Bigquery 中的多个重复结构

Question

跟进此事-Bigquery multiple unnest in a single select

我们使用 bigquery 作为我们的仓储解决方案，并试图通过整合来突破极限。一个简单的例子是客户跟踪。客户产生收入，在我们的网站上有多个接触点，并在我们这里独立维护多个帐户。对于想要对客户进行行为分析的业务用户，他们想要跟踪访问量、产生的收入以及他们的帐户如何影响留存率，我们正在尝试评估嵌套结构是否适合我们

下面是一个例子。我有 3 张桌子。

客户 (C)

C_Key| C_Name

-----|------

1 | ABC

2 |防御

---

帐户 (A)

A_Key | C_Key
11 | 1
12 | 1
21 | 2
22 | 2

23 | 2

收入 (R)

R_Key | C_Key |收入

-------|---------|----------

11 | 1 | $10

12 | 1 | $20

21 | 2 | $10

我使用 array_agg 将这三个组合成一个嵌套表，如下所示：

{Client,
    Accounts:
          [{
            }],
    Revenue:
          [{
              }]
  }

我希望能够在单个查询中使用多个取消嵌套，如下所示

 Select client, Count Distinct(Accounts) and SUM(Revenue) from <single nested 
    table>, unnest accounts, unnest revenue

预期输出为 2 行，

1,2,30 美元

2,3,10 美元

但是，在同一个查询中有多个未嵌套会导致交叉连接。
实际输出为

1,2,60 美元

2,3,30 美元

Answer 1

以下是 BigQuery 标准 SQL

首先让我们澄清single nested table的创建

我希望你做了类似的事情：

#standardSQL
WITH clients AS (
  SELECT 1 AS c_key, 'abc' AS c_name UNION ALL
  SELECT 2, 'def'
), accounts AS (
  SELECT 11 AS a_key, 1 AS c_key UNION ALL
  SELECT 12, 1 UNION ALL
  SELECT 21, 2 UNION ALL
  SELECT 22, 2 UNION ALL
  SELECT 23, 2
), revenue AS (
  SELECT 11 AS r_key, 1 AS c_key, 10 AS revenue UNION ALL
  SELECT 12, 1, 20 UNION ALL
  SELECT 21, 2, 10
), single_nested_table AS (
  SELECT x.c_key, x.c_name, accounts, revenue 
  FROM (
    SELECT c.c_key, c_name, ARRAY_AGG(a) AS accounts --, array_agg(r) as revenue  
    FROM clients AS c
    LEFT JOIN accounts AS a ON a.c_key = c.c_key
    GROUP BY c.c_key, c_name
  ) x
  JOIN (
    SELECT c.c_key, c_name, ARRAY_AGG(r) AS revenue  
    FROM clients AS c
    LEFT JOIN revenue AS r ON r.c_key = c.c_key
    GROUP BY c.c_key, c_name
  ) y
  ON x.c_key = y.c_key
)
SELECT *
FROM single_nested_table  

创建表为

Row c_key c_name accounts.a_key accounts.c_key revenue.r_key revenue.c_key revenue.revenue
1   1     abc    11             1              11            1             10    
                 12             1              12            1             20    
2   2     def    21             2              21            2             10    
                 22             2                
                 23             2                

您用于创建该表的确切查询并不那么重要 - 但清除结构/架构很重要！

那么现在，回到你的问题

#standardSQL
WITH clients AS (
  SELECT 1 AS c_key, 'abc' AS c_name UNION ALL
  SELECT 2, 'def'
), accounts AS (
  SELECT 11 AS a_key, 1 AS c_key UNION ALL
  SELECT 12, 1 UNION ALL
  SELECT 21, 2 UNION ALL
  SELECT 22, 2 UNION ALL
  SELECT 23, 2
), revenue AS (
  SELECT 11 AS r_key, 1 AS c_key, 10 AS revenue UNION ALL
  SELECT 12, 1, 20 UNION ALL
  SELECT 21, 2, 10
), single_nested_table AS (
  SELECT x.c_key, x.c_name, accounts, revenue 
  FROM (
    SELECT c.c_key, c_name, ARRAY_AGG(a) AS accounts --, array_agg(r) as revenue  
    FROM clients AS c
    LEFT JOIN accounts AS a ON a.c_key = c.c_key
    GROUP BY c.c_key, c_name
  ) x
  JOIN (
    SELECT c.c_key, c_name, ARRAY_AGG(r) AS revenue  
    FROM clients AS c
    LEFT JOIN revenue AS r ON r.c_key = c.c_key
    GROUP BY c.c_key, c_name
  ) y
  ON x.c_key = y.c_key
)
SELECT 
  c_key, c_name, 
  ARRAY_LENGTH(accounts) AS distinct_accounts, 
  (SELECT SUM(revenue) FROM UNNEST(revenue)) AS revenue
FROM single_nested_table   

这给出了你所要求的：

Row c_key   c_name  distinct_accounts   revenue  
1   1       abc     2                   30   
2   2       def     3                   10