Question

I’m looking for a way to count the distinct periods without gaps students have been absence:

Each time a student turns in an absence I create a record in the absence table, some absences can overlap, some prolong a previous absence.

StuId    StrPer       EndPer
------   -----------  -----------
111111   2011-01-10   2011-01-15
222222   2011-02-01   2011-02-05
222222   2011-02-06   2011-02-08
333333   2011-04-07   2011-04-14
444444   2011-04-20   2011-04-25
444444   2011-04-23   2011-04-28
111111   2011-05-01   2011-05-03

Now I want to count the number of unique absence periods with gaps, the result should be:

StuId   NbrAbs
------  ------
111111  2
222222  1
333333  1
444444  1

111111 has two absence period with a gap between

222222 also has two absence periods but without gap, so it must be considered as 1 absence period

333333 has only 1 absence period

444444 has 2 periods that overlap, also without gap, so it must be considered as 1 absence period

Can anyone help me write a query for this?

Answer 1

Assuming SQL 2005+ this should work:

SELECT '111111' as stuid,'2011-01-10' as start_date,'2011-01-15' as end_date into #data UNION ALL
SELECT '222222','2011-02-01','2011-02-05' UNION ALL
SELECT '222222','2011-02-06','2011-02-08' UNION ALL
SELECT '333333','2011-04-07','2011-04-14' UNION ALL
SELECT '444444','2011-04-20','2011-04-25' UNION ALL
SELECT '444444','2011-04-23','2011-04-28' UNION ALL
SELECT '111111','2011-05-01','2011-05-03' 

;with periods as
(
select 
stuid
,start_date
,end_date
,row_number() OVER (PARTITION BY stuid ORDER BY end_date ASC) as period_number
FROM #data
)
,periods2 AS
(
SELECT 
p1.stuid
,p1.start_date
,p1.end_date
,p1.period_number
,ISNULL(DATEDIFF(DD,p1.end_date,p2.start_date),1) as period_gap
from periods p1
LEFT OUTER JOIN periods p2 on p2.stuid = p1.stuid
AND p2.period_number = p1.period_number + 1
)
SELECT 
stuid
,count(period_gap) as number_discrete_absences
FROM periods2
WHERE period_gap > 0
GROUP BY stuid