sql - 从累积数据中获取耗气量

Question

这是我在这里的第一篇文章!我处理这个查询已经有一段时间了，我真的需要你的帮助。

我有两张 table

Counters
=============   
ID  Description
-------------------
1   GasCounter1
2   GasCounter2

Measurements
================
ID  TimeStamp              ValCum   CounterID
----------------------------------------------
1   01.01.2015 00:00:00    1        1
2   01.01.2015 00:15:00    2        1
3   01.01.2015 00:30:00    3        1
4   01.01.2015 00:45:00    4        1
---------------------------------------
5   01.01.2015 01:00:00    5        1
6   01.01.2015 01:15:00    6        1
7   01.01.2015 01:30:00    7        1
8   01.01.2015 01:45:00    8        1
---------------------------------------
9   01.01.2015 02:00:00    9        1
10  01.01.2015 02:15:00    10       1
11  01.01.2015 02:30:00    11       1
12  01.01.2015 02:45:00    12       1
---------------------------------------
13  01.01.2015 03:00:00    13       1
14  01.01.2015 03:15:00    14       1
15  01.01.2015 03:30:00    15       1
16  01.01.2015 03:45:00    16       1
---------------------------------------
17  01.01.2015 04:00:00    17       1
18  01.01.2015 04:15:00    18       1
19  01.01.2015 04:30:00    19       1
20  01.01.2015 04:45:00    20       1

.... Here are data only for GasCounter1

ValueCum 是从 GasCounters 读取的累积消耗量

我需要计算每小时的气体消耗量(从选定的时间间隔开始)在此示例中，所有消耗量均为 4。

1->2 = 1 +
2->3 = 1 +
3->4 = 1 +
4->5 = 1 = ===> 4

5->6 = 1
6->7 = 1
7->8 = 1
8->9 = 1 ===> 4

如果实际数据是:

5->6 = 1
6->7 = 1
7->9 = 2
9->12 = 3 ===> 7

好吧，我尝试使用 Group by --> 我可以计算组总和，但那是错误的。当我四处搜索时，我找到了一些running total 和cumulative sum 的解决方案，但这恰恰与我需要的相反。

我已经有了累积值 -> 我需要两者之间的差异:

last record in **previous** group and 
last record inside current group

我希望我清楚这个问题。感谢您的帮助

Answer 1

如果我答对了你的问题，也许这就是你要找的东西。

-- Generate demo data
CREATE TABLE #measure(id int identity(1,1), [TimeStamp] datetime, valcum int, counterid int)

INSERT INTO #measure(TimeStamp, valcum, counterid)
VALUES  (DATEADD(hour,-3,GETDATE()),1,1), (DATEADD(hour,-3,GETDATE()),2,1), (DATEADD(hour,-3,GETDATE()),1,1), (DATEADD(hour,-3,GETDATE()),2,1),
        (DATEADD(hour,-2,GETDATE()),8,1), (DATEADD(hour,-2,GETDATE()),1,1), (DATEADD(hour,-2,GETDATE()),8,1), (DATEADD(hour,-2,GETDATE()),1,1), 
        (DATEADD(hour,-1,GETDATE()),3,1), (DATEADD(hour,-1,GETDATE()),1,1), (DATEADD(hour,-1,GETDATE()),3,1), (DATEADD(hour,-1,GETDATE()),1,1),
        (GETDATE(),2,1), (GETDATE(),1,1), (GETDATE(),10,1), (GETDATE(),2,1), (GETDATE(),1,1), (GETDATE(),10,1)

-- Show demo data
SELECT *
FROM #measure

-- Sum every value grouped per date and hour.
;WITH data AS(
    SELECT id, TimeStamp, valcum, counterid, 
        SUM(valcum) OVER (PARTITION BY  CONVERT(date, m.timestamp), DATEPART(hour, m.TimeStamp)) as sumVal,
        DENSE_RANK() OVER (ORDER BY CONVERT(date, m.timestamp), DATEPART(hour, m.TimeStamp)) as package,
        ROW_NUMBER() OVER (PARTITION BY  CONVERT(date, m.timestamp), DATEPART(hour, m.TimeStamp) ORDER BY id) as numberInPackage
    FROM #measure as m
)
SELECT *, d1.valcum - d2.valcum as diff
FROM data as d1
LEFT JOIN data as d2
        ON d1.package-1 = d2.package
WHERE d1.numberInPackage = 4 and d2.numberInPackage = 4


-- Cleanup
DROP TABLE #measure