如何在 BigQuery 中获取每天的累积记录

Question

我有一个数据如下。

|-----------|-------------|---------------|
|order_date | customer_id | product_id    |
|-----------|-------------|---------------|
|2020-01-01 | 123456      | 0001          |
|-----------|-------------|---------------|
|2020-01-02 | 123456      | 0005          |
|-----------|-------------|---------------|
|2020-01-03 |123456       | 0010          |
|-----------|-------------|---------------|

然后我想像这样每天累积product_id。

|-----------|-------------|----------------------------|
|order_date |customer_id  |count_cumulative_product_id |
|-----------|-------------|----------------------------|
|2020-01-01 |123456       |1                           |
|-----------|-------------|----------------------------|
|2020-01-02 |123456       |2                           |
|-----------|-------------|----------------------------|
|2020-01-03 |123456       |3                           |         
|-----------|-------------|----------------------------|

我不知道什么样的查询可以解决这个问题......

Answer 1

以下是 BigQuery 标准 SQL

#standardSQL
SELECT *, 
  COUNT(1) OVER(PARTITION BY customer_id ORDER BY order_date) count_cumulative_product_id
FROM `project.dataset.table`

您可以使用您问题中的示例数据进行测试，使用上面的示例，如下例所示

#standardSQL
WITH `project.dataset.table` AS (
  SELECT '2020-01-01' order_date, 123456 customer_id, '0001' product_id UNION ALL
  SELECT '2020-01-02', 123456, '0005' UNION ALL
  SELECT '2020-01-03', 123456, '0010' 
)
SELECT *, 
  COUNT(1) OVER(PARTITION BY customer_id ORDER BY order_date) count_cumulative_product_id
FROM `project.dataset.table`
-- ORDER BY order_date   

结果

Row order_date  customer_id product_id  count_cumulative_product_id  
1   2020-01-01  123456      0001        1    
2   2020-01-02  123456      0005        2    
3   2020-01-03  123456      0010        3

Answer 2

如果您不担心distinct product_ids 的累积计数，那么您可以简单地使用“移动窗口”方法：

select 
   order_date,
   customer_id,
   count(product_id) over (order by product_id range between unbounded preceding and current row) as cumulative_product_ids
from `dataset.table`

但是，如果你想要不同 product_ids 的累积计数，那么你可以使用类似的东西：

select order_date, customer_id, count(distinct x) as cumulative_product_ids from ( 
   select 
      order_date, 
      customer_id, 
      array_agg(product_id) over (order by product_id range between unbounded preceding and current row) as cumulative_product_ids
from `dataset.table`
), unnest(cumulative_product_ids) as x
group by 1,2

希望对你有帮助。