当键为空时，Bigquery 合并连接

Posted 2023-02-16

【中文标题】当键为空时，Bigquery 合并连接

【英文标题】：Bigquery coalesce join when key is null

【发布时间】：2016-07-09 18:38:06

【问题描述】：

表一

+---------+-----------+--------+
| user_id | email     | action |
+---------+-----------+--------+
| 1       | [e-mail] | open   |
+---------+-----------+--------+
| 2       | null      | click  |
+---------+-----------+--------+
| 3       | [e-mail] | click  |
+---------+-----------+--------+
| 4       | [e-mail] | open   |
+---------+-----------+--------+

表 2

+---------+-----------+--------+
| user_id | email     | event  |
+---------+-----------+--------+
| 1       | [e-mail] | sent   |
+---------+-----------+--------+
| null    | [e-mail] | none   |
+---------+-----------+--------+
| 2       | [e-mail] | sent   |
+---------+-----------+--------+
| 4       | [e-mail] | sent   |
+---------+-----------+--------+

我想基于 t1.user_id = t2.user_id 加入，但是当 key 为空时，加入 t1.email = t2.email

我尝试了多种方式在 bigquery 中加入： 1.)在 COALESCE(t1.user_id,t1.email) = COALESCE(t2.user_id, t2.email) 2.）在 t2.user_id 不为空时打开，然后 t1.user_id = t2.user_id 否则 t1.email = t2.email 结束

都不行。如何做到这一点？

【问题讨论】：

SQL 有“IS DISTINCT”和“IS NOT DISTINCT”，但我在 BigQuery 中看不到。

【参考方案1】：

我会将这样的连接拆分为两个单独的： 首先 - 通过 user_id 加入

SELECT *
FROM table1 AS t1
JOIN table2 AS t2
ON t1.user_id = t2.user_id

第二次 - 通过电子邮件加入第一次加入时错过的那些 id

SELECT * 
FROM (
  SELECT * FROM table1
  WHERE user_id NOT IN (
    SELECT t1.user_id
    FROM table1 AS t1
    JOIN table2 AS t2
    ON t1.user_id = t2.user_id
  )
) t1
JOIN (
  SELECT * FROM table2
  WHERE user_id NOT IN (
    SELECT t1.user_id
    FROM table1 AS t1
    JOIN table2 AS t2
    ON t1.user_id = t2.user_id
  )
) t2
ON t1.email = t2.email

【讨论】：

谢谢。我真的希望它使用 COALESCE。根据链接，它似乎是可能的。但是，我猜 BigQuery 不支持？ ***.com/questions/5304184/…

per cloud.google.com/bigquery/query-reference#query-grammar : join_predicate: field_from_one_side_of_the_join = field_from_the_other_side_of_the_join [ AND ...]