Records with overlapping dates - sql

Question

I have an addresses table, say:
address_id person_id start_date stop_date address
1 123 01-JAN-15 01-JUN-15 india
2 123 01-MAY-15 null russia
3 321 01-JAN-15 01-JUN-15 us
4 321 10-MAY-15 null india
I want to find all records (address_id values) which have overlapping dates for the same person_id. In this example that would find address_id 2 and 4, as May lies between Jan and Jun.
I then want to update the stop_date to start_date - 1 of the subsequent row belonging to same person so that the overlap is removed. For instance updating stop_date to 09-MAY-2015at row withaddress_id` 3.
So I want to end up with:
address_id person_id start_date stop_date address
1 123 01-JAN-15 30-APR-15 india
2 123 01-MAY-15 null russia
3 321 01-JAN-15 09-MAY-15 us
4 321 10-MAY-15 null india
I have tried:
update (
select * from addresses a1,addresses a2
where a1.person_id = a2.person_id
and a2.start_date > a1.start_date and a2.start_date <a1.stop_date
)
set a1.stop_date = a2.start_date - 1;
This worked fine in Microsoft Access but in Oracle it an invalid identifier error for a2.start_date.
How can I perform this update?

Answer 1

You can use a correlated update:
update addresses a
set stop_date = (
select min(start_date) - 1
from addresses
where person_id = a.person_id
and start_date > a.start_date
and start_date <= a.stop_date
)
where exists (
select null
from addresses
where person_id = a.person_id
and start_date > a.start_date
and start_date <= a.stop_date
);
2 rows updated.
select * from addresses;
ADDRESS_ID PERSON_ID START_DATE STOP_DATE ADDRESS
---------- ---------- ---------- --------- ----------
1 123 01-JAN-15 30-APR-15 india
2 123 01-MAY-15 russia
3 321 01-JAN-15 09-MAY-15 us
4 321 10-MAY-15 india
Both the set subquery and the exists subquery look for a row for the same person whose start date is between the start and stop date of the current row (which is the correlated part). The exists means only accounts which match are updated; without that any rows which don't have an overlap would be updated to null. (You wouldn't see any difference with the sample data, but would if you had more data).