BigQuery 更新在具有多行的表上合并

Posted 2023-03-25

【中文标题】BigQuery 更新在具有多行的表上合并

【英文标题】：BigQuery Update merge on table with multiple rows

【发布时间】：2019-12-17 13:46:05

【问题描述】：

我有以下几点：

Table A:

|uid|info|..
|123|null|..



Table B:
|uid|goodinfo|timestamp|
|123  |  3     |2019-12-12
|123  |  5     |2019-01-12
|234  |  11    |2019-10-12

当我尝试运行更新语句时，我总是收到“UPDATE/MERGE must match at most one source row for each target row”错误，因为在表 B 中我得到了多行并且我没有任何方法进行连接比这更具体。

我试过了：

UPDATE `Table A` a
SET info = (select goodinfo from `Table B` where uid=123
ORDER BY lastmodifieddate DESC
LIMIT 1) b
WHERE 
a.info IS NULL AND
a.user_id=123

-- 这种方法有效，但因为在子查询中我无权访问表 A，我无法将其概括为：

SET info = (select goodinfo from `Table B` where uid=a.uid
ORDER BY lastmodifieddate DESC
LIMIT 1) b

-- 这会抛出一个错误，说他不知道“a.uid”是谁

然后我尝试使用 BigQuery 中的合并：

MERGE `Table A` a 
USING (
  select goodinfo,uid from `Table B` 
  ORDER BY lastmodifieddate DESC
  LIMIT 1
) b
ON a.uid = b.uid 
WHEN MATCHED and a.info is null and DATE(a.timestamp) = "2019-12-12" THEN
  UPDATE SET a.info = b.goodinfo

-- 此查询实际上成功完成，但由于我尚未找到的原因未修改任何行

然后我试过了：

UPDATE `Table A` a 
SET a.info = b.goodinfo
FROM `Table B` b
WHERE a.uid = b.uid
and DATE(a.timestamp) = "2019-12-12"
and a.info IS NULL
//here I get the same error and I cannot filter the data from Table B and get the same error

关于以通用方式更新数据并以某种方式过滤表 B 中的数据并在加入时仅从 goodinfo 中获取值“3”的任何想法？

我也在考虑做一个：

WITH filtered_table_b(
  select uid, goodinfo from Table B
  ORDER BY lastmodifieddate DESC
  LIMIT 1
)

但这无济于事，因为我不知何故需要根据每个用户的时间戳选择最后一个好信息

谢谢

【参考方案1】：

这是您可以使用的标准 SQL：

WITH data AS (
select '123' as uid, 3 as goodinfo, DATE('2019-12-12') as timestamp union all
select '123' as uid, 5 as goodinfo, DATE('2019-01-12') as timestamp union all
select '234' as uid, 11 as goodinfo, DATE('2019-10-12') as timestamp 
),
filterData AS (
select uid, max(timestamp) maxTimestamp from data
group by uid
)

select data.uid, goodinfo, filterData.maxTimestamp as  maxTimestamp 
from data inner join filterData on data.uid = filterData.uid and data.timestamp = filterData.maxTimestamp

这是上面的输出：