从所有行中选择时如何在数据库中选择每个单独记录的 6 个顶部记录

Question

假设我有下表

CREATE TABLE #tblUsersPokemons (
    RecordId int NOT NULL,
    PokemonId int NOT NULL,
    PokemonExp int NOT NULL,
    PokemonLevel int NOT NULL,
    UserId int NOT NULL
)

现在下面的查询可以正常工作了

select 
  SUM(cast(PokemonExp as bigint)) as TotalExp,
  MAX(PokemonLevel) as MaxPokeLevel,
  Count(PokemonId) as TotalPoke,
  UserId 
from #tblUsersPokemons 
group by UserId

这里是此类查询的示例结果

ToplamExp   MaxPokeLevel    TotalPoke   UserId
----------- --------------- ----------- --------
29372294    101             4           1
1134696     98              1           2
1400        98              1           101
24534365    98              4           102
1400        98              1           1102
1400        98              1           1103
1400        98              1           2102
1400        98              1           2103
789220      98              7           2105
1468        98              1           3104

现在我的问题来了

我想限制计数的 PokemonId。我的意思是我想选择最多 6 个相同的 PokemonId 记录。从这些记录中，PokemonExp 排序的前 6 名应该被计算在内。

例如用户有以下记录

从这个表中查询应该采用记录 id : 1,2,3,4,5,6,9 而不是 7,8，因为 PokemonId 1 的前 6 条记录被采用

Answer 1

如果我理解正确，您希望每个用户的前 6 行聚合。您可以使用row_number() 轻松做到这一点：

select SUM(cast(PokemonExp as bigint)) as ToplamExp,
       MAX(PokemonLevel) as MaxPokeLevel,
       Count(PokemonId) as TotalPoke,UserId 
from (select p.*,
             row_number() over (partition by userid order by pokemanexp desc) as seqnum
      from tblUsersPokemons p
     ) p
where seqnum <= 6
group by UserId;

编辑：

我想你想在partition by 子句中包含PokemonId：

select SUM(cast(PokemonExp as bigint)) as ToplamExp,
       MAX(PokemonLevel) as MaxPokeLevel,
       Count(PokemonId) as TotalPoke,UserId 
from (select p.*,
             row_number() over (partition by userid, PokemonId
                                order by pokemanexp desc) as seqnum
      from tblUsersPokemons p
     ) p
where seqnum <= 6
group by UserId;