SQL : 根据条件删除重复项

Question

我大多是 SQL 的新手，因此我对它提供的所有高级选项知之甚少。我目前使用 MS SQL Server 2016（开发版）。

我得到以下结果：

|  Type  |  Role  |                 GUID                 |
|--------|--------|--------------------------------------|
|   B    |   0    |                 ABC                  |
|   B    |   0    |                 KLM                  |
|   A    |   0    |                 CDE                  |
|   A    |   0    |                 EFG                  |
|   A    |   1    |                 CDE                  |
|   B    |   1    |                 ABC                  |
|   B    |   1    |                 GHI                  |
|   B    |   1    |                 IJK                  |
|   B    |   1    |                 KLM                  |

来自以下SELECT：

SELECT DISTINCT
        Type,
        Role,
        GUID

我正在计算遵循这些约束的 GUID :

-> 如果有多个行具有相同的 GUID，只计算 "Role" 设置为“1”的行，否则，计算 "Role" 设置为 0[= 的行26=] -> 如果只有一个，则根据自己的角色值将其计为 "Role 0" 或 "Role 1"。

我的objective是得到以下结果：

|  Type  |  Role  |              COUNT(GUID)             |
|--------|--------|--------------------------------------|
|   A    |   0    |                  1                   | => counted EFG as there was no other row with a "Role" set to 1
|   A    |   1    |                  1                   | => counted CDE with "Role" set to 1, but the row with "Role" set to 0 is ignored
|   B    |   1    |                  4                   |

Answer 1

您的查询没有实现您提到的逻辑。这是一个使用子查询和 window 函数的方法：

select type, role, count(*)
from (select t.*,
             count(*) over (partition by GUID) as guid_cnt
      from t
     ) t
where (guid_cnt > 1 and role = 1) or
      (guid_cnt = 1 and role = 0)
group by type, role;

子查询获取匹配 GUID 的行数。外部 where 然后根据您的条件使用它进行过滤。

注意：role 不是列名的好选择。它是一个关键字（参见 here) and may be reserved in the future (see here）。

Answer 2

A NOT EXISTS 可用于此。

例如：

declare @T table ([Type] char(1), [Role] int, [GUID] varchar(3));

insert into @T ([Type], [Role], [GUID]) values
('A',0,'CDE'),
('A',0,'EFG'),
('A',1,'CDE'),
('B',0,'ABC'),
('B',0,'KLM'),
('B',1,'ABC'),
('B',1,'GHI'),
('B',1,'IJK'),
('B',1,'KLM');

select [Type], [Role], COUNT(DISTINCT [GUID]) as TotalUniqueGuid
from @T t
where not exists (
  select 1 
  from @T t1
  where t.[Type] = t1.[Type]
    and t.[Role] = 0 and t1.[Role] > 0
    and t.[GUID] = t1.[GUID]
)
group by [Type], [Role];

Returns:

Type Role TotalUniqueGuid
A    0    1
A    1    1
B    1    4

SQL : 根据条件删除重复项

SQL : Remove duplicate based on a critera

sql

sql-server

duplicates

sql-server-2016