在 bigquery 中运行多个条件的总计

Question

我需要计算运行总计，但需要在某个条件下重置总计（当预期达到 = 0 并且 product_group 和产品更改时）。在没有两个额外字段的情况下在这里获得帮助：Calculate a running total with a condition in BigQuery 我有这张表，可以使用 product_group 和 product 作为整数或字符串，如下所示。

Date, Product_group, Product, Registrations, Expected Registrations, Expected Reached, Running Total
            2020-03-01,A, Bikes, 5, 4,1, 1
            2020-03-02,A, Bikes, 7, 5,1, 2
            2020-03-03,A, Bikes, 8, 6,1, 3
            2020-03-04,A, Bikes, 2, 5,0, 0
            2020-03-05,A, Bikes, 5, 4,1, 1
            2020-03-06,A, Bikes, 7, 5,1, 2 
            2020-03-04,B, Cars , 2, 5,0, 0
            2020-03-05,B, Cars , 5, 4,1, 1
            2020-03-06,B, Cars , 7, 5,1, 2
            2020-03-07,B, Cars , 8, 6,1, 3 
            2020-03-08,C, Plane, 2, 5,0, 0

有关如何调整此查询（来自另一篇帖子的答案）的任何建议，在没有两个额外字段的情况下可以正常工作-

#standardSQL
SELECT * EXCEPT(grp), 
  SUM(Expected_reached) OVER(PARTITION BY grp ORDER BY `date`) Running_Total
FROM (
  SELECT *, COUNTIF(Expected_reached = 0) OVER(ORDER BY `date`) grp 
  FROM `project.dataset.table`
)

问题是 COUNTIF(Expected_reached = 0) OVER(ORDER BYdate) grp 在 product_group 或 product 更改并且我得到非唯一组时重新开始，因此运行总数 SUM(Expected_reached) OVER(PARTITION BY grp ORDER BYdate) Running_Total 计算不正确。

Answer 1

您只需要在两个分析函数中添加PARTITION BY Product_group, Product

#standardSQL
SELECT * EXCEPT(grp), 
  SUM(Expected_reached) OVER(PARTITION BY Product_group, Product, grp ORDER BY `date`) Running_Total
FROM (
  SELECT *, COUNTIF(Expected_reached = 0) OVER(PARTITION BY Product_group, Product ORDER BY `date`) grp 
  FROM `project.dataset.table`
)

Answer 2

就问题而言，您只需在窗口函数的分区子句中添加两个额外的列，product_group 和 product：

select 
    * except(grp), 
    sum(expected_reached) 
        over(partition by grp, product_group, product order by `date`) running_total
from (
    select 
        *, 
        countif(expected_reached = 0) 
            over(partition by product_group, product order by `date`) grp 
    from `project.dataset.table` 
)