是否可以使用 PARTITION BY 进行计数？

Question

我在 BigQuery 中有一个数据库，其中每条记录都是我网站上的网络流量会话。

我目前有这张表可以告诉我一个人是如何访问我的网站的，还有一个列可以告诉我每个事件的顺序。

最终目标是查看一个人在自然会话之前必须进行多少次“非自然”会话。

我正在尝试创建一个额外的列，每次发生转化时都会返回“包含”（当在非自然会话之后发生自然会话时）

我知道如何使用 Excel 执行此操作，但我不知道如何使用 SQL 执行此操作。我有一种感觉“PARTITION BY”是解决方案，但我不知道如何。

这是我的 Excel 解决方案：

=IF(AND((COUNTIF($B$2:B2,FALSE))>=1,(IF(COUNTIF($B$2:B2,FALSE)>=1,COUNTIFS($B$2:B2,TRUE,$C$2:C2,">1"),0))>=1),"include","exclude")

Answer 1

如果您尝试添加 status 标志，那么您似乎需要以下规则：

• “排除”所有无机会话。
• “包括”在第一个无机会话之前的所有有机会话。
• “包括”所有其他自然会话。

如果是这样，您可以简单地使用：

select t.*
       (case when not is_organic then 'exclude'
             when countif(not is_organic) over (partition by partition by person_id order by sequence) = 0
             then 'exclude'
             else 'include'
        end) as status
from t;

但是，对于这个问题：

最终目标是查看一个人在自然会话之前必须进行多少次“非自然”会话。

我会简单地将聚合与窗口函数一起使用：

select person_id, countif(sequence < first_organic_sequence)
from (select t.*,
             min(case when is_organic then sequence end) over (partition by person_id) as first_organic_sequence
      from t
     ) t
group by person_id;

或者如果sequence 总是以1 开头并且怎么没有间隔：

select person_id, min(case when is_organic then sequence end) - 1
from t
group by person_id;

Answer 2

请看以下内容。我创建了一个中间 CTE 来定义之前的记录是什么，然后使用它来确定第一个有机的会话数以及您的状态。

with sample as (
    SELECT 1 as person_id, FALSE as is_organic, 1 as sequence
    UNION ALL
    SELECT 1 as person_id, TRUE as is_organic, 2 as sequence
    UNION ALL 
    SELECT 1 as person_id, TRUE as is_organic, 3 as sequence
    UNION ALL  
    SELECT 2 as person_id, TRUE as is_organic, 1 as sequence
    UNION ALL
    SELECT 2 as person_id, FALSE as is_organic, 2 as sequence
),
modified  as (
    select *, 
        lag(is_organic)
            over(partition by person_id order by sequence ) as previous_indicator,
    from sample
)
select *,
    case when previous_indicator is not true then count(person_id) over(partition by person_id order by sequence) end as sessions_before_organic, 
    case when previous_indicator IS NOT NULL and is_organic = TRUE THEN 'include'
        else 'exlcude' 
    end as status
from modified;

Answer 3

带有 Partition By 子句的 SQL Count 是 t-SQL 开发人员可以轻松使用的新的强大语法之一。您可以根据此链接一起使用这两个命令： https://www.kodyaz.com/t-sql/sql-count-function-with-partition-by-clause.aspx