Dates Overlaps and Gap

Question

Hi Everyone,
Can you please help me solve this issue? I have dates overlapping and with gaps. I shouldn't have any overlaps, so I need to split them as below.
Here is my test data.
  create table test_dates (
    prov_id varchar2(20 byte),
    start_date date,
    end_date date,
attribute_value varchar2(10 byte)
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-01-01', date '2010-01-31', '1', 'A');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-02-01', date '2010-02-28', '1', 'B');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-04-01', date '2010-04-30', '1', 'C');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-04-15', date '2010-05-31', '1', 'D');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-07-01', date '2010-08-31', '1', 'E');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-07-01', date '2010-07-31', '1', 'F');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-08-15', date '2010-08-01', '1', 'G');
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-09-01', date '2010-12-31', '1', 'H');
commit;
Expected Results:
ID
START
END
1
1/1/2010
1/31/2010
1
2/1/2010
2/28/2010
1
4/1/2010
4/14/2010
1
4/15/2010
4/30/2010
1
5/1/2010
5/31/2010
1
7/1/2010
7/31/2010
1
8/1/2010
8/31/2010
1
9/1/2010
12/31/2010

Answer 1

Hi,
Depending on your requirements:
WITH change_points AS
    SELECT  prov_id
    ,       change_date
    ,       change_type
    ,       SUM  (change_type) OVER ( PARTITION BY  prov_id
                                      ORDER BY      change_date
                                    )  AS active_cnt
    ,       LEAD (change_date) OVER ( PARTITION BY  prov_id
                                      ORDER BY      change_date
                                    )  AS next_change_date
    ,       LEAD (change_type) OVER ( PARTITION BY  prov_id
                                      ORDER BY      change_date
                                    )  AS next_change_type
    FROM    test_dates
    UNPIVOT (    change_date
            FOR  change_type  IN ( start_date  AS  1
                                 , end_date    AS -1
SELECT    prov_id
,         change_date      + CASE
                                 WHEN  change_type = 1
                                 THEN  0
                                 ELSE  1
                             END   AS start_date
,         next_change_date - CASE
                                 WHEN  next_change_type = -1
                                 THEN  0
                                 ELSE  1
                             END   AS end_date
FROM      change_points
WHERE     active_cnt   > 0
AND       change_date  < next_change_date
ORDER BY  prov_id
,         start_date
Thanks for posting the sample data; that's helpful.
It also very helpful to explain how you get the results you want from the given data.  For example, what role does prov_id play in this problem?  (It's  hard to guess when every row in the sample data has the same value for prov_id.)  Also, sometimes, the start_dates or end_dates in the output are not exactly the dates in the table; sometimes they are 1 day earlier or later.  How do you decide which?
In the sample data, start_date can be later than end _date, for example:
insert into test_dates(start_date, end_date, prov_id, attribute_value) values (date '2010-08-15', date '2010-08-01', '1', 'G');
Was that a mistake?