How to find overlapping periods recursively in SQL Server - sql

Question

I have this example dataset
ID StartDate EndDate
------------------------------
1 2014-01-05 2014-01-10
2 2014-01-06 2014-01-11
3 2014-01-07 2014-01-12
4 2014-01-08 2014-01-13
5 2014-01-09 2014-01-14
6 2014-01-26 2014-01-31
7 2014-01-27 2014-02-01
8 2014-01-28 2014-02-02
9 2014-01-29 2014-02-03
10 2014-01-30 2014-02-04
I want to select any rows that overlap a supplied period, and any rows those overlap with and so on.
So If I want to select any rows that over lap period 2014-01-06 to 2014-01-07
The following are directly overlapping (immediate overlaps)
1 2014-01-05 2014-01-10
2 2014-01-06 2014-01-11
But I also need rows that overlap with 1 and 2 (child overlap)
3 2014-01-07 2014-01-12
4 2014-01-08 2014-01-13
5 2014-01-09 2014-01-14
And if 3 to 5 had overlaps, return them too. but in this case, there are none.
This is my attempt so far but it has two problems I see that I don't know how to fix.
;WITH cte
AS
(
SELECT t.ID,
t.StartTime,
t.EndTime
FROM
dbo.Tasks AS t
UNION ALL
SELECT t.ID,
t.StartTime,
t.EndTime
FROM
dbo.Tasks AS t INNER JOIN
cte AS c ON t.StartTime < c.EndTime
AND t.EndTime > c.StartTime
)
SELECT * FROM cte AS a WHERE a.StartTime < #NewEnd
AND a.EndTime > #NewStart
When it goes to get overlapping periods of child overlap, the immediate overlap is re-included and causes an infinite recursion. And Secondly,
SELECT * FROM cte AS a WHERE a.StartTime < #NewEnd
AND a.EndTime > #NewStart
The where clause will filter out any recursively found overlaps.

Answer 1

I would first work out where the islands are in your data set, and only after that, work out which ones are overlapped by your query ranges:
declare #t table (ID int,StartDate date,EndDate date)
insert into #t(ID,StartDate,EndDate) values
(1 ,'20140105','20140110'),
(2 ,'20140106','20140111'),
(3 ,'20140107','20140112'),
(4 ,'20140108','20140113'),
(5 ,'20140109','20140114'),
(6 ,'20140126','20140131'),
(7 ,'20140127','20140201'),
(8 ,'20140128','20140202'),
(9 ,'20140129','20140203'),
(10 ,'20140130','20140204')
declare #Start date
declare #End date
select #Start='20140106',#End='20140107'
;With PotIslands as (
--Find ranges which aren't overlapped at their start
select StartDate,EndDate from #t t where
not exists (select * from #t t2 where
t2.StartDate < t.StartDate and
t2.EndDate >= t.StartDate)
union all
--Extend the ranges by any other ranges which overlap on the end
select pi.StartDate,t.EndDate
from PotIslands pi
inner join
#t t
on
pi.EndDate >= t.StartDate and pi.EndDate < t.EndDate
), Islands as (
select StartDate,MAX(EndDate) as EndDate from PotIslands group by StartDate
)
select * from Islands i where #Start <= i.EndDate and #End >= i.StartDate
Result:
StartDate EndDate
---------- ----------
2014-01-05 2014-01-14
If you need the individual rows, you can now join the selected islands back to the #t table for a simple range query.
This works because, for example, if any row within an island is ever included in a range, the entire remaining rows on an island will always also be included. So we find the islands first.

Answer 2

This should do it, you could put it into a function if needed and use a cross apply to join to another table. I did not test it but it should work with minimal (if any) errors.
declare #rt table
(
ID int not null,
StartTime date not null,
EndTime date not null
)
insert into #rt (ID, StartTime, EndTime)
select t.*
from Tasks t
where (#StartTime <= t.StartTime and #EndTime > t.StartTime)
or (#StartTime < t.EndTime and #EndTime >= t.EndTime)
declare #found int = ##rowcount
while #found > 0
begin
insert into #rt (ID, StartTime, EndTime)
select t.*
from Tasks t
left join #rt rt
on (rt.StartTime <= t.StartTime and rt.EndTime > t.StartTime)
or (rt.StartTime < t.EndTime and rt.EndTime >= t.EndTime)
where t.ID not in (select ID from #rt)
set #found = ##rowcount
end
select * from #rt