Oracle - Join 2 periods of time - oracle

Question

I have 2 rows with 2 periods of time that intersect. For example:
---------------------------------------------
| START_DATE | END_DATE |
---------------------------------------------
| 01/01/2018 08:00:00 | 01/01/2018 09:30:00 |
| 01/01/2018 08:30:00 | 01/01/2018 10:00:00 |
---------------------------------------------
There are 30 minutes where both periods intersect. I want to avoid it. I would like to join both rows in one single column, taking the starting date as the older and the ending date as the newer:
---------------------------------------------
| START_DATE | END_DATE |
---------------------------------------------
| 01/01/2018 08:00:00 | 01/01/2018 10:00:00 |
---------------------------------------------
Have you any idea how can I get the solution I want with a SQL sentence?

Answer 1

For two rows just use greatest() and least(). But the problem is when you have many rows which may overlap in different ways. You can:
add row numbers to each row,
assign groups for overlapping periods using recursive query,
group data using this value and find min and max dates in each group.
dbfiddle demo
with
r(rn, start_date, end_date) as (
select row_number() over(order by start_date), start_date, end_date from t ),
c(rn, start_date, end_date, grp) as (
select rn, start_date, end_date, 1 from r where rn = 1
union all
select r.rn,
case when r.start_date <= c.end_date and c.start_date <= r.end_date
then least(r.start_date, c.start_date) else r.start_date end,
case when r.start_date <= c.end_date and c.start_date <= r.end_date
then greatest(r.end_date, c.end_date) else r.end_date end,
case when r.start_date <= c.end_date and c.start_date <= r.end_date
then grp else grp + 1 end
from c join r on r.rn = c.rn + 1)
select min(start_date), max(end_date) from c group by grp

Answer 2

If all you have is a set of date ranges, with no other correlating or constraining criteria, and you want to reduce that to a set of non overlapping ranges, you can do that with a recursive query like this one:
with recur(start_date, end_date) as (
select * from yourdata yd
where not exists (select 1 from yourdata cyd
where yd.start_Date between cyd.start_date and cyd.end_date
and (yd.start_date <> cyd.start_date or yd.end_date <> cyd.end_date))
union all
select r.start_date
, yd.end_date
from recur r
join yourdata yd
on r.start_date < yd.start_date
and yd.start_date <= r.end_date
and r.end_date < yd.end_date
)
select start_date, max(end_date) end_Date from recur group by start_Date;
In this query the anchor (the part before the union all) select all records whose start date is not contained in any other range.
The recursive part (the part after the union all) then select ranges that extend the current range. In both halves the original start date is returned while in the recursive part the new extended end date is returned. This results in a set of over lapping ranges with a common start date.
Finally the output query returns the start date and max end date grouped by start date.