SQL 计算从指定日期的第一天到指定日期的差异
SQL to calculate difference from 1st day of Month to Date specified
我正在处理 sql 查询以获得差异。我有一个 table 包含一个读数，读数的时间戳，id。我的最终目标是获得三个差异。 1. 前一天读数与第二天读数之间的差异，2. 时间戳值前 7 天读数之间的差异，3. 从当月的第一天开始到每个指定日期的读数之间的差异。
我破解了第一个两项。现在我正在尝试破解第三个。我知道它会很容易使用功能，任何人都可以帮助我处理第三个请求。
预期结果：11 月 1 日的读数为 1000，11 月 2 日的读数为 1020，11 月 3 日的读数为 1050，11 月 2 日的差值应为 20，11 月 3 日的差值应为 50。
如果一个月的第一天没有数据，则取可用日期的最少数据。例如，september 只有从 24 开始，所以从 9 月 24 日开始阅读。
下面是例子table。
+----+-----------+---------+----------------+----------------+-----------------+
| id | timestamp | Reading | 1DayDifference | 7DayDifference | monthDifference |
+----+-----------+---------+----------------+----------------+-----------------+
| A1 | 11/20/18 | 44182 | 0 | 300 | 541 |
| A1 | 11/19/18 | 44182 | 0 | 338 | 541 |
| A1 | 11/18/18 | 44182 | 0 | 338 | 541 |
| A1 | 11/17/18 | 44182 | 38 | 338 | 541 |
| A1 | 11/16/18 | 44144 | 197 | 300 | 503 |
| A1 | 11/15/18 | 43947 | 26 | 103 | |
| A1 | 11/14/18 | 43921 | 39 | 158 | |
| A1 | 11/13/18 | 43882 | 38 | 158 | |
| A1 | 11/12/18 | 43844 | 0 | 120 | |
| A1 | 11/11/18 | 43844 | 0 | 120 | |
| A1 | 11/10/18 | 43844 | 0 | 160 | |
| A1 | 11/09/18 | 43844 | 0 | 203 | |
| A1 | 11/08/18 | 43844 | 81 | 241 | |
| A1 | 11/06/18 | 43763 | 39 | 198 | |
| A1 | 11/05/18 | 43724 | 0 | 198 | |
| A1 | 11/04/18 | 43724 | 0 | 198 | |
| A1 | 11/03/18 | 43724 | 40 | 198 | |
| A1 | 11/02/18 | 43684 | 43 | 199 | |
| A1 | 11/01/18 | 43641 | 38 | 194 | |
| A1 | 10/31/18 | 43603 | 38 | 275 | 237 |
| A1 | 10/30/18 | 43565 | 39 | 317 | |
| A1 | 10/29/18 | 43526 | 0 | 317 | |
| A1 | 10/28/18 | 43526 | 0 | 317 | |
| A1 | 10/27/18 | 43526 | 41 | 317 | |
| A1 | 10/26/18 | 43485 | 38 | 276 | |
| A1 | 10/25/18 | 43447 | 119 | 238 | |
| A1 | 10/24/18 | 43328 | 80 | 119 | |
+----+-----------+---------+----------------+----------------+-----------------+
我习惯了第一种的SQL两种。
SELECT id,
timestamp,
Reading,
Reading - lead(Reading,1,0) OVER( partition BY [id] ORDER BY timestamp desc) [OneDayDifference],
Reading - lead(Reading,7,0) OVER( partition BY [id] ORDER BY timestamp desc) [SevDayDifference]
FROM [dbo].[test_example] s
ORDER BY id, timestamp desc
下面是生成上述数据的脚本。
CREATE TABLE [dbo].[test_Example](
[id] [nvarchar](50) NOT NULL,
[timestamp] [datetime2](7) NOT NULL,
[reading] [int] NOT NULL,
[OneDayDifference] [int] NOT NULL,
[SevDayDifference] [int] NOT NULL
) ON [PRIMARY]
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-19T00:01:38.0000000' AS DateTime2), 44182, 0, 338)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-18T00:01:44.0000000' AS DateTime2), 44182, 0, 338)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-17T00:01:35.0000000' AS DateTime2), 44182, 38, 338)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-16T00:01:39.0000000' AS DateTime2), 44144, 197, 300)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-15T00:01:47.0000000' AS DateTime2), 43947, 26, 103)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-14T00:01:40.0000000' AS DateTime2), 43921, 39, 158)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-13T00:01:38.0000000' AS DateTime2), 43882, 38, 158)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-12T00:02:39.0000000' AS DateTime2), 43844, 0, 120)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-11T00:01:37.0000000' AS DateTime2), 43844, 0, 120)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-10T00:01:37.0000000' AS DateTime2), 43844, 0, 160)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-09T00:01:37.0000000' AS DateTime2), 43844, 0, 203)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-08T00:01:46.0000000' AS DateTime2), 43844, 81, 241)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-06T00:01:36.0000000' AS DateTime2), 43763, 39, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-05T00:02:27.0000000' AS DateTime2), 43724, 0, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-04T00:01:37.0000000' AS DateTime2), 43724, 0, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-03T00:01:48.0000000' AS DateTime2), 43724, 40, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-02T00:01:33.0000000' AS DateTime2), 43684, 43, 199)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-01T00:01:41.0000000' AS DateTime2), 43641, 38, 194)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-31T00:01:32.0000000' AS DateTime2), 43603, 38, 275)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-30T00:01:34.0000000' AS DateTime2), 43565, 39, 43565)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-29T00:02:45.0000000' AS DateTime2), 43526, 0, 43526)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-28T00:01:43.0000000' AS DateTime2), 43526, 0, 43526)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-27T00:01:31.0000000' AS DateTime2), 43526, 41, 43526)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-26T00:01:30.0000000' AS DateTime2), 43485, 38, 43485)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-25T00:01:35.0000000' AS DateTime2), 43447, 119, 43447)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-24T00:01:43.0000000' AS DateTime2), 43328, 43328, 43328)
GO
不使用 Lead()，而是使用子查询获取相同 Id、year 和 month 的前 1 行，按 [=14= 排序] 并计算您与子查询返回的该行的 reading 的差异。
查找一个月的第一天需要向后查看可变数量的行，因此您可以在 apply 中使用相关子查询而不是 LEAD() 或 LAG()。请注意，因为您是“向后看”，所以我更喜欢使用 LAG() 而不是颠倒时间戳和 LEAD() 的顺序，但是两者都会产生相同的结果。
nb：此子查询将查找任何月份中最早的时间戳，如果不需要，则将 and t.timestamp < dateadd(dd,1,dateadd(mm,datediff(mm,0,s.timestamp),0)) 添加到 where 子句中
SELECT
id
, timestamp
, Reading
, Reading - LAG( Reading, 1, 0 ) OVER (PARTITION BY [id] ORDER BY timestamp) [OneDayDifference]
, Reading - LAG( Reading, 7, 0 ) OVER (PARTITION BY [id] ORDER BY timestamp) [SevDayDifference]
, reading - oa.prev_reading [ThisMonthDiff]
FROM [dbo].[test_example] s
outer apply (
select top(1) t.reading prev_reading
from [dbo].[test_example] t
where s.id = t.id
and t.timestamp >= dateadd(mm,datediff(mm,0,s.timestamp),0)
-- and t.timestamp < dateadd(dd,1,dateadd(mm,datediff(mm,0,s.timestamp),0))
order by t.timestamp
) oa
ORDER BY
id
, timestamp DESC
;
结果：
+----+----+------------+---------+------------------+------------------+---------------+
| | id | timestamp | Reading | OneDayDifference | SevDayDifference | ThisMonthDiff |
+----+----+------------+---------+------------------+------------------+---------------+
| 1 | A1 | 2018-11-19 | 44182 | 0 | 338 | 541 |
| 2 | A1 | 2018-11-18 | 44182 | 0 | 338 | 541 |
| 3 | A1 | 2018-11-17 | 44182 | 38 | 338 | 541 |
| 4 | A1 | 2018-11-16 | 44144 | 197 | 300 | 503 |
| 5 | A1 | 2018-11-15 | 43947 | 26 | 103 | 306 |
| 6 | A1 | 2018-11-14 | 43921 | 39 | 158 | 280 |
| 7 | A1 | 2018-11-13 | 43882 | 38 | 158 | 241 |
| 8 | A1 | 2018-11-12 | 43844 | 0 | 120 | 203 |
| 9 | A1 | 2018-11-11 | 43844 | 0 | 120 | 203 |
| 10 | A1 | 2018-11-10 | 43844 | 0 | 160 | 203 |
| 11 | A1 | 2018-11-09 | 43844 | 0 | 203 | 203 |
| 12 | A1 | 2018-11-08 | 43844 | 81 | 241 | 203 |
| 13 | A1 | 2018-11-06 | 43763 | 39 | 198 | 122 |
| 14 | A1 | 2018-11-05 | 43724 | 0 | 198 | 83 |
| 15 | A1 | 2018-11-04 | 43724 | 0 | 198 | 83 |
| 16 | A1 | 2018-11-03 | 43724 | 40 | 198 | 83 |
| 17 | A1 | 2018-11-02 | 43684 | 43 | 199 | 43 |
| 18 | A1 | 2018-11-01 | 43641 | 38 | 194 | 0 |
| 19 | A1 | 2018-10-31 | 43603 | 38 | 275 | 275 |
| 20 | A1 | 2018-10-30 | 43565 | 39 | 43565 | 237 |
| 21 | A1 | 2018-10-29 | 43526 | 0 | 43526 | 198 |
| 22 | A1 | 2018-10-28 | 43526 | 0 | 43526 | 198 |
| 23 | A1 | 2018-10-27 | 43526 | 41 | 43526 | 198 |
| 24 | A1 | 2018-10-26 | 43485 | 38 | 43485 | 157 |
| 25 | A1 | 2018-10-25 | 43447 | 119 | 43447 | 119 |
| 26 | A1 | 2018-10-24 | 43328 | 43328 | 43328 | 0 |
+----+----+------------+---------+------------------+------------------+---------------+
上面我使用了 outer apply，它的作用类似于外连接（如果未找到匹配结果，仍会返回源行）。如果这不是不必要的，那么请改用 cross apply。
编辑
SELECT
id
, format(timestamp, 'yyyy-MM-dd') [timestamp]
, Reading
, COALESCE(Reading - LAG( Reading, 1) OVER (PARTITION BY [id] ORDER BY timestamp),0) [OneDayDifference]
, COALESCE(Reading - LAG( Reading, 7) OVER (PARTITION BY [id] ORDER BY timestamp),0) [SevDayDifference]
, reading - ca.tr [ThisMonthDiff]
FROM [dbo].[test_example] s
cross apply (
select top(1) t.reading tr
from [dbo].[test_example] t
where s.id = t.id
and t.timestamp >= dateadd(mm,datediff(mm,0,s.timestamp),0)
order by t.timestamp
) ca
ORDER BY
id
, timestamp DESC
;
+----+----+------------+---------+------------------+------------------+---------------+
| | id | timestamp | Reading | OneDayDifference | SevDayDifference | ThisMonthDiff |
+----+----+------------+---------+------------------+------------------+---------------+
| 1 | A1 | 2018-11-19 | 44182 | 0 | 338 | 541 |
| 2 | A1 | 2018-11-18 | 44182 | 0 | 338 | 541 |
| 3 | A1 | 2018-11-17 | 44182 | 38 | 338 | 541 |
| 18 | A1 | 2018-11-01 | 43641 | 38 | 194 | 0 |
| 19 | A1 | 2018-10-31 | 43603 | 38 | 275 | 275 |
| 20 | A1 | 2018-10-30 | 43565 | 39 | 0 | 237 |
| 21 | A1 | 2018-10-29 | 43526 | 0 | 0 | 198 |
| 22 | A1 | 2018-10-28 | 43526 | 0 | 0 | 198 |
| 23 | A1 | 2018-10-27 | 43526 | 41 | 0 | 198 |
| 24 | A1 | 2018-10-26 | 43485 | 38 | 0 | 157 |
| 25 | A1 | 2018-10-25 | 43447 | 119 | 0 | 119 |
| 26 | A1 | 2018-10-24 | 43328 | 0 | 0 | 0 |
+----+----+------------+---------+------------------+------------------+---------------+
我正在处理 sql 查询以获得差异。我有一个 table 包含一个读数，读数的时间戳，id。我的最终目标是获得三个差异。 1. 前一天读数与第二天读数之间的差异，2. 时间戳值前 7 天读数之间的差异，3. 从当月的第一天开始到每个指定日期的读数之间的差异。
我破解了第一个两项。现在我正在尝试破解第三个。我知道它会很容易使用功能，任何人都可以帮助我处理第三个请求。
预期结果：11 月 1 日的读数为 1000，11 月 2 日的读数为 1020，11 月 3 日的读数为 1050，11 月 2 日的差值应为 20，11 月 3 日的差值应为 50。
如果一个月的第一天没有数据，则取可用日期的最少数据。例如，september 只有从 24 开始，所以从 9 月 24 日开始阅读。
下面是例子table。
+----+-----------+---------+----------------+----------------+-----------------+
| id | timestamp | Reading | 1DayDifference | 7DayDifference | monthDifference |
+----+-----------+---------+----------------+----------------+-----------------+
| A1 | 11/20/18 | 44182 | 0 | 300 | 541 |
| A1 | 11/19/18 | 44182 | 0 | 338 | 541 |
| A1 | 11/18/18 | 44182 | 0 | 338 | 541 |
| A1 | 11/17/18 | 44182 | 38 | 338 | 541 |
| A1 | 11/16/18 | 44144 | 197 | 300 | 503 |
| A1 | 11/15/18 | 43947 | 26 | 103 | |
| A1 | 11/14/18 | 43921 | 39 | 158 | |
| A1 | 11/13/18 | 43882 | 38 | 158 | |
| A1 | 11/12/18 | 43844 | 0 | 120 | |
| A1 | 11/11/18 | 43844 | 0 | 120 | |
| A1 | 11/10/18 | 43844 | 0 | 160 | |
| A1 | 11/09/18 | 43844 | 0 | 203 | |
| A1 | 11/08/18 | 43844 | 81 | 241 | |
| A1 | 11/06/18 | 43763 | 39 | 198 | |
| A1 | 11/05/18 | 43724 | 0 | 198 | |
| A1 | 11/04/18 | 43724 | 0 | 198 | |
| A1 | 11/03/18 | 43724 | 40 | 198 | |
| A1 | 11/02/18 | 43684 | 43 | 199 | |
| A1 | 11/01/18 | 43641 | 38 | 194 | |
| A1 | 10/31/18 | 43603 | 38 | 275 | 237 |
| A1 | 10/30/18 | 43565 | 39 | 317 | |
| A1 | 10/29/18 | 43526 | 0 | 317 | |
| A1 | 10/28/18 | 43526 | 0 | 317 | |
| A1 | 10/27/18 | 43526 | 41 | 317 | |
| A1 | 10/26/18 | 43485 | 38 | 276 | |
| A1 | 10/25/18 | 43447 | 119 | 238 | |
| A1 | 10/24/18 | 43328 | 80 | 119 | |
+----+-----------+---------+----------------+----------------+-----------------+
我习惯了第一种的SQL两种。
SELECT id,
timestamp,
Reading,
Reading - lead(Reading,1,0) OVER( partition BY [id] ORDER BY timestamp desc) [OneDayDifference],
Reading - lead(Reading,7,0) OVER( partition BY [id] ORDER BY timestamp desc) [SevDayDifference]
FROM [dbo].[test_example] s
ORDER BY id, timestamp desc
下面是生成上述数据的脚本。
CREATE TABLE [dbo].[test_Example](
[id] [nvarchar](50) NOT NULL,
[timestamp] [datetime2](7) NOT NULL,
[reading] [int] NOT NULL,
[OneDayDifference] [int] NOT NULL,
[SevDayDifference] [int] NOT NULL
) ON [PRIMARY]
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-19T00:01:38.0000000' AS DateTime2), 44182, 0, 338)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-18T00:01:44.0000000' AS DateTime2), 44182, 0, 338)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-17T00:01:35.0000000' AS DateTime2), 44182, 38, 338)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-16T00:01:39.0000000' AS DateTime2), 44144, 197, 300)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-15T00:01:47.0000000' AS DateTime2), 43947, 26, 103)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-14T00:01:40.0000000' AS DateTime2), 43921, 39, 158)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-13T00:01:38.0000000' AS DateTime2), 43882, 38, 158)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-12T00:02:39.0000000' AS DateTime2), 43844, 0, 120)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-11T00:01:37.0000000' AS DateTime2), 43844, 0, 120)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-10T00:01:37.0000000' AS DateTime2), 43844, 0, 160)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-09T00:01:37.0000000' AS DateTime2), 43844, 0, 203)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-08T00:01:46.0000000' AS DateTime2), 43844, 81, 241)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-06T00:01:36.0000000' AS DateTime2), 43763, 39, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-05T00:02:27.0000000' AS DateTime2), 43724, 0, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-04T00:01:37.0000000' AS DateTime2), 43724, 0, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-03T00:01:48.0000000' AS DateTime2), 43724, 40, 198)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-02T00:01:33.0000000' AS DateTime2), 43684, 43, 199)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-11-01T00:01:41.0000000' AS DateTime2), 43641, 38, 194)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-31T00:01:32.0000000' AS DateTime2), 43603, 38, 275)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-30T00:01:34.0000000' AS DateTime2), 43565, 39, 43565)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-29T00:02:45.0000000' AS DateTime2), 43526, 0, 43526)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-28T00:01:43.0000000' AS DateTime2), 43526, 0, 43526)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-27T00:01:31.0000000' AS DateTime2), 43526, 41, 43526)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-26T00:01:30.0000000' AS DateTime2), 43485, 38, 43485)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-25T00:01:35.0000000' AS DateTime2), 43447, 119, 43447)
GO
INSERT [dbo].[test_Example] ([id], [timestamp], [reading], [OneDayDifference], [SevDayDifference]) VALUES (N'A1', CAST(N'2018-10-24T00:01:43.0000000' AS DateTime2), 43328, 43328, 43328)
GO
不使用 Lead()，而是使用子查询获取相同 Id、year 和 month 的前 1 行，按 [=14= 排序] 并计算您与子查询返回的该行的 reading 的差异。
查找一个月的第一天需要向后查看可变数量的行，因此您可以在 apply 中使用相关子查询而不是 LEAD() 或 LAG()。请注意，因为您是“向后看”，所以我更喜欢使用 LAG() 而不是颠倒时间戳和 LEAD() 的顺序，但是两者都会产生相同的结果。
nb：此子查询将查找任何月份中最早的时间戳，如果不需要，则将 and t.timestamp < dateadd(dd,1,dateadd(mm,datediff(mm,0,s.timestamp),0)) 添加到 where 子句中
SELECT
id
, timestamp
, Reading
, Reading - LAG( Reading, 1, 0 ) OVER (PARTITION BY [id] ORDER BY timestamp) [OneDayDifference]
, Reading - LAG( Reading, 7, 0 ) OVER (PARTITION BY [id] ORDER BY timestamp) [SevDayDifference]
, reading - oa.prev_reading [ThisMonthDiff]
FROM [dbo].[test_example] s
outer apply (
select top(1) t.reading prev_reading
from [dbo].[test_example] t
where s.id = t.id
and t.timestamp >= dateadd(mm,datediff(mm,0,s.timestamp),0)
-- and t.timestamp < dateadd(dd,1,dateadd(mm,datediff(mm,0,s.timestamp),0))
order by t.timestamp
) oa
ORDER BY
id
, timestamp DESC
;
结果：
+----+----+------------+---------+------------------+------------------+---------------+
| | id | timestamp | Reading | OneDayDifference | SevDayDifference | ThisMonthDiff |
+----+----+------------+---------+------------------+------------------+---------------+
| 1 | A1 | 2018-11-19 | 44182 | 0 | 338 | 541 |
| 2 | A1 | 2018-11-18 | 44182 | 0 | 338 | 541 |
| 3 | A1 | 2018-11-17 | 44182 | 38 | 338 | 541 |
| 4 | A1 | 2018-11-16 | 44144 | 197 | 300 | 503 |
| 5 | A1 | 2018-11-15 | 43947 | 26 | 103 | 306 |
| 6 | A1 | 2018-11-14 | 43921 | 39 | 158 | 280 |
| 7 | A1 | 2018-11-13 | 43882 | 38 | 158 | 241 |
| 8 | A1 | 2018-11-12 | 43844 | 0 | 120 | 203 |
| 9 | A1 | 2018-11-11 | 43844 | 0 | 120 | 203 |
| 10 | A1 | 2018-11-10 | 43844 | 0 | 160 | 203 |
| 11 | A1 | 2018-11-09 | 43844 | 0 | 203 | 203 |
| 12 | A1 | 2018-11-08 | 43844 | 81 | 241 | 203 |
| 13 | A1 | 2018-11-06 | 43763 | 39 | 198 | 122 |
| 14 | A1 | 2018-11-05 | 43724 | 0 | 198 | 83 |
| 15 | A1 | 2018-11-04 | 43724 | 0 | 198 | 83 |
| 16 | A1 | 2018-11-03 | 43724 | 40 | 198 | 83 |
| 17 | A1 | 2018-11-02 | 43684 | 43 | 199 | 43 |
| 18 | A1 | 2018-11-01 | 43641 | 38 | 194 | 0 |
| 19 | A1 | 2018-10-31 | 43603 | 38 | 275 | 275 |
| 20 | A1 | 2018-10-30 | 43565 | 39 | 43565 | 237 |
| 21 | A1 | 2018-10-29 | 43526 | 0 | 43526 | 198 |
| 22 | A1 | 2018-10-28 | 43526 | 0 | 43526 | 198 |
| 23 | A1 | 2018-10-27 | 43526 | 41 | 43526 | 198 |
| 24 | A1 | 2018-10-26 | 43485 | 38 | 43485 | 157 |
| 25 | A1 | 2018-10-25 | 43447 | 119 | 43447 | 119 |
| 26 | A1 | 2018-10-24 | 43328 | 43328 | 43328 | 0 |
+----+----+------------+---------+------------------+------------------+---------------+
上面我使用了 outer apply，它的作用类似于外连接（如果未找到匹配结果，仍会返回源行）。如果这不是不必要的，那么请改用 cross apply。
编辑
SELECT
id
, format(timestamp, 'yyyy-MM-dd') [timestamp]
, Reading
, COALESCE(Reading - LAG( Reading, 1) OVER (PARTITION BY [id] ORDER BY timestamp),0) [OneDayDifference]
, COALESCE(Reading - LAG( Reading, 7) OVER (PARTITION BY [id] ORDER BY timestamp),0) [SevDayDifference]
, reading - ca.tr [ThisMonthDiff]
FROM [dbo].[test_example] s
cross apply (
select top(1) t.reading tr
from [dbo].[test_example] t
where s.id = t.id
and t.timestamp >= dateadd(mm,datediff(mm,0,s.timestamp),0)
order by t.timestamp
) ca
ORDER BY
id
, timestamp DESC
;
+----+----+------------+---------+------------------+------------------+---------------+
| | id | timestamp | Reading | OneDayDifference | SevDayDifference | ThisMonthDiff |
+----+----+------------+---------+------------------+------------------+---------------+
| 1 | A1 | 2018-11-19 | 44182 | 0 | 338 | 541 |
| 2 | A1 | 2018-11-18 | 44182 | 0 | 338 | 541 |
| 3 | A1 | 2018-11-17 | 44182 | 38 | 338 | 541 |
| 18 | A1 | 2018-11-01 | 43641 | 38 | 194 | 0 |
| 19 | A1 | 2018-10-31 | 43603 | 38 | 275 | 275 |
| 20 | A1 | 2018-10-30 | 43565 | 39 | 0 | 237 |
| 21 | A1 | 2018-10-29 | 43526 | 0 | 0 | 198 |
| 22 | A1 | 2018-10-28 | 43526 | 0 | 0 | 198 |
| 23 | A1 | 2018-10-27 | 43526 | 41 | 0 | 198 |
| 24 | A1 | 2018-10-26 | 43485 | 38 | 0 | 157 |
| 25 | A1 | 2018-10-25 | 43447 | 119 | 0 | 119 |
| 26 | A1 | 2018-10-24 | 43328 | 0 | 0 | 0 |
+----+----+------------+---------+------------------+------------------+---------------+