SQL number of rows valid in a time range grouped by time - sql

Question

SQL Server
I have a table with 2 time stamps, time_start and time_end.
e.g.
ID time_start time_end
---- ------------------- -------------------
1 2019-01-01 08:30:00 2019-01-01 09:40:00
2 2019-01-01 09:10:24 2019-01-01 15:14:19
3 2019-01-01 09:21:15 2019-01-01 09:21:19
4 2019-01-01 10:39:45 2019-01-01 10:58:12
5 2019-01-01 11:39:45 2019-01-01 11:40:10
and I would like to group them so I can have the number of rows grouped by a variable time interval.
e.g.
time_interval row_count
------------------- ---------
2019-01-01 07:00:00 0
2019-01-01 08:00:00 1
2019-01-01 09:00:00 3
2019-01-01 10:00:00 2
2019-01-01 11:00:00 1
2019-01-01 12:00:00 0
My interval could be 1 hour, 1 minute, 30 minutes, 1 day, etc...
Think of it as a log-in/log-out situation, and I want to see how may users were logged at any given minute, hour, day, etc...

Answer 1

Try this,
DECLARE #start_date datetime='2019-01-01',
#end_date datetime='2019-01-02',
#i_minutes int=60
DECLARE #t TABLE
(
id int identity(1,1),time_start datetime,time_end datetime
)
INSERT INTO #t(time_start,time_end)VALUES
('2019-01-01 08:30:00','2019-01-01 09:40:00'),
('2019-01-01 09:10:24','2019-01-01 15:14:19'),
('2019-01-01 09:21:15','2019-01-01 09:21:19'),
('2019-01-01 10:39:45','2019-01-01 10:58:12'),
('2019-01-01 11:39:45','2019-01-01 11:40:10')
--SELECT #start_date=min(time_start),#end_date=max(time_end)
--FROM #t
;WITH CTE_time_Interval AS
(
SELECT #start_date AS time_int,#i_minutes AS i_minutes
UNION ALL
SELECT dateadd(minute,#i_minutes,time_int),i_minutes+ #i_minutes
FROM CTE_time_Interval
WHERE time_int<=#end_date
)
,CTE1 AS
(
SELECT ROW_NUMBER()OVER(ORDER BY time_int)AS r_no,time_int
FROM CTE_time_Interval
)
,CTE2 AS
(
SELECT a.time_int AS Int_start_time,b.time_int AS Int_end_time
FROM CTE1 a
INNER JOIN CTE1 b ON a.r_no+1=b.r_no
)
SELECT a.Int_start_time,a.Int_end_time,sum(iif(b.time_start is not null,1,0)) AS cnt
FROM CTE2 a
LEFT JOIN #t b ON
(
b.time_start BETWEEN a.Int_start_time AND a.Int_end_time
OR
b.time_end BETWEEN a.Int_start_time AND a.Int_end_time
OR
a.Int_start_time BETWEEN b.time_start AND b.time_end
OR
a.Int_end_time BETWEEN b.time_start AND b.time_end
)
GROUP BY a.Int_start_time,a.Int_end_time

Answer 2

Hi this is my workaround.
I created a table "test" with your data.
First I get the min and max intervals and after, I get all the intervals between these values with a CTE.
Finally, with this CTE and a left join with the intervals between time_start and time_end I got the answer.
This is for intervals of 1 hour
DECLARE #minDate AS DATETIME;
DECLARE #maxDate AS DATETIME;
SET #minDate = (select case
when (select min(time_start) from test) < (select min(time_end) from test)
then (select min(time_start) from test)
else (select min(time_end) from test) end )
SET #minDate = FORMAT(#minDate, 'dd-MM.yyyy HH:00:00')
SET #maxDate = (select case
when (select max(time_start) from test) > (select max(time_end) from test)
then (select max(time_start) from test)
else (select max(time_end) from test) end )
SET #maxDate = FORMAT(#maxDate, 'dd-MM.yyyy HH:00:00')
;WITH Dates_CTE
AS (SELECT #minDate AS Dates
UNION ALL
SELECT Dateadd(hh, 1, Dates)
FROM Dates_CTE
WHERE Dates < #maxDate)
SELECT d.Dates as time_interval, count(*) as row_count
FROM Dates_CTE d
LEFT JOIN test t on d.Dates
between (FORMAT(t.time_start, 'dd-MM.yyyy HH:00:00'))
and (FORMAT(t.time_end, 'dd-MM.yyyy HH:00:00'))
GROUP BY d.Dates
For intervals of 10 minutes you need some changes.
First I format dates getting minutes (dd-MM.yyyy HH:mm:00 instead of dd-MM.yyyy HH:00:00)
and in the left join I approach time_start and time_end to their 10 minutes time (9:32:00 in 9:30:00) (dateadd(minute, 10 * (datediff(minute, 0, time_start) / 10), 0)):
DECLARE #minDate AS DATETIME;
DECLARE #maxDate AS DATETIME;
SET #minDate = (select case
when (select min(time_start) from test) < (select min(time_end) from test)
then (select min(time_start) from test)
else (select min(time_end) from test) end )
SET #minDate = FORMAT(#minDate, 'dd-MM.yyyy HH:mm:00')
SET #maxDate = (select case
when (select max(time_start) from test) > (select max(time_end) from test)
then (select max(time_start) from test)
else (select max(time_end) from test) end )
SET #maxDate = FORMAT(#maxDate, 'dd-MM.yyyy HH:mm:00')
;WITH Dates_CTE
AS (SELECT #minDate AS Dates
UNION ALL
SELECT Dateadd(minute, 10, Dates)
FROM Dates_CTE
WHERE Dates < #maxDate)
SELECT d.Dates as time_interval, count(*) as row_count
FROM Dates_CTE d
LEFT JOIN test t on d.Dates
between dateadd(minute, 10 * (datediff(minute, 0, time_start) / 10), 0)
and dateadd(minute, 10 * (datediff(minute, 0, time_end) / 10), 0)
GROUP BY d.Dates
And finally I get this results for 1 hour intervals:
+---------------------+-----------+
| time_interval | row_count |
+---------------------+-----------+
| 01/01/2019 08:00:00 | 1 |
| 01/01/2019 09:00:00 | 3 |
| 01/01/2019 10:00:00 | 2 |
| 01/01/2019 11:00:00 | 2 |
| 01/01/2019 12:00:00 | 1 |
| 01/01/2019 13:00:00 | 1 |
| 01/01/2019 14:00:00 | 1 |
| 01/01/2019 15:00:00 | 1 |
+---------------------+-----------+
I hope it works for you.

Answer 3

You need to specify the time intervals. The rest is a LEFT JOIN/GROUP BY or correlated subquery:
with dates as (
select convert(datetime, '2019-01-01 07:00:00') as dt
union all
select dateadd(hour, 1, dt)
from dates
where dt < '2019-01-01 12:00:00'
)
select dates.dt, count(t.id)
from dates left join
t
on dates.dt < t.time_end and
dates.dt >= dateadd(hour, 1, t.time_start)
group by dates.dt
order by dates.dt;
If you have lots of data and lots of time periods, you might find that this has poor performance. If this is the case, ask a new question, with more information about sizing and performance.