将 BigQuery 数组解析为列的最佳方法是什么

Question

我有一张这样的桌子

select 'Alice' as Name, ['a=1','b=2','c=3'];

我希望它把它变成这个

select 'Alice' as Name, 1 as a, 2 as b, 3 as c

最好的方法是什么？

我正在考虑也许首先使用结构

select 'Alice' as Name, [struct('a' as Letter, 1 as Number),struct('b' as Letter, 2 as Number) ,struct('c' as Letter, 3 as Number)]  as struct_column

Answer 1

假设您事先不知道“未来”列的名称和数量 - 我建议您将其展平，如下例所示（BigQuery 标准 SQL）

#standardSQL
WITH `project.dataset.table` AS (
  SELECT 'Alice' AS Name, ['a=1','b=2','c=3'] attributes UNION ALL
  SELECT 'Cheshire Cat', ['a=4', 'x=5'] UNION ALL
  SELECT 'White Rabbit', ['a=6', 'c=7'] 
)
SELECT Name, 
  SPLIT(kv, '=')[OFFSET(0)] key, 
  SPLIT(kv, '=')[SAFE_OFFSET(1)] value  
FROM `project.dataset.table`, UNNEST(attributes) kv   

结果

Row Name            key value    
1   Alice           a   1    
2   Alice           b   2    
3   Alice           c   3    
4   Cheshire Cat    a   4    
5   Cheshire Cat    x   5    
6   White Rabbit    a   6    
7   White Rabbit    c   7     

我确实提前知道属性...

在这种情况下，下面将起作用

#standardSQL
WITH `project.dataset.table` AS (
  SELECT 'Alice' AS Name, ['a=1','b=2','c=3'] attributes UNION ALL
  SELECT 'Cheshire Cat', ['a=4', 'b=5'] UNION ALL
  SELECT 'White Rabbit', ['a=6', 'c=7'] 
)
SELECT Name,
  MAX(IF(key = 'a', value, NULL)) a,
  MAX(IF(key = 'b', value, NULL)) b,
  MAX(IF(key = 'c', value, NULL)) c
FROM (
  SELECT Name, 
    SPLIT(kv, '=')[OFFSET(0)] key, 
    SPLIT(kv, '=')[SAFE_OFFSET(1)] value  
  FROM `project.dataset.table`, UNNEST(attributes) kv   
)
GROUP BY Name   

结果

Row Name            a       b       c    
1   Alice           1       2       3    
2   Cheshire Cat    4       5       null     
3   White Rabbit    6       null    7