Question

我有2个表格产品和商品，表格中的数据如下所示

select * from psales;
+-------------+---------------+--+
| psales.pid  | psales.sales  |
+-------------+---------------+--+
| 1           | 100           |
| 1           | 150           |
| 1           | 200           |
| 2           | 75            |
| 2           | 45            |
| 2           | 145           |
| 3           | 176           |
| 3           | 99            |
| 1           | 27            |
| 4           | 51            |
+-------------+---------------+--+

select * from product;
+--------------+----------------+--+
| product.pid  | product.pname  |
+--------------+----------------+--+
| 1            | p1             |
| 2            | p2             |
| 3            | p3             |
| 4            | p4             |
+--------------+----------------+--+

目标是获得组合销售额第二高的产品。

这是我当前用于获取组合销售额最高（正常工作）的产品的查询

select p1.pname, p1.total_sales
from (select p.pid as pid, p.pname as pname, s.sales as sales,  
      sum(s.sales) over (partition by p.pid order by p.pid) as total_sales 
      from product p
      inner join psales s on (p.pid = s.pid) 
      order by total_sales desc) p1 
limit 1;

我如何获得销售总额第二高的产品？

当我尝试在内部查询中获取row_num时，出现以下错误：

select p1.pname as pname, p1.total_sales as total_sales, row_num() over (partition by pname order by pname) as rownum 
from (select p.pid as pid, p.pname as pname, s.sales as sales,  
      sum(s.sales) over (partition by p.pid order by p.pid) as total_sales, 
      row_num() over (partition by p.pid) as rownum 
      from product p 
      inner join psales s on (p.pid = s.pid) 
      order by total_sales desc) p1 
where rownum =2;

错误：编译语句时出错：失败：SemanticException无法将窗口调用分解为组。至少一组必须仅取决于输入列。还要检查循环依赖性。潜在错误：无效的函数row_num（状态= 42000，代码= 40000）

谢谢您的帮助。

Answer 1

在上部子查询中使用row_number()函数。看来您不需要解析sum()，简单的group by就可以做到：

select p1.pname, p1.pid, p1.total_sales
  from
(
select p1.pname, p1.pid, p1.total_sales, 
       row_number() over (order by total_sales  desc) rn
  from 
     (select p.pid, p.pname, sum(s.sales) as total_sales 
        from product p 
             inner join psales s on p.pid = s.pid
        group by p.pid, p.pname
     )p1
)s
where rn=2
;

如果要选择所有具有相同销售额的产品，请使用dense_rank()代替row_number()。

Answer 2

您可以使用dense_rank对每个pid名称组合的总和进行排名。

select p1.pname,p1.pid,p1.total_sales
from (select p.pid, p.pname,sum(s.sales) as total_sales,
      dense_rank() over(order by sum(s.sales) desc) as rnk 
      from product p 
      join psales s on p.pid = s.pid
      group by p.pid,p.pname
     ) p1
where rnk=2

蜂巢SQL内部联接-如何在同一查询中获取sum和row_num

2 个答案: