Cash' , to_date('2010-01-01 9:00:00') , to_date('2010-01-01 16:00:00') from dual union all
select 101, 'Break', to_date('2010-01-01 10:00:00'), to_date('2010-01-01 10:15:00') from dual union all
select 101, 'Lunch', to_date('2010-01-01 12:00:00'), to_date('2010-01-01 13:00:00') from dual union all
select 101, 'Break', to_date('2010-01-01 14:30:00'), to_date('2010-01-01 14:45:00') from dual union all
select 101, 'Close', to_date('2010-01-01 16:00:00'), to_date('2010-01-01 17:00:00') from dual
),
prep ( empno, activity, start_time, end_time, flag ) as (
select empno, activity, start_time, end_time, 1
from table1
where activity not in ('Shift', 'Cash')
union all select empno, 'Shift', start_time, start_time, 0
from table1
where activity = 'Shift'
union all select empno, 'Shift', end_time, end_time, 2
from table1
where activity = 'Shift'
),
with_cash_intervals ( empno, activity, start_time, end_time ) as (
select empno, activity, start_time, end_time
from prep
where activity != 'Shift'
union all
select empno, 'Cash', lag(end_time) over (partition by empno, trunc(start_time)
order by flag, start_time), start_time
from prep
)
select empno, activity, start_time, end_time
from with_cash_intervals
where start_time < end_time
order by empno, start_time -- if needed
Output:
EMPNO ACTIVITY START_TIME END_TIME
----- -------- ------------------- -------------------
101 Open 2010-01-01 08:00:00 2010-01-01 09:00:00
101 Cash 2010-01-01 09:00:00 2010-01-01 10:00:00
101 Break 2010-01-01 10:00:00 2010-01-01 10:15:00
101 Cash 2010-01-01 10:15:00 2010-01-01 12:00:00
101 Lunch 2010-01-01 12:00:00 2010-01-01 13:00:00
101 Cash 2010-01-01 13:00:00 2010-01-01 14:30:00
101 Break 2010-01-01 14:30:00 2010-01-01 14:45:00
101 Cash 2010-01-01 14:45:00 2010-01-01 16:00:00
101 Close 2010-01-01 16:00:00 2010-01-01 17:00:00
9 rows selected.