sql server - How do you group by any time based interval?

Question

In SQL SERVER How do you group by any time based interval?

To save someone time I have come up with this solution, For me it works very well. You can generate any time base then group by any interval. Great for doing time weighted averages. If someone has a better way of doing this I would love to hear from you.

Hours

declare @startdate datetime2
declare @enddate datetime2
declare @interval int
set @startdate = '2017-01-01 00:00:00'
set @enddate   = '2017-01-31 00:00:00'
set @interval = 4 --Group by Every 4 hours
;with 
ALL_INTERVALS 
AS (
    SELECT TOP (DATEDIFF(HOUR,@startdate,@enddate))
    TIMES = DATEADD(HOUR,CONVERT(INT, ROW_NUMBER() OVER (ORDER BY s1.[object_id])),@startdate),
    1 AS VALUE
        FROM sys.all_objects AS s1
        CROSS JOIN
        sys.all_objects AS s2
)
select DATEADD(HOUR,((DATEDIFF(HOUR, @startdate,TIMES)/@interval)*@interval),@startdate) AS TIMES,SUM(VALUE) AS TESTDATA 
from ALL_INTERVALS
group by DATEADD(HOUR,((DATEDIFF(HOUR, @startdate,TIMES)/@interval)*@interval),@startdate)
order by DATEADD(HOUR,((DATEDIFF(HOUR, @startdate,TIMES)/@interval)*@interval),@startdate)

Minutes

Note. you can set your interval to 60 to achieve hours, 1440 to achieve days....

declare @startdate datetime2
declare @enddate datetime2
declare @interval int
set @startdate = '2017-01-01 00:00:00'
set @enddate   = '2017-01-31 00:00:00'
set @interval = 7
;with 
ALL_INTERVALS 
AS (
    SELECT TOP (DATEDIFF(MINUTE,@startdate,@enddate))
    TIMES = DATEADD(MINUTE,CONVERT(INT, ROW_NUMBER() OVER (ORDER BY s1.[object_id])),@startdate),
    1 AS VALUE
        FROM sys.all_objects AS s1
        CROSS JOIN
        sys.all_objects AS s2
)
select DATEADD(MINUTE,((DATEDIFF(MINUTE, @startdate,TIMES)/@interval)*@interval),@startdate) AS TIMES,SUM(VALUE) AS TESTDATA 
from ALL_INTERVALS
group by DATEADD(MINUTE,((DATEDIFF(MINUTE, @startdate,TIMES)/@interval)*@interval),@startdate)
order by DATEADD(MINUTE,((DATEDIFF(MINUTE, @startdate,TIMES)/@interval)*@interval),@startdate)

Answer 1

I think you are over complicating thins.
You can use GROUP BY (DATEDIFF(MINUTE, '2017-01-01', TheDateTime) / 30 for grouping by every 30 minutes. Of course, the date I've chosen is a just a random date. You can choose, if you want, the first (or last) date in your sample data.
And you can also use this technique to get every interval of any time part - just change the keyword MINUTE to any date part you want to use, and the intreval 30 to any interval you want.

Consider the following sample data:

;WITH CTE AS 
(
    SELECT CAST('2017-01-01T00:00:00' as datetime) As TheDateTime, 0 as rn
    UNION ALL
    SELECT DATEADD(MINUTE, 1, TheDateTime), rn + 1
    FROM CTE
    WHERE rn < 60
)

SELECT TheDateTime, rn INTO #T
FROM CTE
OPTION(MAXRECURSION 0)

#T now contains the following data:

TheDateTime                 rn
2017-01-01 00:00:00.000     0
2017-01-01 00:01:00.000     1
2017-01-01 00:02:00.000     2
2017-01-01 00:03:00.000     3
...
2017-01-01 00:59:00.000     59
2017-01-01 01:00:00.000     60

To get the maximum rn grouped by 30 minutes you just need this:

SELECT DATEDIFF(MINUTE, '2017-01-01', TheDateTime) / 30, MAX(rn)
FROM #T
GROUP BY DATEDIFF(MINUTE, '2017-01-01', TheDateTime) / 30

Results:

interval    max_rn
0           29
1           59
2           60