Sql Server - Count occurrences of datetime group by specific interval - sql

Question

I have this table with sample data:
MY_TABLE
------------------------------------------
ID DateVal other columns
------------------------------------------
1 2017-01-14 11:00:00 ...
2 2017-01-14 11:01:00 ...
3 2017-01-14 11:02:00 ...
4 2017-01-14 11:03:00 ...
5 2017-01-14 11:11:00 ...
6 2017-01-14 11:11:30 ...
7 2017-01-14 11:15:00 ...
8 2017-01-14 11:15:01 ...
9 2017-01-14 11:18:00 ...
I need to have this kind of result:
start end occurrences
-----------------------------------------------------------
2017-01-14 11:00 2017-01-14 11:05 4
2017-01-14 11:05 2017-01-14 11:10 0
2017-01-14 11:10 2017-01-14 11:15 3
2017-01-14 11:15 2017-01-14 11:20 2
...
In specific I need a query that extracts all the occurrences of raws in MY_TABLE in 5 mins range (range value is variable).
Someone could help me?
Best regards,

Answer 1

You need to generate the timeframes you want and then left join. Here is one method:
select v.dt, dateadd(minute, 5, v.dt) as end_dt, count(t.id)
from (values (convert(datetime, '2017-01-14 11:00')),
(convert(datetime, '2017-01-14 11:05')),
. . .
) v(dt) left join
my_table t
on t.dateval >= v.dt and
t.dateval < dateadd(minute, 5, v.dt)
group by v.dt;
Note: If you want to do this for a wider range of time, then using a tally table or recursive CTE is handy.

Answer 2

Let's have a row generator, that generates dates every rangeSize from a startDate:
DECLARE #rangeSize INT = 5;
DECLARE #startDate DATETIME = '2020-01-01 00:00';
WITH RG(D,D2) AS (
SELECT #startDate AS D, DATEADD(MINUTE, #rangeSize, #startDate) AS D2
UNION ALL
SELECT DATEADD(MINUTE, #rangeSize, D), DATEADD(MINUTE, #rangeSize, D2)
FROM RG a
WHERE D < DATEADD(MINUTE, #rangeSize * 100, #startDate)
)
SELECT D,D2
FROM RG
OPTION (MAXRECURSION 100);
Now let's hook it up to your data and count the data:
DECLARE #rangeSize INT = 5;
DECLARE #startDate DATETIME = '2020-01-01 00:00';
WITH RG(D,D2) AS (
SELECT #startDate AS D, DATEADD(MINUTE, #rangeSize, #startDate) AS D2
UNION ALL
SELECT DATEADD(MINUTE, #rangeSize, D), DATEADD(MINUTE, #rangeSize, D2)
FROM RG a
WHERE D < DATEADD(MINUTE, #rangeSize * 100, #startDate)
)
SELECT r.D as StartDate, r.D2 as EndDate, COUNT(m.ID) as Count as EndDate
FROM
RG r
LEFT JOIN
my_table m ON m.dateval > r.D AND m.dateval <= r.D2
GROUP BY r.D, r.D2
OPTION (MAXRECURSION 100);
Note I used > and <= for the range because you seem to classify from e.g 15:01 to 20:00 as a range, whereas it feels more natural to me to have 15:00 to 19:59 as "belonging to the 15-20 range"

Answer 3

If you don't need your start and end to reflect the exact range:
You can use something like DATEDIFF(minute, '2000-01-01', DateVal)/5 as your grouping, then use MIN(DateVal) and MAX(DateVal) for your start and end; but those values will be of the first and last transaction in the interval, not the bounds of the interval.
Alternatively, you can use a recursive CTE to generate the intervals, and then join that to your data:
; WITH intervals AS (
SELECT CAST('2017-11-01 00:00:00' AS DATETIME) AS `start`
, CAST ('2017-11-01 00:05:00' AS DATETIME) AS `end`
UNION ALL
SELECT DATEADD(minute, 5, `start`) AS, DATEADD(minute, 5, `end`) AS end
FROM intervals
WHERE intervals.end < '2017-11-02 00:00:00'
)
SELECT i.`start`, i.`end`, COUNT(t.ID) AS occurrences
FROM intervals AS i
INNER JOIN MY_TABLE AS t ON t.DateVal >= i.`start` AND t.DateVal < i.End
GROUP BY i.`start`, i.`end`
ORDER BY i.`start`, i.`end`
;
Notes:
the literal date values can be adjusted to reflect the actual range you want to query
If you want intervals without activity to be included, the INNER can be changed to LEFT
Since your question as stated has overlapping intervals, I went with the assumption that a DateVal on the 5 minute mark belongs to the interval that starts with that value.