Oracle Dynamic Range - sql

Question

So, i have this sample data:
Department | InitialDate | FinalDate
-------------------------------------------------------
1 | 01/01/2017 01:12:00 | 01/03/2017 00:00:08
1 | 01/03/2017 00:00:08 | 01/04/2017 05:00:01
1 | 01/04/2017 05:00:01 | 01/05/2017 02:00:00
2 | 01/05/2017 10:00:00 | 01/06/2017 11:00:08
2 | 01/06/2017 11:00:08 | 01/07/2017 04:04:00
3 | 01/07/2017 04:00:00 | 01/07/2017 15:00:22
1 | 01/07/2017 14:00:00 | 01/07/2017 18:00:08
1 | 01/07/2017 18:15:00 | 01/08/2017 22:00:00
3 | 01/12/2017 01:30:03 | 01/12/2017 18:00:00
1 | 01/13/2017 23:12:00 | 01/13/2017 23:59:08
and want to group it like this
Department | InitialDate | FinalDate
-------------------------------------------------------
1 | 01/01/2017 01:12:00 | 01/05/2017 02:00:00
2 | 01/05/2017 10:00:00 | 01/07/2017 04:04:00
3 | 01/07/2017 04:00:00 | 01/07/2017 15:00:22
1 | 01/07/2017 14:00:00 | 01/08/2017 22:00:00
3 | 01/12/2017 01:30:03 | 01/12/2017 18:00:00
1 | 01/13/2017 23:12:00 | 01/13/2017 23:59:08
I need to make groups by department and get the first and last date of each group, but the departments can repeat and for each time it occurs, I want the first and last date of that specific window. I already tried Analytic functions but nothing seems to work.

Answer 1

You can do it using the LAG analytic function to compare each row with the previous row:
SELECT department,
MIN( InitialDate ) AS InitialDate,
MIN( FinalDate ) AS FinalDate
FROM (
SELECT department,
InitialDate,
FinalDate,
SUM( grp_inc ) OVER ( ORDER BY FinalDate ) AS grp
FROM (
SELECT department,
InitialDate,
FinalDate,
CASE WHEN LAG( department ) OVER ( ORDER BY FinalDate ) = department
THEN 0
ELSE 1
END AS grp_inc
FROM table_name
)
)
GROUP BY department, grp

Answer 2

This is a type of "gaps-and-islands" problem. One method of solving it is by determining where groups of overlapping times start. Then use a cumulative sum to define each group:
select departmentid, min(initialdate), max(finaldate)
from (select t.*, sum(grp_starts) over (partition by departmentid order by initialdate) as grp
from (select t.*,
(case when exists (select 1
from t t2
where t2.departmentid = t.departmentid and
t.initialdate > t2.initialdate and
t.initialdate <= t2.finaldate
)
then 0 else 1
end) as grp_starts
from t
) t
) t
group by departmentid, grp;

Answer 3

Since you are looking for where the department changes and not where the department changes or the initialdate is not the same as the previous row's finaldate, you can use tabibitosan
WITH sample_data AS (SELECT 1 department, to_date('01/01/2017 01:12:00', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/03/2017 00:00:08', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 1 department, to_date('01/03/2017 00:00:08', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/04/2017 05:00:01', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 1 department, to_date('01/04/2017 05:00:01', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/05/2017 02:00:00', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 2 department, to_date('01/05/2017 10:00:00', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/06/2017 11:00:08', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 2 department, to_date('01/06/2017 11:00:08', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/07/2017 04:04:00', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 3 department, to_date('01/07/2017 04:00:00', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/07/2017 15:00:22', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 1 department, to_date('01/07/2017 14:00:00', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/07/2017 18:00:08', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 1 department, to_date('01/07/2017 18:15:00', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/08/2017 22:00:00', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 3 department, to_date('01/12/2017 01:30:03', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/12/2017 18:00:00', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual union all
SELECT 1 department, to_date('01/13/2017 23:12:00', 'mm/dd/yyyy hh24:mi:ss') initialdate, to_date('01/13/2017 23:59:08', 'mm/dd/yyyy hh24:mi:ss') finaldate from dual)
SELECT department,
MIN(initialdate) initialdate,
MAX(finaldate) finaldate
FROM (SELECT department,
initialdate,
finaldate,
row_number() OVER (ORDER BY initialdate)
- row_number() OVER (PARTITION BY department ORDER BY initialdate) grp
FROM sample_data sd)
GROUP BY department, grp
ORDER BY initialdate;
DEPARTMENT INITIALDATE FINALDATE
---------- ------------------- -------------------
1 01/01/2017 01:12:00 01/05/2017 02:00:00
2 01/05/2017 10:00:00 01/07/2017 04:04:00
3 01/07/2017 04:00:00 01/07/2017 15:00:22
1 01/07/2017 14:00:00 01/08/2017 22:00:00
3 01/12/2017 01:30:03 01/12/2017 18:00:00
1 01/13/2017 23:12:00 01/13/2017 23:59:08
This works by walking through and numbering all rows ordered by initial date and comparing them to walking through and numbering the rows for each department. When the department changes, the difference between the numbers changes. Where a department has consecutive rows in the initial dataset, the difference will remain the same for those rows. E.g. in your data, department 1 has 6 rows, the first 3 rows are the same as the first 3 rows of the initial data set, so the difference for those three rows is 0. The fourth and fifth department 1 rows are the 7th and 8th rows in the dataset, so the difference is 3 for those rows, etc.
This gives us a number that we can use, in conjunction with the department number, to group the data by. It's then a simple matter of finding the min/max dates within that group.