Date range windowing query...I think

Question

Trying to do something I've not done before perhaps you can help or point me in the right direction.
I've got a simple table with events, something like name, start_date, end_date. These events can be duplicates, they can be distinct, they can have the same start_date and different names, or they can have the same name and start_date but different end_dates - any kind of overlap is allowed. There can also be gaps in the events where there is no event that covers a particular date. Really the only thing I can count on is that none of the values will be null. Basically a mess, as far as determining into which event's start-end date range a particular date falls, when there is overlap. However, that's what I need to do. The other wrinkle is that it's okay for me to ignore the end date of an event, for the purposes of gap-filling. That is to say, if there is no event that covers Jun 1, 2010, then I can 'choose' the event that ends prior to Jun 1, 2010.
I would like to, assuming some analytic function magic that I don't have, take this table and produce another that recomputes the start and end dates so that there is no overlap and the gaps are filled. As an overly simplistic example, given the input:
name start end
foo1 jan 1 jan 31
foo2 jan 15 jan 31
foo3 mar 1 mar 30
it would generate
foo1 jan 1 feb 28
foo3 mar 1 mar 30
or, alternatively
foo1 jan 1 jan 14
foo2 jan 15 feb 28
foo3 mar 1 mar 30
almost any reasonable elimination of overlaps is okay, so long as it makes sense. Are there any of the analytic functions that make this less painful?

Answer 1

I'm not sure what the PK of your table is and how "funny" the data can get, but your second result can be got by this:
create table test1 (
name varchar2(10),
startd date,
endd date);
insert into test1 values('foo1', to_date('20100101','yyyymmdd'),to_date('20100131','yyyymmdd'));
insert into test1 values('foo2', to_date('20100115','yyyymmdd'),to_date('20100131','yyyymmdd'));
insert into test1 values('foo3', to_date('20100301','yyyymmdd'),to_date('20100330','yyyymmdd'));
commit;
select name,to_char(startd,'mon dd') startd,
to_char(lead(startd -1,1,endd) over (order by startd),'mon dd')endd
from test1
order by startd;
or, alternatively
foo1 jan 1 jan 14
foo2 jan 15 feb 28
foo3 mar 1 mar 30foo1     jan 01     jan 14
foo2     jan 15     feb 28
foo3     mar 01     mar 30
- andy