Reputation: 149
Let's say I have a table with data ranges
create table ranges (id number, date_from date, date_to date);
insert into ranges values (1, to_date('01.01.2017', 'dd.mm.rrrr'), to_date('03.01.2017', 'dd.mm.rrrr'));
insert into ranges values (2, to_date('05.02.2017', 'dd.mm.rrrr'), to_date('08.02.2017', 'dd.mm.rrrr'));
and my output should by one row for every date in these ranges
id | the_date
----------------
1 | 01.01.2017
1 | 02.01.2017
1 | 03.01.2017
2 | 05.02.2017
2 | 06.02.2017
2 | 07.02.2017
2 | 08.02.2017
But connect by gives me ORA-01436 Connect by Loop
SELECT connect_by_root(id), Trunc(date_from, 'dd') + LEVEL - 1 AS the_date
FROM ranges
CONNECT BY PRIOR id = id AND Trunc(date_from, 'dd') + LEVEL - 1 <= Trunc(date_to, 'dd')
ORDER BY id, the_date
What's wrong?
Upvotes: 1
Views: 603
Reputation: 191415
You can add a call to a non-deterministic function, e.g.
AND PRIOR dbms_random.value IS NOT NULL
So it becomes:
SELECT connect_by_root(id), Trunc(date_from, 'dd') + LEVEL - 1 AS the_date
FROM ranges
CONNECT BY PRIOR id = id
AND PRIOR dbms_random.value IS NOT NULL
AND Trunc(date_from, 'dd') + LEVEL - 1 <= Trunc(date_to, 'dd')
ORDER BY id, the_date;
CONNECT_BY_ROOT(ID) THE_DATE
------------------- ---------
1 01-JAN-17
1 02-JAN-17
1 03-JAN-17
2 05-FEB-17
2 06-FEB-17
2 07-FEB-17
2 08-FEB-17
7 rows selected.
There's an explanation of why it is necessary in this Oracle Community post; that uses sys_guid()
instead of dbms_random.value
, but the principle is the same.
If you're on 11gR2 or higher you could use recursive subquery factoring instead:
WITH rcte (root_id, the_date, date_to) AS (
SELECT id, date_from, date_to
FROM ranges
UNION ALL
SELECT root_id, the_date + 1, date_to
FROM rcte
WHERE the_date < date_to
)
SELECT root_id, the_date
FROM rcte
ORDER BY root_id, the_date;
ROOT_ID THE_DATE
---------- ---------
1 01-JAN-17
1 02-JAN-17
1 03-JAN-17
2 05-FEB-17
2 06-FEB-17
2 07-FEB-17
2 08-FEB-17
7 rows selected.
Upvotes: 3