选择最常见的值以及分组依据

Question

我有以下 SQL 表

eventdate   userid  traffic location    
18.09.2023  user_1  10      A
18.09.2023  user_1  20      A
18.09.2023  user_2  10      B
18.09.2023  user_2  20      B
18.09.2023  user_2  30      B
18.09.2023  user_3  100     A
19.09.2023  user_1  50      B
19.09.2023  user_2  10      B
19.09.2023  user_2  20      B
19.09.2023  user_3  150     C
19.09.2023  user_3  250     C
20.09.2023  user_1  50      A
20.09.2023  user_1  20      A
20.09.2023  user_2  30      B
20.09.2023  user_3  110     C
20.09.2023  user_3  120     C

我想要的结果如下 eventdate - 每周开始，userid - 每周唯一用户 ID，流量 - 所有流量的总和，位置 - 本周出现最频繁的位置

例如

    eventdate   userid  traffic location
    18.09.2023  user_1  150    A
    18.09.2023  user_2  120    B
    18.09.2023  user_3  730    C

我能够通过以下查询获得结果

SELECT t1.eventdate, t1.userid, t1.traffic, t2.location
  FROM (SELECT TO_CHAR(TRUNC(TO_DATE('2023-09-18', 'yyyy-mm-dd'), 'IW'),
                       'yyyy-mm-dd') AS eventdate,
               tk.userid,
               SUM(tk.traffic) AS traffic
          FROM test_kt tk
         GROUP BY tk.userid) t1
  JOIN (
         WITH cte AS 
        (
         SELECT tk2.userid,
                tk2.location,
                ROW_NUMBER() OVER 
                (PARTITION BY tk2.userid ORDER BY COUNT(tk2.location) DESC) rn
           FROM test_kt tk2
          GROUP BY tk2.userid, tk2.location
        )
        SELECT userid, location 
          FROM cte 
         WHERE rn = 1
       ) t2 
    ON t1.userid = t2.userid;

有没有有效的方法可以做到这一点？

Answer 1

您可以使用聚合函数实现相同的目的STATS_MODE：

with a(eventdate, userid, traffic, location) as (
  select to_date('18.09.2023', 'dd.mm.yyyy'), 'user_1', 10,  'A' from dual union all
  select to_date('18.09.2023', 'dd.mm.yyyy'), 'user_1', 20,  'A' from dual union all
  select to_date('18.09.2023', 'dd.mm.yyyy'), 'user_2', 10,  'B' from dual union all
  select to_date('18.09.2023', 'dd.mm.yyyy'), 'user_2', 20,  'B' from dual union all
  select to_date('18.09.2023', 'dd.mm.yyyy'), 'user_2', 30,  'B' from dual union all
  select to_date('18.09.2023', 'dd.mm.yyyy'), 'user_3', 100, 'A' from dual union all
  select to_date('19.09.2023', 'dd.mm.yyyy'), 'user_1', 50,  'B' from dual union all
  select to_date('19.09.2023', 'dd.mm.yyyy'), 'user_2', 10,  'B' from dual union all
  select to_date('19.09.2023', 'dd.mm.yyyy'), 'user_2', 20,  'B' from dual union all
  select to_date('19.09.2023', 'dd.mm.yyyy'), 'user_3', 150, 'C' from dual union all
  select to_date('19.09.2023', 'dd.mm.yyyy'), 'user_3', 250, 'C' from dual union all
  select to_date('20.09.2023', 'dd.mm.yyyy'), 'user_1', 50,  'A' from dual union all
  select to_date('20.09.2023', 'dd.mm.yyyy'), 'user_1', 20,  'A' from dual union all
  select to_date('20.09.2023', 'dd.mm.yyyy'), 'user_2', 30,  'B' from dual union all
  select to_date('20.09.2023', 'dd.mm.yyyy'), 'user_3', 110, 'C' from dual union all
  select to_date('20.09.2023', 'dd.mm.yyyy'), 'user_3', 120, 'C' from dual
)
select
  trunc(eventdate, 'iw') as eventdate,
  userid,
  sum(traffic) as traffic,
  stats_mode(location) as location
from a
group by
  trunc(eventdate, 'iw'),
  userid

活动日期	用户身份	交通	地点
2023-09-18	用户_1	150	A
2023-09-18	用户_2	120	乙
2023-09-18	用户_3	第730章	C

小提琴

UPD：您还可以查阅此函数的文档，其中显示了示例查询以获取多种模式（如果存在）（您将DENSE_RANK在原始查询中使用它来代替ROW_NUMBER）。