google-bigquery - Bigquery - 'Having' 无法按预期工作

Question

我有一个疑问:

SELECT date(date) as day, App, count(*) as cnt,
avg(cnt) over (partition by App) as avg_per_day

FROM [DATABASE]

group by date,two_hour, App

现在我基本上只想只使用平均每天下载量超过 5 次的应用程序。但是，当我添加having avg_per_day>5时，我得到

Field 'avg_per_day' not found.

另外输入 where avg(cnt) over (partition by App) > 5 不起作用。

我做错了什么？

我不想使用双SELECT，因为每个查询都要花钱，而且它们应该每小时运行一次。

谢谢

Answer 1

OVER() 函数在 HAVING 步骤之后运行 - 因此在 HAVING 阶段，avg_per_day 列仍然不存在。

而不是:

SELECT date(date) as day, App, count(*) as cnt,
avg(cnt) over (partition by App) as avg_per_day
FROM [DATABASE]
group by date,two_hour, App
having avg_per_day>5

尝试:

SELECT day, App, cnt, avg_per_day
FROM (
SELECT date(date) as day, App, count(*) as cnt,
avg(cnt) over (partition by App) as avg_per_day
FROM [DATABASE]
group by date,two_hour, App
)
WHERE avg_per_day>5

成本方面应该是相同的。