sql - Snowflake 中的 STDDEV Windows 函数 : not getting a uniform value according to the partition

Question

我想获得按reg、日期和小时划分的加速度的标准差。因此，我希望使用以下代码根据这些分区获得统一的值，但我为每一行获得不同的值:

SELECT *,
       STDDEV(ACCELERATION)OVER(PARTITION BY REG, DATE, HOUR ORDER BY TIMESTAMP)
  
FROM 

(
  
SELECT 
       *,
       SPEED_DIFFERENCE/ TIMESTAMP_DIFFERENCE AS ACCELERATION

FROM 

(



SELECT *,
       NEXT_MILES_PER_HOUR - MILES_PER_HOUR AS SPEED_DIFFERENCE,
       TIMESTAMPDIFF(second,timestamp,Next_timestamp) AS TIMESTAMP_DIFFERENCE
       
FROM

(

SELECT *, 
       TO_TIME (DATE_AND_HOUR) AS HOUR,
       LEAD(TIMESTAMP)OVER (PARTITION BY VIN ORDER BY TIMESTAMP) AS NEXT_TIMESTAMP,
       LEAD(MILES_PER_HOUR)OVER (PARTITION BY REG ORDER BY TIMESTAMP) AS NEXT_MILES_PER_HOUR

FROM

(

SELECT Reg, 
       TIMESTAMP,
       Miles_per_hour
       TO_DATE (TIMESTAMP) AS DATE,
       TO_TIME (TIMESTAMP) AS TIME,
       date_trunc('HOUR', TIMESTAMP) as DATE_AND_Hour,
       DATA
       
FROM Motorycle_Data_Refurbished_Models_DuPage_County

)
)
)
)

Answer 1

要为每个分区获取单个值，应删除 ORDER BY，因为它将函数切换为累积函数:

SELECT *,
       STDDEV(ACCELERATION)OVER(PARTITION BY REG, DATE, HOUR)
...

示例:

CREATE  OR REPLACE TABLE t(i INTEGER);
INSERT INTO t (i) VALUES  (6), (10), (14);

SELECT STDDEV(i) OVER(ORDER BY i) FROM t;

输出:

对比

SELECT STDDEV(i) OVER() FROM t;

输出:

---

应用于 STDDEV 时的 ORDER BY :

If the OVER clause contains an ORDER BY subclause, then:

A window frame is required. If no window frame is specified explicitly, then the ORDER BY implies a cumulative window frame:

   RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW

窗口正在从分区的开头移动到当前行。

可以覆盖它:

SELECT STDDEV(i) OVER(ORDER BY i 
                      RANGE BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) 
FROM t;

输出: