Get all overlapping date ranges when all overlap at the same time - sql

Question

I'm struggling with this for a few days... trying to write an SQL query to get all date ranges when all units overlap at the same time. It's better to see it graphically.
Here is the simplified table with the image for reference:
UnitId Start End
====== ========== ==========
1 05/01/2018 09/01/2018
1 10/01/2018 13/01/2018
2 04/01/2018 15/01/2018
2 19/01/2018 23/01/2018
3 06/01/2018 12/01/2018
3 14/01/2018 22/01/2018
Expected result:
Start End
====== ==========
06/01/2018 09/01/2018
10/01/2018 12/01/2018
What I currently have:
DECLARE #sourceTable TABLE (UnitId int, StartDate datetime, EndDate datetime);
INSERT INTO #sourceTable VALUES
(1, '2018-01-05', '2018-01-09')
,(1, '2018-01-10', '2018-01-13')
,(2, '2018-01-04', '2018-01-15')
,(2, '2018-01-19', '2018-01-23')
,(3, '2018-01-06', '2018-01-12')
,(3, '2018-01-14', '2018-01-22');
SELECT DISTINCT
(SELECT max(v) FROM (values(A.StartDate), (B.StartDate)) as value(v)) StartDate
,(SELECT min(v) FROM (values(A.EndDate), (B.EndDate)) as value(v)) EndDate
FROM #sourceTable A
JOIN #sourceTable B
ON A.startDate <= B.endDate AND A.endDate >= B.startDate AND A.UnitId != B.UnitId

Answer 1

I believe it is "count number of overlapping intervals" problem (this picture should help). Here is one solution to it:
DECLARE #t TABLE (UnitId INT, [Start] DATE, [End] DATE);
INSERT INTO #t VALUES
(1, '2018-01-05', '2018-01-09'),
(1, '2018-01-10', '2018-01-13'),
(2, '2018-01-04', '2018-01-15'),
(2, '2018-01-19', '2018-01-23'),
(3, '2018-01-06', '2018-01-12'),
(3, '2018-01-14', '2018-01-22');
WITH cte1(date, val) AS (
SELECT [Start], 1 FROM #t AS t
UNION ALL
SELECT [End], 0 FROM #t AS t
UNION ALL
SELECT DATEADD(DAY, 1, [End]), -1 FROM #t AS t
), cte2 AS (
SELECT date, SUM(val) OVER (ORDER BY date, val) AS usage
FROM cte1
)
SELECT date, MAX(usage) AS usage
FROM cte2
GROUP BY date
It will give you a list of all dates at which the use count (possibly) changed:
date usage
2018-01-04 1
2018-01-05 2
2018-01-06 3
2018-01-09 3
2018-01-10 3
2018-01-12 3
2018-01-13 2
2018-01-14 2
2018-01-15 2
2018-01-16 1
2018-01-19 2
2018-01-22 2
2018-01-23 1
2018-01-24 0
With this approach you do not need a calendar table or rCTE to build missing dates. Converting the above to ranges (2018-01-05 ... 2018-01-15, 2018-01-19 ... 2018-01-22 etc) is not very difficult.
DECLARE #t TABLE (UnitId INT, [Start] DATE, [End] DATE);
INSERT INTO #t VALUES
(1, '2018-01-05', '2018-01-09'),
(1, '2018-01-10', '2018-01-13'),
(2, '2018-01-04', '2018-01-15'),
(2, '2018-01-19', '2018-01-23'),
(3, '2018-01-06', '2018-01-12'),
(3, '2018-01-14', '2018-01-22');
WITH cte1(date, val) AS (
SELECT [Start], 1 FROM #t AS t -- starting date increments counter
UNION ALL
SELECT [End], 0 FROM #t AS t -- we need all edges in the result
UNION ALL
SELECT DATEADD(DAY, 1, [End]), -1 FROM #t AS t -- end date + 1 decrements counter
), cte2 AS (
SELECT date, SUM(val) OVER (ORDER BY date, val) AS usage -- running sum for counter
FROM cte1
), cte3 AS (
SELECT date, MAX(usage) AS usage -- group multiple events on same date together
FROM cte2
GROUP BY date
), cte4 AS (
SELECT date, usage, CASE
WHEN usage > 1 AND LAG(usage) OVER (ORDER BY date) > 1 THEN 0
WHEN usage < 2 AND LAG(usage) OVER (ORDER BY date) < 2 THEN 0
ELSE 1
END AS chg -- start new group if prev and curr usage are on opposite side of 1
FROM cte3
), cte5 AS (
SELECT date, usage, SUM(chg) OVER (ORDER BY date) AS grp -- number groups for each change
FROM cte4
)
SELECT MIN(date) date1, MAX(date) date2
FROM cte5
GROUP BY grp
HAVING MIN(usage) > 1
Result:
date1 date2
2018-01-05 2018-01-15
2018-01-19 2018-01-22

Answer 2

You are looking for date ranges where all units overlap. So look for start dates where all units exist and end dates where all units exist and then join the two.
I'm using ROW_NUMBER to join the first start date with the first end date, the second start date with the second end date and so on.
select s.startdate, e.enddate
from
(
select startdate, row_number() over (order by startdate) as rn
from #sourceTable s1
where
(
select count(*)
from #sourceTable s2
where s1.startdate between s2.startdate and s2.enddate
) = (select count(distinct unitid) from #sourceTable)
) s
join
(
select enddate, row_number() over (order by startdate) as rn
from #sourceTable s1
where
(
select count(*)
from #sourceTable s2
where s1.enddate between s2.startdate and s2.enddate
) = (select count(distinct unitid) from #sourceTable)
) e on e.rn = s.rn
order by s.startdate;
There may be more elegant ways to solve this, but I guess this query is at least easy to understand :-)
Rextester demo: https://rextester.com/GRRSW89045