属于时间段的时间

Question

I am looking for a way to calculate the time that falls into a datetime period.

I have a dataset that looks like this:

+------+------------+-------------------------+-------------------------+
| Name |    Date    |          Start          |           End           |
+------+------------+-------------------------+-------------------------+
| Me   | 06.02.2018 | 2018-02-06 22:26:00.000 | 2018-02-07 05:21:00.000 |
| Me   | 08.02.2018 | 2018-02-08 19:00:00.000 | 2018-02-08 22:04:00.000 |
+------+------------+-------------------------+-------------------------+

I now want to check how much time falls into a certain time window, eg:

• 0600 to 2000
• 2000 to 2300
• 2300 to 0600 (next day)
• 0000 to 0400 (next day)

So the result would look like this for the example above:

+------+------------+-------------------------+-------------------------+----------+----------+----------+----------+
| Name |    Date    |          Start          |           End           |  6 - 20  | 20 - 23  |  23 - 6  |  0 - 4   |
+------+------------+-------------------------+-------------------------+----------+----------+----------+----------+
| Me   | 06.02.2018 | 2018-02-06 22:26:00.000 | 2018-02-07 05:21:00.000 | 00:00:00 | 00:34:00 | 06:21:00 | 04:00:00 |
| Me   | 08.02.2018 | 2018-02-08 19:00:00.000 | 2018-02-08 22:04:00.000 | 01:00    | 02:04    | 00:00:00 | 00:00:00 |
+------+------------+-------------------------+-------------------------+----------+----------+----------+----------+

I am struggling with a large line of code:

CONVERT(varchar(5),IIF(a.end<(a.date+'00:00'),0,DATEADD(minute, DATEDIFF(MINUTE, IIF(a.start<(a.date+'00:00'),(a.date+'00:00'),a.start), IIF(a.end>(a.date+0+'04:00'),(a.date+0+'04:00'),a.end)), 0)), 114) [00-04],

This is exemplary for what I am trying to achieve but it doesn't work for this time period. The +0 is there because in some cases I do have +1 and I am generating the lines of SQL in Excel at the moment.

Can anyone help me out here?

Answer 1

What about the PIVOT? Check it on rextester.com .

WITH
  t AS (
    SELECT 'Me' "Name",
      CAST(a AS DATETIME) "Date",
      CAST(b AS DATETIME) "Start",
      CAST(c AS DATETIME) "End"
    FROM (
      VALUES ('20180206',
              '20180206 22:26:00.000',
              '20180207 05:21:00.000'),
             ('20180208',
              '20180208 19:00:00.000',
              '20180208 22:04:00.000')
    ) t(a, b, c)
  ),
  d AS (
    SELECT *
    FROM (
      VALUES (6, 20), (20, 23), (23, 30), (24, 28)
    ) t(beg, fin)
  ),
  a AS (
    SELECT t."Date", t."Start", t."End",
      LTRIM(STR(d.beg)) + ' - ' + LTRIM(STR(d.fin)) Range,
      DATEADD(hour, d.beg, t."Date") beg,
      DATEADD(hour, d.fin, t."Date") fin
    FROM t JOIN d
    ON t."Start" <= DATEADD(hour, d.fin, t."Date") AND
       t."End" >= DATEADD(hour, d.beg, t."Date")
  ),
  b AS (
    SELECT "Date", "Start", "End", Range,
      LEFT(
        CONVERT(CHAR(8),
          DATEADD(minute,
            DATEDIFF(minute,
              IIF("Start" > beg, "Start", beg),
              IIF("End" > fin, fin, "End")), 0), 8), 5) Duration
    FROM a
  )
SELECT "Date", "Start", "End",
  COALESCE("6 - 20", '00:00') "6 - 20",
  COALESCE("20 - 23", '00:00') "20 - 23",
  COALESCE("23 - 30", '00:00') "23 - 6",
  COALESCE("24 - 28", '00:00') "0 - 4"
FROM b
PIVOT (
  MIN(Duration)
    FOR Range
      IN ("6 - 20", "20 - 23", "23 - 30", "24 - 28")
) AS p;

Answer 2

Here's one way. I got all dates from your table and formed needed time periods. Then looked for overlapping periods and calculated the time. This query will not return rows which startDate and endDate on same day and before 6 o'clock

declare @t table (
    Date date
    , StartDate datetime
    , EndDate datetime
)
insert into @t 
values 
    ('20180206', '20180206 22:26:00.000', '20180207 05:21:00.000')
    , ('20180207', '20180207 05:26:00.000', '20180207 22:21:00.000')
    , ('20180207', '20180207 05:16:00.000', '20180207 05:21:00.000')

;with time_periods as (
    select 
        distinct q.Date, t.n
        , start_date = dateadd(hh, t.b, cast(q.Date as datetime))
        , end_date = dateadd(hh, t.c, cast(q.Date as datetime))
    from 
        @t q
        cross apply (values (1, 6, 20), (2, 20, 23), (3, 23, 30), (4, 24, 28)) t(n, b, c)
)

select
    Date, StartDate, EndDate
    , [6 - 20] = right(concat('0', v1 / 3600), 2) + ':' + right(concat('0', v1 % 3600 / 60), 2) + ':' + right(concat('0', v1 % 60), 2)
    , [20 - 23] = right(concat('0', v2 / 3600), 2) + ':' + right(concat('0', v2 % 3600 / 60), 2) + ':' + right(concat('0', v2 % 60), 2)
    , [23 - 6] = right(concat('0', v3 / 3600), 2) + ':' + right(concat('0', v3 % 3600 / 60), 2) + ':' + right(concat('0', v3 % 60), 2)
    , [0 - 4] = right(concat('0', v4 / 3600), 2) + ':' + right(concat('0', v4 % 3600 / 60), 2) + ':' + right(concat('0', v4 % 60), 2)
from (
    select
        Date, StartDate, EndDate
        , v1 = max(iif(n = 1, diff, 0))
        , v2 = max(iif(n = 2, diff, 0))
        , v3 = max(iif(n = 3, diff, 0))
        , v4 = max(iif(n = 4, diff, 0))
    from (
        select
            m.Date, m.StartDate, m.EndDate, c.n
            , diff = datediff(ss, iif(m.StartDate < c.start_date, c.start_date, m.StartDate), iif(m.EndDate < c.end_date, m.EndDate, c.end_date))
        from
            @t m
            join time_periods c on m.StartDate <= c.end_date and m.EndDate >= c.start_date and c.Date = m.Date
    ) t
    group by Date, StartDate, EndDate
) t

Output

Date        StartDate               EndDate                 [6 - 20]    [20 - 23]   [23 - 6]    [0 - 4]
------------------------------------------------------------------------------------------------------------
2018-02-06  2018-02-06 22:26:00.000 2018-02-07 05:21:00.000 00:00:00    00:34:00    06:21:00    04:00:00
2018-02-07  2018-02-07 05:26:00.000 2018-02-07 22:21:00.000 14:00:00    02:21:00    00:00:00    00:00:00