在 BigQuery 表中展平多个相同大小的数组列

Posted 2023-02-23

【中文标题】在 BigQuery 表中展平多个相同大小的数组列

【英文标题】：Flatten multiple same-sized array columns in BigQuery table

【发布时间】：2020-03-01 17:33:20

【问题描述】：

我有一个包含几列的表，其中一些列是长度相同的数组。我想解除它们的嵌套，以获得一个结果，其中包含来自不同行的数组中的值。

所以有这样一张桌子：

我想去：

这是其中一个数组列的工作方式：

WITH data AS
(
  SELECT 1001 as id, ['a', 'b', 'c'] as array_1, [1, 2, 3] as array_2
  UNION ALL
  SELECT 1002 as id, ['d', 'e', 'f', 'g'] as array_1, [4, 5, 6, 7] as array_2
  UNION ALL
  SELECT 1003 as id, ['h', 'i'] as array_1, [8, 9] as array_2
)
SELECT id, a1
FROM data,
UNNEST(array_1) as a1

有没有一些优雅的方法可以同时解除两个数组的嵌套？我想避免单独取消嵌套每一列，然后将所有内容连接在一起。

【参考方案1】：

您可以使用with offset 和join：

WITH data AS
(
  SELECT 1001 as id, ['a', 'b', 'c'] as array_1, [1, 2, 3] as array_2
  UNION ALL
  SELECT 1002 as id, ['d', 'e', 'f', 'g'] as array_1, [4, 5, 6, 7] as array_2
  UNION ALL
  SELECT 1003 as id, ['h', 'i'] as array_1, [8, 9] as array_2
)
SELECT id, a1, a2
FROM data cross join
     UNNEST(array_1) as a1 with offset n1 JOIN
     UNNEST(array_2) as a2 with offset n2 
     on n1 = n2

【参考方案2】：

以下是 BigQuery 标准 SQL

#standardSQL
SELECT id, a1, a2
FROM data, UNNEST(array_1) AS a1 WITH OFFSET 
JOIN UNNEST(array_2) AS a2 WITH OFFSET
USING(OFFSET)

【参考方案3】：

所以我自己做了一些关于在 SQL 中取消嵌套的研究，并提出了这个解决方案：

WITH data AS
(
  SELECT 1001 as id, ['a', 'b', 'c'] as array_1, [1, 2, 3] as array_2
  UNION ALL
  SELECT 1002 as id, ['d', 'e', 'f', 'g'] as array_1, [4, 5, 6, 7] as array_2
  UNION ALL
  SELECT 1003 as id, ['h', 'i'] as array_1, [8, 9] as array_2
)
SELECT id, a1, array_2[OFFSET(off)] AS a2
FROM data
CROSS JOIN UNNEST(array_1) AS a1 WITH OFFSET off

优点是它不需要取消嵌套所有数组，只需要一个。

【讨论】：

谢谢，这解决了我遇到的类似问题。您能否指向文档以阅读有关 offset 关键字的更多信息？我似乎不明白它在这种情况下是如何工作的。