每个组的第二个最大值 - SQL

Question

我有这个数据集：

shopID    supplier    supply_count
1         a           12
2         b           12
2         f           12
2         e           10
3         b           5
3         a           2
4         f           15
4         c           11

我已使用此查询来获得每个组的第二高值：

select s1.shopId, max(s1.supply_count)
from supply s1
where supply_count NOT IN (
   select max(supply_count) 
   from supply s2
   where s1.shopId = s2.shopId
)
group by s1.shopId

我得到的结果是：

shopID   supply_count
2        10
3        2
4        11

期望的输出是： （如果没有第二高我想呈现最高的）

shopID   supply_count
1        12
2        10
3        2
4        11

基于这个问题： Retrieve 2nd highest count by each group

Answer 1

select 
shopID,
supply_count
from
(select shopID,
supply_count, dense_rank() over(partition by shopID order by supply_count desc) as rn
from supply) a
where rnk=2
group by
shopID,
supply_count
union
select shopID,
supply_count
from supply
where shopID in (select shopID from supply group by shopID having count(distinct supply_count) =1) a
group shopID,
supply_count;

我相信这应该可行。我在这里尝试的是通过为每个组分配一个排名并与只有一条记录的 shopID 进行联合来确定第二高的供应数量。

如果您不明白 dense_rank() 的作用，请阅读：https://towardsdatascience.com/how-to-use-sql-rank-and-dense-rank-functions-7c3ebf84b4e8

Answer 2

通过计算每个 ShopId 的行数，您可以获得所需的输出以包含每个 shopId 只有一行的值。

基于一些更多样化的数据和 cmets，我相信以下应该会给出预期的结果：

with c as (
    select *,
        case when dense_rank() over(partition by shopid order by supply_count desc)=2 
            or Min(supply_count) over(partition by shopid)
              =Max(supply_count) over(partition by shopid) 
        then 1 else 0 end as valid
    from t
)
select distinct shopId, supply_count
from c
where valid=1

例如DBFiddle with some more varied data