使用max和分组依据后获取多行

编程问答 2022-06-10

问题描述

我正在尝试根据xml_id具有最活跃属性的哪个段来为xml_id分配一个段。我对此进行了查询：

我正在创建一个带有如下选择的表：

create table schema.table4 as
select
    yyyy_mm_dd,xml_id,segment as xml_segment,max(property_count)
from(
    select
        t1.yyyy_mm_dd,t2.xml_id,t3.segment,count(t1.hotel_id) as property_count
    from(
        select
            yyyy_mm_dd,hotel_id
        from
            schema.table1
        where
            is_active = 1
            and yyyy_mm_dd = "2020-10-01"
    ) t1
    left join(
        select
            yyyy_mm_dd,hotel_id,xml_id
        from
            schema.table2
        where
            yyyy_mm_dd = "2020-10-01"
            and xml_id is not null
    ) t2 on t2.hotel_id = t1.hotel_id and t2.yyyy_mm_dd = t1.yyyy_mm_dd
    inner join
        schema.table3 t3 on t3.hotel_id = t1.hotel_id
    group by
        1,2,3
) x
group by
    1,3

但是，当我查询该表时，每个xml_id可以看到多个细分。我希望只插入带有MAX() property_count的段。怎么会这样呢？

select
    xml_id,count(*)
from
    schema.table4

显示多行，其中xml_id具有两行和三行。我只需要一行，该段应该是property_count最高的那一行。

下面的一些示例输出。 xml_id继承其属性最多的段。

t1：

| yyyy_mm_dd | hotel_id | is_active |
|------------|----------|-----------|
| 2020-10-01 | 1        | 1         |
| 2020-10-01 | 2        | 1         |
| 2020-10-01 | 3        | 1         |
| 2020-10-01 | 4        | 1         |
| 2020-10-01 | 5        | 1         |
| 2020-10-01 | 6        | 1         |
| 2020-10-01 | 7        | 0         |

t2：

| yyyy_mm_dd | hotel_id | xml_id |
|------------|----------|--------|
| 2020-10-01 | 1        | 444    |
| 2020-10-01 | 2        | 444    |
| 2020-10-01 | 3        | 444    |
| 2020-10-01 | 4        | 920    |
| 2020-10-01 | 5        | 920    |
| 2020-10-01 | 6        | 920    |
| 2020-10-01 | 7        | null   |

t3：

| hotel_id | segment |
|----------|---------|
| 1        | Home    |
| 2        | Core    |
| 3        | Core    |
| 4        | Core    |
| 5        | Home    |
| 6        | Home    |
| 7        | Chain   |

预期输出：

| yyyy_mm_dd | xml_id | segment |
|------------|--------|---------|
| 2020-10-01 | 444    | Core    |
| 2020-10-01 | 920    | Home    |

解决方法

如果我的理解正确，您想按日期xml_id和segment来统计活动数量。这是该计数的基本查询：

select t2.yyyy_mm_dd,t2.xml_id,t3.segment,count(*) as num_actives
from t2 join
     t1
     on t2.hotel_id = t1.hotel_id and
        t2.yyyy_mm_dd  = t1.yyyy_mm_dd join
     t3
     on t3.hotel_id = t2.hotel_id
where t1.is_active = 1
group by t2.yyyy_mm_dd,t3.segment;

要获得最大值，可以使用窗口函数：

select t.*
from (select t2.yyyy_mm_dd,count(*) as num_actives,row_number() over (partition by t2.yyyy_mm_dd,t2.xml_id order by count(*) desc) as seqnum
      from t2 join
           t1
           on t2.hotel_id = t1.hotel_id and
              t2.yyyy_mm_dd  = t1.yyyy_mm_dd join
           t3
           on t3.hotel_id = t2.hotel_id
      where t1.is_active = 1
      group by t2.yyyy_mm_dd,t3.segment
     ) t
where seqnum = 1;
hivehiveqlsqlsql