Oracle SQL: how to group same value in different group - sql

Question

Database:
Oracle Database 12c Release 12.2.0.1.0
Following is my test case script:
create table test
(
id number(1),
sdate date,
tdate date,
prnt_id number(1)
);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('10/17/2012','mm/dd/yyyy'), to_date('10/16/2014','mm/dd/yyyy'), 2);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('10/16/2014','mm/dd/yyyy'), to_date('2/16/2016','mm/dd/yyyy'), 2);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('2/16/2016','mm/dd/yyyy'), to_date('9/30/2016','mm/dd/yyyy'), 3);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('9/30/2016','mm/dd/yyyy'), to_date('3/16/2017','mm/dd/yyyy'), 3);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('3/16/2017','mm/dd/yyyy'), to_date('1/16/2019','mm/dd/yyyy'), 2);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('1/16/2019','mm/dd/yyyy'), to_date('10/16/2019','mm/dd/yyyy'), 2);
insert into test (id, sdate, tdate, prnt_id) values (1, to_date('10/16/2019','mm/dd/yyyy'), to_date('12/1/2999','mm/dd/yyyy'), 2);
commit;
select * from test order by sdate;
Question:
I want to modify the above Select SQL which returns all 7 rows from test table, selects all the columns plus two more columns.
First additional column (min_sdate) will return 10/17/2012 for rows 1,2 and 2/16/2016 for rows 3,4 and 3/16/2017 for rows 5,6,7.
Second additional column (max_tdate) will return 2/16/2016 for rows 1,2 and 3/16/2017 for rows 3,4 and 12/1/2999 for rows 5,6,7.
Basically, I'm trying to group by prnt_id column but instead of two groups (prnt_id: 2 and 3), I want three groups (prnt_id: 2,3,2), and then for those three groups get the min(sdate) and max(tdate).
I was thinking I could use analytical function min() and max() with window clause to achieve this, but not sure how to frame the SQL.
Any or all help will be appreciated. Thanks!

Answer 1

This is a form of gaps-and-islands. Assuming that the dates tile with no gaps, you can use the difference of row numbers to identify the islands:
select t.*,
min(sdate) over (partition by id, prnt_id, seqnum - seqnum_2),
max(edate) over (partition by id, prnt_id, seqnum - seqnum_2)
from (select t.*,
row_number() over (partition by id order by sdate) as seqnum,
row_number() over (partition by id, prnt_id order by sdate) as seqnum_2
from test t
) t;
Why this works is a little tricky to explain. But if you look at the results of the subquery, you will be able to see how the difference in row numbers defines the groups you want to define.