如果值在连续时间段内出现，SQL 需要应用可变速率

Question

我有一张如下所示的表格：

在查询中，我需要找到两个时间段（行）内出现的最大导入值，其中该值大于定义的阈值并应用速率。如果它发生在两个以上的时间段内，将使用不同的速率

Threshold = 1000 
Rate 1 (2 consecutive) = 100
Rate 2 (> 2 consecutive) = 200

Id    DateTime            Import       Export Total
1     2016-01-13 00:00    1000         500    1500
2     2016-01-13 00:15    2500         100    3000
3     2016-01-13 00:30    1900         200    2100
4     2016-01-13 01:00    900          100    1200

ID 2 和 3 是 > 阈值，因此查询应返回那些 (2500,1900) = 1900 减去阈值 (1000) = 900 的最小值。应用速率 Rate1 * 900 = 9000

如果我们将 Id 4 的值更改为 1200，那么 MIN 值将是 1200。减去阈值 = 200。200 * Rate2 = 4000

任何帮助将不胜感激！

反馈后更新。我的挑战似乎是我没有获得第二高的价值。以下是数据集的示例： Dataset example

我添加了另一个变量来缩小列表以测试间隙和孤岛部分。这是一个较小的子集： Subset

代码如下：

    WITH CTE AS (  

SELECT LogTable.[LocalTimestamp] as thetime,LogTable.[SystemImport] as import, LogTable.[Id] - ROW_NUMBER() OVER (ORDER BY LogTable.[Id]) AS grp
  FROM {System_KWLogRaw} LogTable
  WHERE LogTable.[SystemImport] between @DemandThreshold and @In1 and 
    DATEPART(year,@inDate) = DATEPART(year, LogTable.[LocalTimestamp]) and
    DATEPART(month,@inDate) = DATEPART(month, LogTable.[LocalTimestamp]) and
    DATEPART(day,@inDate) = DATEPART(day, LogTable.[LocalTimestamp])
),

counted AS (
    SELECT *, COUNT(*) OVER (PARTITION BY grp) AS cnt
    FROM CTE

    )

SELECT MAX(counted.import) as again1
FROM counted
WHERE cnt > 3 and counted.import < (SELECT MAX(counted.import) FROM counted)

这将返回 3555.53 而不是 3543.2，后者是第二高的值

Answer 1

这将满足您的要求：

with x as (
select
t1.Id,
t1.DateTime,
t1.Import,
t1.Export,
t1.Total,
count(t2.Import) over (partition by 1) as [QualifyingImports],
min(t2.Import) over (partition by 1) as [MinQualifyingImport]

from
myTable t1
left join myTable t2 on t2.Import > 1000 and t2.Id = t1.Id

where
t1.DateTime >= '2016-01-13'
and t1.DateTime < dateadd(d, 1,'2016-01-13')

)

select
x.Id,
x.DateTime,
x.Import,
x.Export,
x.Total,
case when x.[QualifyingImports] > 2 then (x.MinQualifyingImport - 1000) * 200 else (x.MinQualifyingImport - 1000) * 100 end as [Rate]
from x

我已将Fiddle 放在一起，因此您可以使用 Id #4 的不同值。

我真的很想把阈值和周期之类的值变成@variables，但它似乎在 CTE 中不受支持，所以我只好硬编码它们。

编辑

原来 CTE 是多余的，你可以把它缩小到这个并使用 @variables，耶！

declare @period smalldatetime = '2016-01-13'
declare @threshold float = 1000
declare @rate1 float = 100
declare @rate2 float = 200

select
t1.Id,
t1.DateTime,
t1.Import,
t1.Export,
t1.Total,
case
  when count(t2.Import) over (partition by 1) > 2 then (min(t2.Import) over (partition by 1) - @threshold) * @rate2
  else (min(t2.Import) over (partition by 1) - @threshold) * @rate1
  end as [Rate]

from
myTable t1
left join myTable t2 on t2.Import > @threshold and t2.Id = t1.Id

where
t1.DateTime >= @period
and t1.DateTime < dateadd(d, 1, @period)

新Fiddle