BigQuery连接广播限制错误

Question

我正在运行一个类似于这样的查询:

SELECT date, group, COUNT(a.user_id) as count
FROM (
SELECT user_id, DATE(timestamp) as date
FROM [db.log_2015_08]
GROUP EACH BY user_id, date) as a
JOIN EACH (
SELECT user_id, group
FROM [db.users]) as b
ON a.user_id = b.user_id
GROUP BY date, group

并得到以下错误:

Error: Shuffle reached broadcast limit for table __I0 (broadcasted at least 137619698 bytes). Consider using partitioned joins instead of broadcast joins .

我已经使用BigQuery有一段时间了，这对我来说是一个新的!这是一个相当大的连接，但我觉得我以前做过类似的没有这个错误。

Answer 1

尝试将GROUP EACH BY更改为GROUP BY。EACH关键字不再需要聚合，事实上，如果您删除它，查询规划器可以更聪明一点。