#成为 SQL 大师#groupby 中不能有聚合函数

背景

select 
    t1.p_date,
    t1.member_id,
    top3_domains,
    CASE WHEN label_code = 'realtime' THEN COLLECT_SET(bayes_first) END as realtime_bys
    from 
    create_top3 t1 
    join consume_profile t2 
    on t1.p_date=t2.p_date 
    and cast(t1.member_id as string)=cast(t2.member_id as string)
    group by 1,2,3,4


报错：
aggregate functions are not allowed in GROUP BY, but found CASE WHEN (t2.`label_code` = 'realtime') THEN collect_set(t2.`bayes_first`) END;
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

原因&解决方案

你遇到的错误是因为在 GROUP BY 子句中，不能包含聚合函数（如 COLLECT_SET）之外的其他列或表达式。在这种情况下，你可以使用子查询来解决问题。以下是一种修改方法：

SELECT 
    t1.p_date,
    t1.member_id,
    top3_domains,
    MAX(CASE WHEN t2.label_code = 'realtime' THEN t2.collect_set_bayes_first END) AS realtime_bys
FROM 
    create_top3 t1 
JOIN 
    (
        SELECT
            p_date,
            member_id,
            label_code,
            COLLECT_SET(bayes_first) AS collect_set_bayes_first
        FROM 
            consume_profile
        WHERE 
            label_code IN ('text_bayes_first_realtime', 'text_bayes_first_short_term', 'text_bayes_first_long_term')
        GROUP BY 
            p_date,
            member_id,
            label_code
    ) t2
ON 
    t1.p_date = t2.p_date 
    AND cast(t1.member_id as string) = cast(t2.member_id as string)
GROUP BY 
    1, 2, 3;

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

声明：本文章有 AI 辅助