我正在尝试将开始时间和结束时间按整小时计时，并计算每小时使用的秒数。以整小时计算，我的意思是例如15:00:00到16:00:00。由于在午夜和跨越行的边界问题，我无法找出在SQL中执行此操作的方法。我有以下输入和输出：
Table data:
Start End
2020-03-06 15:30:40.000 2020-03-06 17:09:01.000
2020-03-06 22:47:52.000 2020-03-06 23:48:52.000
2020-03-06 23:49:52.000 2020-03-07 00:47:52.000
2020-03-09 17:05:26.000 2020-03-09 18:05:26.000
2020-03-09 18:05:32.000 2020-03-09 19:05:26.000
Columns in the output are: Year Month Day Hour SecondsInFullHour
Output (query on 2020-03-06)
.
.
.
2020 03 06 12 0
2020 03 06 13 0
2020 03 06 14 0
2020 03 06 15 1760
2020 03 06 16 3600
2020 03 06 17 541
2020 03 06 18 0
2020 03 06 19 0
2020 03 06 20 0
2020 03 06 21 0
2020 03 06 21 0
2020 03 06 22 728
2020 03 06 23 3540 (truncated due to day limit of 00:00:00 the following day)
Output (query on 2020-03-07)
2020 03 07 00 2872 (calculated from row 3)
Output (query on 2020-03-09)
2020 03 09 17 3274
2020 03 09 18 3594
2020 03 09 19 326
我想像上面的示例一样在特定的日期进行查询（后来我也将尝试每周执行一次，并在全天执行相同的操作），但是对于整小时的装箱操作，我尝试使用此SQL脚本：
select Duration, datepart(HOUR, Start) AS hourN, datepart(day, Start) AS dayN
from [dbo].[DataInput]
where Start >= '2020-03-06 00:00' and Start <= '2020-03-06 23:59'
group by
datepart(day, Start), datepart(HOUR, Start)
在我的数据行中，我还有一个名为“持续时间”的列，该列只是以总秒数计算的开始到结束。但是，它没有考虑跨越小时/天的时间。我正在使用SQL Server。结果是，我认为这是您通常在一小时的条形图中会发现的结果。
这里是使用临时提示表和少许蛮力的方法。
示例
Select [Date] = convert(date,DT)
,[Hour] = datepart(hour,DT)
,[Seconds]=sum(1)
From (
Select B.*
,DT=DateAdd(SECOND,N,[Start])
From #YourTable A
Cross Apply ( Select Top (datediff(second,[Start],[End]))
N=-1+Row_Number() Over (Order By (Select NULL))
From master..spt_values n1, master..spt_values n2
) B
) A
Group By convert(date,DT)
,datepart(hour,DT)
Order By [Date],[Hour]
返回
Date Hour Seconds
2020-03-06 15 1760
2020-03-06 16 3600
2020-03-06 17 541
2020-03-06 22 728
2020-03-06 23 3540
2020-03-07 0 2872
2020-03-09 17 3274 << Not clear if you wanted a WHERE
2020-03-09 18 3594 <<
2020-03-09 19 326 <<
可能的方法-不是最漂亮的，也许不是最好的方法-并且不确定它是否可以在您可能拥有的所有期间都适用
从https://dba.stackexchange.com/questions/174035/split-duration-hourly-depending-on-start-and-end-time改编的代码
declare @workdate datetime = '2020-03-06'
;
-- may need more rows here depending on how long the events can be
WITH E1(N) AS (
SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL
SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL
SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1 UNION ALL SELECT 1
),
numbers (n) AS ( SELECT top 24 row_number() over (order by a.n) - 1 FROM E1 a, E1 b)
, inpdata as
(
select convert(datetime, starttime) as Start_timestamp
, convert(datetime, Endtime) as End_timestamp
from (values ('2020-03-06 15:30:40.000', '2020-03-06 17:09:01.000')
,('2020-03-06 22:47:52.000', '2020-03-06 23:48:52.000')
,('2020-03-06 23:49:52.000', '2020-03-07 00:47:52.000')
,('2020-03-09 17:05:26.000', '2020-03-09 18:05:26.000')
,('2020-03-09 18:05:32.000', '2020-03-09 19:05:26.000')
) t (Starttime, Endtime)
)
, workday as
(
SELECT
convert(varchar(13), start_final, 121) as period
, convert(date, start_final) as perioddt
, sum(DATEDIFF(second, start_final, end_final)) Duration
FROM
(
SELECT Start_timestamp
, End_Timestamp
, CASE WHEN t.Start_timestamp > n.start_from_numbers THEN t.Start_timestamp ELSE n.start_from_numbers END start_final
, CASE WHEN t.End_timestamp > n.end_from_numbers THEN n.end_from_numbers ELSE t.End_timestamp END end_final
FROM inpdata t
CROSS APPLY
(
SELECT dateadd(hour, n.n + datediff(hour, 0, t.Start_timestamp), 0) start_from_numbers
, dateadd(hour, 1 + n.n + datediff(hour, 0, t.Start_timestamp), 0) end_from_numbers
FROM numbers n WHERE DATEDIFF(HOUR, t.Start_timestamp, t.End_Timestamp) >= n.n
) n
) t2
group by convert(varchar(13), start_final, 121)
,convert(date, start_final)
)
select workdate
, coalesce(work.duration, 0) as duration
from numbers
cross apply (select convert(varchar(13), dateadd(hour, numbers.n, @workdate), 121) as workdate) dt
outer apply (select *
from workday wd
where wd.period = dt.workdate
) work