如何在 Google Big Query 中计算上周一列的总和？答案

【问题标题】：How to calculate the sum of a column for last week in Google Big Query?如何在 Google Big Query 中计算上周一列的总和？
【发布时间】：2018-04-19 14:53:17
【问题描述】：

基于米哈伊尔 (How to calculate rolling average of same weekday as current row in Big Query?) 的回答

我希望复制这一点，但改为使用周数。以下是我的查询。似乎因为一次 7 行的周数相同，所以它不能按预期工作

SELECT id, 
       date, 
       weeknumber, 
       sales_total,
       SUM(net_sales_total) OVER(rolling_last_week) last_week
FROM (
     SELECT id, 
            date,  
            sales_total,
            EXTRACT(WEEK FROM entry_date) weeknumber
     FROM `mydataset.mytable`   
)
WINDOW rolling_last_week AS (
   PARTITION BY id, weeknumber
   ORDER BY entry_date 
   ROWS BETWEEN 1 PRECEDING AND 1 PRECEDING
)
ORDER BY weeknumber, entry_date

表_A

+-------+---------+----------+-----------+
| id    |  date   |weeknumber|sales_total|
+-------+---------+----------+-----------+
| 1     | 01-01-17|    1     |     5     |
| 1     | 01-02-17|    1     |     5     |
| 1     | 01-03-17|    1     |     5     |
| 1     | 01-04-17|    1     |     5     |
| 1     | 01-05-17|    1     |     5     |
| 1     | 01-06-17|    1     |     5     |
| 1     | 01-07-17|    1     |     5     |
| 1     | 01-08-17|    2     |     10    |
| 1     | 01-09-17|    2     |     10    |
| 1     | 01-10-17|    2     |     10    |
| 1     | 01-11-17|    2     |     10    |
| 1     | 01-12-17|    2     |     10    |
| 1     | 01-13-17|    2     |     10    |
| 1     | 01-14-17|    2     |     10    |
+-------+---------+----------+-----------+

我希望结果查询返回上周每一天的销售额总和

+-------+---------+----------+-----------+-----------+
| id    |  date   |weeknumber|sales_total| last_week |
+-------+---------+----------+-----------+-----------+
| 1     | 01-01-17|    1     |     5     | .         |
| 1     | 01-02-17|    1     |     5     | .         |
| 1     | 01-03-17|    1     |     5     | .         |
| 1     | 01-04-17|    1     |     5     | .         |
| 1     | 01-05-17|    1     |     5     | .         |
| 1     | 01-06-17|    1     |     5     | .         |
| 1     | 01-07-17|    1     |     5     | .         |
| 1     | 01-08-17|    2     |     10    | 35        |
| 1     | 01-09-17|    2     |     10    | 35        |
| 1     | 01-10-17|    2     |     10    | 35        |
| 1     | 01-11-17|    2     |     10    | 35        |
| 1     | 01-12-17|    2     |     10    | 35        |
| 1     | 01-13-17|    2     |     10    | 35        |
| 1     | 01-14-17|    2     |     10    | 35        |
+-------+---------+----------+-----------+-----------+

【问题讨论】：

标签： sql google-bigquery

【解决方案1】：

以下是 BigQuery 标准 SQL 的示例

#standardSQL
SELECT *, 
  IFNULL(SUM(sales_total) 
    OVER(rolling_last_week), 0) last_week 
FROM `project.dataset.your_table`
WINDOW rolling_last_week AS (
   PARTITION BY id
   ORDER BY DATE_DIFF(PARSE_DATE('%m-%d-%y', sales_date), '2000-01-01', WEEK)
   RANGE BETWEEN 1 PRECEDING AND 1 PRECEDING
)
ORDER BY id, sales_date

您可以使用您问题中的虚拟数据来测试/玩它，如下所示

#standardSQL
WITH `project.dataset.your_table` AS (
  SELECT 1 id, '01-01-17' sales_date, 1 weeknumber, 5 sales_total UNION ALL
  SELECT 1, '01-02-17', 1, 5 UNION ALL
  SELECT 1, '01-03-17', 1, 5 UNION ALL
  SELECT 1, '01-04-17', 1, 5 UNION ALL
  SELECT 1, '01-05-17', 1, 5 UNION ALL
  SELECT 1, '01-06-17', 1, 5 UNION ALL
  SELECT 1, '01-07-17', 1, 5 UNION ALL
  SELECT 1, '01-08-17', 2, 10 UNION ALL
  SELECT 1, '01-09-17', 2, 10 UNION ALL
  SELECT 1, '01-10-17', 2, 10 UNION ALL
  SELECT 1, '01-11-17', 2, 10 UNION ALL
  SELECT 1, '01-12-17', 2, 10 UNION ALL
  SELECT 1, '01-13-17', 2, 10 UNION ALL
  SELECT 1, '01-14-17', 2, 10 
)
SELECT *, 
  IFNULL(SUM(sales_total) 
    OVER(rolling_last_week), 0) last_week 
FROM `project.dataset.your_table`
WINDOW rolling_last_week AS (
   PARTITION BY id
   ORDER BY DATE_DIFF(PARSE_DATE('%m-%d-%y', sales_date), '2000-01-01', WEEK)
   RANGE BETWEEN 1 PRECEDING AND 1 PRECEDING
)
ORDER BY id, sales_date

结果

Row id  sales_date  weeknumber  sales_total last_week    
1   1   01-01-17    1           5           0    
2   1   01-02-17    1           5           0    
3   1   01-03-17    1           5           0    
4   1   01-04-17    1           5           0    
5   1   01-05-17    1           5           0    
6   1   01-06-17    1           5           0    
7   1   01-07-17    1           5           0    
8   1   01-08-17    2           10          35   
9   1   01-09-17    2           10          35   
10  1   01-10-17    2           10          35   
11  1   01-11-17    2           10          35   
12  1   01-12-17    2           10          35   
13  1   01-13-17    2           10          35   
14  1   01-14-17    2           10          35

【讨论】：