使用 SUM OVER 子句，如何仅在输出不大于某个值时检查期间的总和，否则使用当前月份值？
Using the SUM OVER clause, how to check sum over period only when output is not greater than a certain value, otherwise use current month value?
示例数据：
select date, agent, sales
from agentsales
date agent sales
2021-01-03 00:00:00.000 Agent A 10
2021-02-05 00:00:00.000 Agent A 15
2021-03-10 00:00:00.000 Agent A 10
2021-01-05 00:00:00.000 Agent B 5
2021-02-06 00:00:00.000 Agent B 28
2021-03-10 00:00:00.000 Agent B 5
2021-01-02 00:00:00.000 Agent C 35
2021-02-04 00:00:00.000 Agent C 25
2021-03-08 00:00:00.000 Agent C 15
2021-01-01 00:00:00.000 Agent D 5
2021-02-02 00:00:00.000 Agent D 35
2021-03-10 00:00:00.000 Agent D 31
我想获得销售量超过 30 次的代理商数量，这样 如果他们的销售量从未超过 30 次，则考虑当前和之前月份的总和，否则仅考虑当月。
预期输出：
YrMon Count_Agent_more_than_30_sales
Jan21 1
Feb21 2
Mar21 2
逻辑：
Jan21 - 1 since only C has crossed 30 sales
Feb21 - 2 since B and D have crossed 30 sales. Agent D has crossed the 30 mark in the month, and B has crossed over period for first time. C is not considered as it previously crossed the 30 mark.
Mar21 - 2 since A and D have crossed 30 sales. Agent A has crossed mark over period for 1st time. D has crossed for the month. B is not considered as periodic case was already considered in last month. C is not considered as it already crossed 30 mark last month.
如上所述，我想获得销售量超过 30 次的代理商的数量，这样 如果他们的销售量从未超过 30 次，则考虑当前和之前月份的总和，否则只有当月。
我计算期间总和的查询：
;WITH CTE AS (SELECT CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR) YRMON, AGENT, SUM(SALES) SALES
FROM AgentSales
GROUP BY CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR), AGENT
)
SELECT *, SUM(SALES) OVER(PARTITION BY AGENT ORDER BY YRMON) SUMOVERPERIOD FROM CTE
ORDER BY 2,1
YRMON AGENT SALES SUMOVERPERIOD
2021 1 Agent A 10 10
2021 2 Agent A 15 25
2021 3 Agent A 10 35
2021 1 Agent B 5 5
2021 2 Agent B 28 33
2021 3 Agent B 5 38
2021 1 Agent C 35 35
2021 2 Agent C 25 60
2021 3 Agent C 15 75
2021 1 Agent D 5 5
2021 2 Agent D 35 40
2021 3 Agent D 31 71
现在我尝试将逻辑应用于计算的总和：
;WITH CTE AS (SELECT CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR) YRMON, AGENT, SUM(SALES) SALES
FROM AgentSales
GROUP BY CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR), AGENT
)
SELECT *, SUM(SALES) OVER(PARTITION BY AGENT ORDER BY YRMON) SUMOVERPERIOD,
CASE WHEN SUM(SALES) OVER(PARTITION BY AGENT ORDER BY YRMON)>30 THEN 1 ELSE 0 END AS CALC
FROM CTE
ORDER BY 2,1
YRMON AGENT SALES SUMOVERPERIOD CALC
2021 1 Agent A 10 10 0
2021 2 Agent A 15 25 0
2021 3 Agent A 10 35 1
2021 1 Agent B 5 5 0
2021 2 Agent B 28 33 1
2021 3 Agent B 5 38 1
2021 1 Agent C 35 35 1
2021 2 Agent C 25 60 1
2021 3 Agent C 15 75 1
2021 1 Agent D 5 5 0
2021 2 Agent D 35 40 1
2021 3 Agent D 31 71 1
此查询始终考虑当前和上一期间的总和。
如何检查销售额之前是否超过了 30 个销售额标记，并且对于这种情况如何排除进行期间求和？例如，我们可以对 SUM OVER 列的结果应用 LAG 吗？
看起来这应该适合你
- 您需要预先汇总每个
agent 和 month 的销售额，然后得到该汇总的 运行 总和
- 然后通过将当前数据与 运行 总和
进行比较，简单地检查本月每一行是否交叉
SELECT
YrMon = FORMAT(Month, 'yyyy MM'),
Count_Agent_more_than_30_sales =
COUNT(CASE WHEN SumOverPeriod >= 30 AND SumOverPeriod - sales < 30 OR sales >= 30 THEN 1 END)
FROM (
SELECT
Month = EOMONTH(date),
agent,
sales = SUM(sales),
SumOverPeriod = SUM(SUM(sales)) OVER (PARTITION BY agent ORDER BY EOMONTH(date)
ROWS UNBOUNDED PRECEDING)
FROM AgentSales
GROUP BY EOMONTH(date), agent
) sales
GROUP BY Month;
请检查其中一项是否符合您的需求（我认为描述混乱）
选项 1
-- If you want to count only the first time [agent] crossed 30 sales
;With MyCTE01 as (
SELECT
[date] = EOMONTH([date], -1),
[agent],[sales],
S = SUM([sales]) OVER (PARTITION BY [agent] ORDER BY [date] ROWS BETWEEN UNBOUNDED PRECEDING and CURRENT ROW)
FROM [AgentSales]
),
MyCTE02 as (
SELECT [date],[agent],[sales], S
FROM MyCTE01
-- The idea of using "and S - [sales] < 30" instead of ROW_NUMBER came from @Charlieface, but it is better to do the work on DATE data type and not on string
WHERE S > 30 and S - [sales] < 30
)
SELECT DATENAME(month,[Date]), YEAR([Date]), COUNT(*)
FROM MyCTE02
GROUP BY [date]
GO
选项 2
-- If you want to count all the [agent] crossed 30 sales till now
;With MyCTE01 as (
SELECT
[date] = DATEADD(DAY, 1, EOMONTH([date], -1)),
[agent],[sales],
S = SUM([sales]) OVER (PARTITION BY [agent] ORDER BY [date] ROWS BETWEEN UNBOUNDED PRECEDING and CURRENT ROW)
FROM [AgentSales]
)
,MyCTE02 as (
SELECT [date],[agent],[sales], S
FROM MyCTE01
WHERE S > 30
)
SELECT DATENAME(month,[Date]), YEAR([Date]), COUNT(*)
FROM MyCTE02
GROUP BY [date]
GO
选项 3
-- If you want to count only the first time [agent] crossed 30 sales or when the sales or over 30
;With MyCTE01 as (
SELECT
[date] = DATEADD(DAY,1,EOMONTH([date], -1)),
[agent],[sales],
S = SUM([sales]) OVER (PARTITION BY [agent] ORDER BY [date] ROWS BETWEEN UNBOUNDED PRECEDING and CURRENT ROW)
FROM [AgentSales]
)
,MyCTE02 as (
SELECT [date],[agent],[sales], S
FROM MyCTE01
-- The idea of using "and S - [sales] < 30" instead of ROW_NUMBER came from @Charlieface, but it is better to do the work on DATE data type and not on string
WHERE (S > 30 and S - [sales] < 30) or sales > 30
)
SELECT DATENAME(month,[Date]), YEAR([Date]), COUNT(*)
FROM MyCTE02
GROUP BY [date]
GO
DDL+DML
USE tempdb
GO
DROP TABLE IF EXISTS [AgentSales]
GO
CREATE TABLE [AgentSales](id INT IDENTITY(1,1), [date] DATE, agent VARCHAR(100), sales INT)
GO
INSERT [AgentSales]([date],[agent],[sales]) VALUES
('2021-01-03 00:00:00.000','Agent A', 10),
('2021-02-05 00:00:00.000','Agent A', 15),
('2021-03-10 00:00:00.000','Agent A',10),
('2021-01-05 00:00:00.000','Agent B',5 ),
('2021-02-06 00:00:00.000','Agent B',28),
('2021-03-10 00:00:00.000','Agent B',5 ),
('2021-01-02 00:00:00.000','Agent C',35),
('2021-02-04 00:00:00.000','Agent C',25),
('2021-03-08 00:00:00.000','Agent C',15),
('2021-01-01 00:00:00.000','Agent D',5 ),
('2021-02-02 00:00:00.000','Agent D',35),
('2021-03-10 00:00:00.000','Agent D',31)
GO
SELECT [id],[date],[agent],[sales]
FROM [AgentSales]
GO
示例数据：
select date, agent, sales
from agentsales
date agent sales
2021-01-03 00:00:00.000 Agent A 10
2021-02-05 00:00:00.000 Agent A 15
2021-03-10 00:00:00.000 Agent A 10
2021-01-05 00:00:00.000 Agent B 5
2021-02-06 00:00:00.000 Agent B 28
2021-03-10 00:00:00.000 Agent B 5
2021-01-02 00:00:00.000 Agent C 35
2021-02-04 00:00:00.000 Agent C 25
2021-03-08 00:00:00.000 Agent C 15
2021-01-01 00:00:00.000 Agent D 5
2021-02-02 00:00:00.000 Agent D 35
2021-03-10 00:00:00.000 Agent D 31
我想获得销售量超过 30 次的代理商数量，这样 如果他们的销售量从未超过 30 次，则考虑当前和之前月份的总和，否则仅考虑当月。
预期输出：
YrMon Count_Agent_more_than_30_sales
Jan21 1
Feb21 2
Mar21 2
逻辑：
Jan21 - 1 since only C has crossed 30 sales
Feb21 - 2 since B and D have crossed 30 sales. Agent D has crossed the 30 mark in the month, and B has crossed over period for first time. C is not considered as it previously crossed the 30 mark.
Mar21 - 2 since A and D have crossed 30 sales. Agent A has crossed mark over period for 1st time. D has crossed for the month. B is not considered as periodic case was already considered in last month. C is not considered as it already crossed 30 mark last month.
如上所述，我想获得销售量超过 30 次的代理商的数量，这样 如果他们的销售量从未超过 30 次，则考虑当前和之前月份的总和，否则只有当月。
我计算期间总和的查询：
;WITH CTE AS (SELECT CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR) YRMON, AGENT, SUM(SALES) SALES
FROM AgentSales
GROUP BY CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR), AGENT
)
SELECT *, SUM(SALES) OVER(PARTITION BY AGENT ORDER BY YRMON) SUMOVERPERIOD FROM CTE
ORDER BY 2,1
YRMON AGENT SALES SUMOVERPERIOD
2021 1 Agent A 10 10
2021 2 Agent A 15 25
2021 3 Agent A 10 35
2021 1 Agent B 5 5
2021 2 Agent B 28 33
2021 3 Agent B 5 38
2021 1 Agent C 35 35
2021 2 Agent C 25 60
2021 3 Agent C 15 75
2021 1 Agent D 5 5
2021 2 Agent D 35 40
2021 3 Agent D 31 71
现在我尝试将逻辑应用于计算的总和：
;WITH CTE AS (SELECT CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR) YRMON, AGENT, SUM(SALES) SALES
FROM AgentSales
GROUP BY CAST(YEAR([DATE]) AS VARCHAR)+' '+CAST(MONTH([DATE]) AS VARCHAR), AGENT
)
SELECT *, SUM(SALES) OVER(PARTITION BY AGENT ORDER BY YRMON) SUMOVERPERIOD,
CASE WHEN SUM(SALES) OVER(PARTITION BY AGENT ORDER BY YRMON)>30 THEN 1 ELSE 0 END AS CALC
FROM CTE
ORDER BY 2,1
YRMON AGENT SALES SUMOVERPERIOD CALC
2021 1 Agent A 10 10 0
2021 2 Agent A 15 25 0
2021 3 Agent A 10 35 1
2021 1 Agent B 5 5 0
2021 2 Agent B 28 33 1
2021 3 Agent B 5 38 1
2021 1 Agent C 35 35 1
2021 2 Agent C 25 60 1
2021 3 Agent C 15 75 1
2021 1 Agent D 5 5 0
2021 2 Agent D 35 40 1
2021 3 Agent D 31 71 1
此查询始终考虑当前和上一期间的总和。
如何检查销售额之前是否超过了 30 个销售额标记，并且对于这种情况如何排除进行期间求和？例如，我们可以对 SUM OVER 列的结果应用 LAG 吗？
看起来这应该适合你
- 您需要预先汇总每个
agent和month的销售额，然后得到该汇总的 运行 总和 - 然后通过将当前数据与 运行 总和 进行比较，简单地检查本月每一行是否交叉
SELECT
YrMon = FORMAT(Month, 'yyyy MM'),
Count_Agent_more_than_30_sales =
COUNT(CASE WHEN SumOverPeriod >= 30 AND SumOverPeriod - sales < 30 OR sales >= 30 THEN 1 END)
FROM (
SELECT
Month = EOMONTH(date),
agent,
sales = SUM(sales),
SumOverPeriod = SUM(SUM(sales)) OVER (PARTITION BY agent ORDER BY EOMONTH(date)
ROWS UNBOUNDED PRECEDING)
FROM AgentSales
GROUP BY EOMONTH(date), agent
) sales
GROUP BY Month;
请检查其中一项是否符合您的需求（我认为描述混乱）
选项 1
-- If you want to count only the first time [agent] crossed 30 sales
;With MyCTE01 as (
SELECT
[date] = EOMONTH([date], -1),
[agent],[sales],
S = SUM([sales]) OVER (PARTITION BY [agent] ORDER BY [date] ROWS BETWEEN UNBOUNDED PRECEDING and CURRENT ROW)
FROM [AgentSales]
),
MyCTE02 as (
SELECT [date],[agent],[sales], S
FROM MyCTE01
-- The idea of using "and S - [sales] < 30" instead of ROW_NUMBER came from @Charlieface, but it is better to do the work on DATE data type and not on string
WHERE S > 30 and S - [sales] < 30
)
SELECT DATENAME(month,[Date]), YEAR([Date]), COUNT(*)
FROM MyCTE02
GROUP BY [date]
GO
选项 2
-- If you want to count all the [agent] crossed 30 sales till now
;With MyCTE01 as (
SELECT
[date] = DATEADD(DAY, 1, EOMONTH([date], -1)),
[agent],[sales],
S = SUM([sales]) OVER (PARTITION BY [agent] ORDER BY [date] ROWS BETWEEN UNBOUNDED PRECEDING and CURRENT ROW)
FROM [AgentSales]
)
,MyCTE02 as (
SELECT [date],[agent],[sales], S
FROM MyCTE01
WHERE S > 30
)
SELECT DATENAME(month,[Date]), YEAR([Date]), COUNT(*)
FROM MyCTE02
GROUP BY [date]
GO
选项 3
-- If you want to count only the first time [agent] crossed 30 sales or when the sales or over 30
;With MyCTE01 as (
SELECT
[date] = DATEADD(DAY,1,EOMONTH([date], -1)),
[agent],[sales],
S = SUM([sales]) OVER (PARTITION BY [agent] ORDER BY [date] ROWS BETWEEN UNBOUNDED PRECEDING and CURRENT ROW)
FROM [AgentSales]
)
,MyCTE02 as (
SELECT [date],[agent],[sales], S
FROM MyCTE01
-- The idea of using "and S - [sales] < 30" instead of ROW_NUMBER came from @Charlieface, but it is better to do the work on DATE data type and not on string
WHERE (S > 30 and S - [sales] < 30) or sales > 30
)
SELECT DATENAME(month,[Date]), YEAR([Date]), COUNT(*)
FROM MyCTE02
GROUP BY [date]
GO
DDL+DML
USE tempdb
GO
DROP TABLE IF EXISTS [AgentSales]
GO
CREATE TABLE [AgentSales](id INT IDENTITY(1,1), [date] DATE, agent VARCHAR(100), sales INT)
GO
INSERT [AgentSales]([date],[agent],[sales]) VALUES
('2021-01-03 00:00:00.000','Agent A', 10),
('2021-02-05 00:00:00.000','Agent A', 15),
('2021-03-10 00:00:00.000','Agent A',10),
('2021-01-05 00:00:00.000','Agent B',5 ),
('2021-02-06 00:00:00.000','Agent B',28),
('2021-03-10 00:00:00.000','Agent B',5 ),
('2021-01-02 00:00:00.000','Agent C',35),
('2021-02-04 00:00:00.000','Agent C',25),
('2021-03-08 00:00:00.000','Agent C',15),
('2021-01-01 00:00:00.000','Agent D',5 ),
('2021-02-02 00:00:00.000','Agent D',35),
('2021-03-10 00:00:00.000','Agent D',31)
GO
SELECT [id],[date],[agent],[sales]
FROM [AgentSales]
GO