Select 项在每个组中按列不同

Question

我有这个样品table

+--------+-------------+
| DBName | Description |
+--------+-------------+
| A      | Car         |
| A      | Boat        |
| B      | Car         |
| B      | Plane       |
| C      | Car         |
| C      | Boat        |
| C      | Plane       |
+--------+-------------+

我只想获取每个 DBName 上不存在的描述，并显示没有描述的 DBName。

我想要的查询结果

+--------+-------------+
| DBName | Description |
+--------+-------------+
| A      | Plane       |
| B      | Boat        |
+--------+-------------+

请记住，在 DBName 上它不仅仅是 A、B、C。

Answer 1

使用NOT EXISTS

SELECT *
FROM   yourtable t
WHERE  NOT EXISTS
       (
           SELECT *
           FROM   yourtable x
           WHERE  x.Description = t.Description 
           AND    x.DBName     <> t.DBName 
       )

Answer 2

有趣的问题。这里有几个解决方案。围绕这些技术进行了讨论 here，并提出了一些处理此类情况的其他途径的建议。

SQL Fiddle Example

select DBName, Description
from (
    select DBName, Description
    from (select distinct DBName from demo) a
    cross join (select distinct Description from demo) b
) c
except
select DbName, Description from demo

此解决方案的工作原理是获取所有可能的组合（通过交叉连接每列的不同值），然后通过 except 子句排除所有已存在的组合。

SQL Fiddle Example

select [each].DBName, missing.Description
from (select distinct DBName from demo) [each]
cross join (select distinct Description from demo) [missing]
where not exists 
(
  select top 1 1
  from demo [exists]
  where [exists].DbName = [each].DBName
  and [exists].Description = [missing].Description
)

此解决方案与上述相同，只是我们使用 where not exists 代替 except 子句来删除现有组合。

Answer 3

理想情况下，您应该有一个数据主列表。如果你不这样做，你应该从数据中导出它，然后像下面这样对它们进行检查：

SQL Fiddle Example

select 
masterlistDbname.Dbname,
masterlistDesc.Description 
from
(
    select distinct Description from yourtable
) masterlistDesc
cross join
(
    select distinct Dbname from yourtable
) masterlistDbname
left join
yourtable t1 
on t1.Dbname = masterlistDbname.Dbname
and t1.Description = masterlistDesc.Description 
where t1.Dbname is NULL

Answer 4

你应该多扔一点样本数据。

试试这个，

create table #test(DBName varchar(50),Descriptions varchar(50) )

insert into #test VALUES
('A','Car')        
,('A','Boat')       
,('B','Car')        
,('B','Plane')      
,('C','Car')        
,('C','Boat')       
,('C','Plane')      

;

WITH CTE
AS (
    SELECT *
        ,ROW_NUMBER() OVER (
            ORDER BY (
                    SELECT NULL
                    )
            ) rn
        ,ROW_NUMBER() OVER (
            PARTITION BY DBName ORDER BY (
                    SELECT NULL
                    )
            ) rn1
    FROM #test
    )
SELECT t.DBName
    ,t1.Descriptions
FROM cte t
CROSS APPLY (
    SELECT TOP 1 Descriptions
    FROM cte t1
    WHERE t1.rn > t.rn
        AND t.Descriptions <> t1.Descriptions
        AND t.dbname <> t1.dbname
    ORDER BY t1.rn
    ) t1
WHERE t.rn1 = 1



drop table #test

Select 项在每个组中按列不同

Select item that is different in each of group by a column

sql

tsql

sql-server

sql-server-2008-r2