Google BigQuery：具有重复记录值的表连接

Question

我有一张这样的桌子：

Row field1  field2  field3.a    field3.b     
1   value1  1       id1         key5     
                    id2         key6     
2   value2  2       id3         key7     
                    id4         key8

前 2 列是标准字段，第三列是 REPEATED RECORD 字段。

我还有 2 张桌子

Row  id     value    
1    id1    my-valueA    
2    id2    my-valueB    
3    id3    my-valueC    
4    id4    my-valueD

和

Row id      value    
1   key5    my-valueE    
2   key6    my-valueF    
3   key7    my-valueG    
4   key8    my-valueH

将第一个表中的 id/key 映射到给定的值（我使用了术语 key 和 id 以避免混淆，但最后的概念是相同的）

这里是重现结构的完整声明

#standardSQL
WITH my_table AS (
  SELECT "value1" as field1, 1 as field2, [STRUCT("id1" as a, "key5" as b),STRUCT("id2" as a, "key6" as b)] as field3
  UNION ALL
  SELECT "value2" as field2, 2 as field2, [STRUCT("id3" as a, "key7" as b),STRUCT("id4" as a, "key8" as b)] as field3
),

ids_table AS (
  SELECT "id1" as id, "my-valueA" as value
  UNION ALL
  SELECT "id2" as id, "my-valueB" as value
  UNION ALL
  SELECT "id3" as id, "my-valueC" as value
  UNION ALL
  SELECT "id4" as id, "my-valueD" as value
),

keys_table AS (
  SELECT "key5" as id, "my-valueE" as value
  UNION ALL
  SELECT "key6" as id, "my-valueF" as value
  UNION ALL
  SELECT "key7" as id, "my-valueG" as value
  UNION ALL
  SELECT "key8" as id, "my-valueH" as value
)

-- SELECT * FROM my_table
-- SELECT * FROM ids_table
-- SELECT * FROM keys_table

我的目标是用其他 2 个表给出的值替换第一个表中的键/id 值，就像经典的 id 连接一样。

这是预期的输出

Row field1  field2  t2_value    t3_value     
1   value1  1       my-valueA   my-valueE    
                    my-valueB   my-valueF    
2   value2  2       my-valueC   my-valueG    
                    my-valueD   my-valueH

一开始我想用UNNEST操作符来获取扁平线，所以可以做一个简单的JOIN来解析值，然后用替换的值重新加入数组。

SELECT my_table.* EXCEPT(field3),
  t2.value as t2_value,
  t3.value as t3_value
FROM my_table CROSS JOIN UNNEST(my_table.field3) AS t1 
  LEFT JOIN ids_table AS t2 ON t1.a = t2.id
  LEFT JOIN keys_table as t3 ON t1.b = t3.id

使用此语句，值可以从 id 正确替换为值，但现在我无法重现之前的 RECORD REPEATED 结构

Answer 1

以下是 BigQuery 标准 SQL

#standardSQL
WITH my_table AS (
  SELECT "value1" AS field1, 1 AS field2, [STRUCT("id1" AS a, "key5" AS b),STRUCT("id2" AS a, "key6" AS b)] AS field3 UNION ALL
  SELECT "value2" AS field2, 2 AS field2, [STRUCT("id3" AS a, "key7" AS b),STRUCT("id4" AS a, "key8" AS b)] AS field3
), ids_table AS (
  SELECT "id1" AS id, "my-valueA" AS value UNION ALL
  SELECT "id2" AS id, "my-valueB" AS value UNION ALL
  SELECT "id3" AS id, "my-valueC" AS value UNION ALL
  SELECT "id4" AS id, "my-valueD" AS value
),keys_table AS (
  SELECT "key5" AS id, "my-valueE" AS value UNION ALL
  SELECT "key6" AS id, "my-valueF" AS value UNION ALL
  SELECT "key7" AS id, "my-valueG" AS value UNION ALL
  SELECT "key8" AS id, "my-valueH" AS value
)
SELECT 
  field1, field2,
  (
    SELECT ARRAY_AGG(STRUCT<a_value STRING, b_value STRING>(t2.value, t3.value)) 
    FROM UNNEST(field3) t1
      LEFT JOIN ids_table AS t2 ON t1.a = t2.id
      LEFT JOIN keys_table AS t3 ON t1.b = t3.id
  ) AS field3
FROM my_table    

下面的输出

field1  field2  field3.a_value  field3.b_value   
value1  1       my-valueA       my-valueE    
                my-valueB       my-valueF    
value2  2       my-valueC       my-valueG    
                my-valueD       my-valueH