使用 SQL 服务器中的存储过程提取值

Extract values with stored procedure in SQL Server

我需要你的帮助来在 SQL 服务器 (v12.0.6024.0) 中创建一个视图。我的一位客户有一个 table,其中一些时间段以这种格式保存:

ID ID_EVENT Time Slot
1000 24 08:30:00.0000
1000 24 09:00:00.0000
1000 24 09:30:00.0000

每个时间段持续 30 分钟,上面的示例表示 ID 为 24 的事件(保存在另一个 table 中)持续形式为 8:30 到 10:00(第 3 个时间段开始于 9:30,持续了 30 分钟,所以在 10:00 结束。问题是在某些情况下时间值不连续,中间可能会有停顿,所以我会这样:

ID ID_EVENT Time Slot
1000 24 08:30:00.0000
1000 24 09:00:00.0000
1000 24 09:30:00.0000
1000 24 11:30:00.0000
1000 24 12:00:00.0000
1000 24 12:30:00.0000

在这种情况下,ID 为 24 的事件从 8:30 持续到 10,停止,然后从 11:30 到 13:00 再次开始。我被要求为外部开发人员准备一个视图,其中我不仅要报告事件开始的时间(在我的示例中,8:30),而且它永远停止的时间(在我的示例中,13:00) 以及暂停开始的时间(在我的示例中 10:00)和暂停结束的时间(在我的示例中 11:30)。

我对前 2 个值没有问题,但我不知道如何提取其他两个值。我认为我们可以考虑在 2 个时间段不连续时发生暂停,同一事件不能超过多个时间段。我想我需要一个过程,但发现很难写;我需要一个显示

的视图
ID ID_EVENT Time1 Time2 Time3 Time4
1000 24 08:30:00.0000 10:00:00.0000 11:30:00.0000 13:00:00.0000

有什么帮助吗?

declare @t table(ID int, ID_EVENT int, TimeSlot time)
insert into @t
values
(1000,  24, '08:30:00.0000'),
(1000,  24, '09:00:00.0000'),
(1000,  24, '09:30:00.0000'),
--
(1000,  24, '11:30:00.0000'),
(1000,  24, '12:00:00.0000'),
(1000,  24, '12:30:00.0000'),
--
(1000,  24, '15:00:00.0000'),
(1000,  24, '15:30:00.0000'),
(1000,  24, '16:00:00.0000'),
--
(1000,  25, '15:30:00.0000'),
(1000,  25, '16:30:00.0000');


select Id, ID_EVENT, 
    min(TimeSlot) as StartTimeSlot, 
    dateadd(minute, 30, max(TimeSlot)) as EndTimeSlot
from 
(
    select *,
    datediff(minute, '00:00:00', Timeslot)/30 - row_number() over(partition by Id, ID_EVENT order by TimeSlot) as grpid
    from @t
) as t
group by Id, ID_EVENT, grpid;



--first two groups per event&id row
select Id, ID_EVENT,
--1
min(case when grpordinal = 1 then TimeSlot end) as StartSlot1,
dateadd(minute, 30, max(case when grpordinal = 1 then TimeSlot end)) as EndSlot1,
--2
min(case when grpordinal = 2 then TimeSlot end) as StartSlot2,
dateadd(minute, 30, max(case when grpordinal = 2 then TimeSlot end)) as EndSlot2
from 
(
    select Id, ID_EVENT, TimeSlot,
        dense_rank() over(partition by Id, ID_EVENT order by grpid) as grpordinal
    from 
    (
        select *,
        datediff(minute, '00:00:00', Timeslot)/30 - row_number() over(partition by Id, ID_EVENT order by TimeSlot) as grpid
        from @t
    ) as t
) as src
--where grpordinal <= 2 --not really needed
group by Id, ID_EVENT;



--!!!!only when there are max two groups/periods
--if there could be more than 2 periods this will not work
select Id, ID_EVENT,
--1
min(case when grpid = 0 then TimeSlot end) as StartSlot1,
dateadd(minute, 30, max(case when grpid = 0 then TimeSlot end)) as EndSlot1,
--2
min(case when grpid <> 0 then TimeSlot end) as StartSlot2,
dateadd(minute, 30, max(case when grpid <> 0 then TimeSlot end)) as EndSlot2
from
(
select *,
/*
    1
    + datediff(minute, '00:00:00', Timeslot)/30 - row_number() over(partition by Id, ID_EVENT order by TimeSlot) 
    - datediff(minute, '00:00:00', min(Timeslot) over(partition by Id, ID_EVENT)) /30
*/
    1
    + datediff(minute, min(Timeslot) over(partition by Id, ID_EVENT), TimeSlot)/30  
    - row_number() over(partition by Id, ID_EVENT order by TimeSlot)
    as grpid --1st groupid is always 0
from @t
) as t
group by Id, ID_EVENT;

这看起来像是一个间隙和孤岛问题,您要在其中识别“相邻”时隙并将其组合在一起。

我建议将范围放在行中而不是列中。为此,您可以使用 window 这样的函数:

select id, id_event, 
    min(timeslot) as timeslot_start, max(timeslot) as timeslot_end
from (
    select t.*, 
        row_number() over(partition by id, id_event order by timeslot) rn
    from mytable t
) t
group by id, id_event, datediff(minute, - rn * 30, timeslot)

如果只想查看每个事件的前两个范围 - 都在结果集中的同一行 - 那么我们可以在该查询之上使用条件聚合:

select id, id_event,
    max(case when rn = 1 then timeslot_start end) as timeslot_start_1,
    max(case when rn = 1 then timeslot_end   end) as timeslot_end_1,
    max(case when rn = 2 then timeslot_start end) as timeslot_start_2,
    max(case when rn = 2 then timeslot_end   end) as timeslot_end_2
from (
    select id, id_event, 
        min(timeslot) as timeslot_start, max(timeslot) as timeslot_end,
        row_number() over(partition by id, id_event order by min(timeslot)) rn
    from (
        select t.*, 
            row_number() over(partition by id, id_event order by timeslot) rn
        from mytable t
    ) t
    group by id, id_event, datediff(minute, - rn * 30, timeslot)
) t
where rn <= 2
group by id, id_event