How to find contiguous dates in numerous rows in SQL Server - sql

Question

We have a table with service provisions for people. For example:
id people_id dateStart dateEnd
1 1 28.07.14 19.07.16
2 2 14.04.15 16.02.16
3 2 16.02.16 18.04.16
4 2 18.04.16 27.06.16
5 2 27.06.16 19.07.16
6 2 19.07.16 NULL
7 3 24.02.12 17.06.12
8 3 23.07.12 19.09.12
9 3 18.08.14 NULL
10 4 28.06.15 NULL
11 5 19.01.16 NULL
I need to extract distinct people_id's (clients) with real start date of unfinished uninterrupted service that lasts more than year and then count days pass. 'Start date' and 'End date' of two different rows should be the same to be count as contiguous. One client can only have one unfinished service.
So the perfect result for the table above would be:
people_id dateStart lasts(days)
2 14.04.15 472
3 18.08.14 711
4 28.06.15 397
I didn't have problem with a single service:
SELECT
--some other columns from PEOPLE,
p.PEOPLE_ID,
s.DATESTART,
DATEDIFF(DAY, s.DATESTART, GETDATE()) as lasts
FROM
PEOPLE p
INNER JOIN service s on s.ID =
(
SELECT TOP 1 s2.ID
FROM service s2
WHERE s2.PEOPLE_ID = p.PEOPLE_ID
AND s2.DATESTART IS NOT NULL
AND s2.DATEEND IS NULL
ORDER BY s2.DATESTART DESC
)
WHERE
DATEDIFF(DAY, s.DATESTART , GETDATE()) >= 365
But I can't figure out how to determine contiguous services.

Answer 1

You can find where periods of "continuous" service begin by using lag(). Then a cumulative sum of this flag provides a group, which can be used for aggregation:
select people_id, min(datestart) as datestart,
(case when count(dateend) = count(*) then max(dateend) end) as dateend
from (select t.*,
sum(case when prev_dateend = datestart then 0 else 1 end) over
(partition by people_id order by datestart) as grp
from (select t.*,
lag(dateend) over (partition by people_id order by date_start) as prev_dateend
from t
) t
) t
group by people_id, grp
having count(*) > count(dateend);

Answer 2

Try this query:
select PeopleId, min(dateStart) as dateStart, sum(diff) as [lasts(days)] from
(
select P.*, datediff(day,datestart, DateEnd) as diff from
(select peopleId, dateStart,
isnull(dateend, cast(getdate() as date)) as DateEnd
from People
) P
where Dateend in
(select DateStart from People
where PeopleId = P.PeopleId)
or DateEnd = cast(getdate() as date ) -- check for continuous dates
) P1 group by PeopleId having sum(diff)> 365 --check for > one year
The comments in the query should explain things