Sql 前十名活动持续时间

我有一个SQL Server表，如下所示：

TimeStamp (datetime)| Alarm (string) | Active (bool)
01.08.2013 1:30:05    | Alarm 1             | false
01.08.2013 1:29:54    | Alarm 2             | true
01.08.2013 1:28:43    | Alarm 1             | true
01.08.2013 1:27:21    | Alarm 3             | false
01.08.2013 1:26:35    | Alarm 1             | false
01.08.2013 1:25:34    | Alarm 1             | true

我已经显示了前10个报警事件：

SELECT TOP 10 Alarm, COUNT(Alarm) AS Occurrence
FROM MyTable
WHERE (Active = 'True')
GROUP BY Alarm
ORDER BY Occurrence DESC

我现在需要前10个警报持续时间

目标是获得每个报警的持续时间总和（从true到false）

我真的被阻止了，我想我需要迭代每一行并为每个met报警求和时间值

但我不知道如何使用SQL查询来实现这一点。任何帮助或指导都将不胜感激

提前谢谢

---

@Roman Pekar:SQL server 2008

@Gordon Linoff：预期输出=>前10个报警持续时间（当报警为真时））

---

您可以使用

交叉应用

查找

打开

后的第一个

关闭

：

select  top 10 active.Alarm
,       active.TimeStamp as TimeOn
,       active.TimeStamp as TimeOff
,       datediff(second, active.TimeStamp, inactive.TimeStamp) as Duration
from    YourTable active
cross apply
        (
        select  top 1 *
        from    YourTable inactive
        where   inactive.Active = 'false'
                and inactive.Alarm = active.Alarm
                and inactive.TimeStamp > active.TimeStamp
        order by
                inactive.TimeStamp
        ) inactive
where   active.Active = 'true'
order by
        Duration desc

---

这里是另一个可能更有效的解决方案

;with a as
(
    select a.Alarm
    , a.Active
    , a.[Timestamp]
    , Sequence = dense_rank() over
                 (
                     partition by a.Alarm, a.Active
                     order by a.[TimeStamp]
                 )
    from #YourTable a
),
b as
(
    select a.Alarm
    , Start = b.[Timestamp]
    , Stop = a.[Timestamp]
    from a 
    left outer join a b
    on b.Alarm = a.Alarm
    and b.Sequence = a.Sequence
    and b.Active = 'true'
    where a.Active = 'false'
)
select b.Alarm
, DurationMinutes = sum(datediff(millisecond, b.Start, b.Stop) / 1000.0 / 60.0)
from b
group by b.Alarm

---

您有什么版本的SQL Server？请提供示例输出。不清楚您希望问题中数据的结果是什么。由于报警2和报警3只出现一次（报警2=1真，报警3=1假），您如何计算报警2和报警3的时间？另外，对于报警1，您需要两行还是只需要一行。我的意思是，你想计算总持续时间或两者的持续时间，1:25:34到1:26:35，然后1:28:43到1:30:05？非常感谢你的回答，这对我帮助很大！可能更高效，但也要求更高的数据一致性。如果第一行是前一个报警的结尾（“关闭”条目），则查询将返回负持续时间。@Andomar True。您可以在“on”子句中添加“and b.Timestamp;with a as ( select a.Alarm , a.Active , a.[Timestamp] , Sequence = dense_rank() over ( partition by a.Alarm, a.Active order by a.[TimeStamp] ) from #YourTable a ), b as ( select a.Alarm , Start = b.[Timestamp] , Stop = a.[Timestamp] from a left outer join a b on b.Alarm = a.Alarm and b.Sequence = a.Sequence and b.Active = 'true' where a.Active = 'false' ) select b.Alarm , DurationMinutes = sum(datediff(millisecond, b.Start, b.Stop) / 1000.0 / 60.0) from b group by b.Alarm