Question

我有两个数据集：

id_description

id  description
1   The cat sat
2   The dog barked
2   The dog barked
3   The parrot
4   The dog barked
4   The dog barked

person_description

person  description
John    The cat sat
Jane    The dog barked
James   The parrot
Mary    The dog barked

我需要构造一个看起来像这样的第三个数据集（以下两个选项之一）：

id  person  description
1   John    The cat sat
2   Jane    The dog barked
3   James   The parrot
4   Mary    The dog barked

id  person  description
1   John    The cat sat
2   Mary    The dog barked
3   James   The parrot
4   Jane    The dog barked

我首先尝试：

SELECT distinct a.id, b.person, a.description
FROM id_description a
LEFT OUTER JOIN person_description b ON a.description = b.description

这将产生此数据集：

id  person  description
1   John    The cat sat
2   Jane    The dog barked
2   Mary    The dog barked
3   James   The parrot
4   Jane    The dog barked
4   Mary    The dog barked

由于在description上的连接，person可以复制两个或更多 id个数字。我将如何到达目标数据集？

person和id的每个数字都代表一次，而id附加到哪个person则无关紧要（即2 / Jane和4 / Mary是相当于2 / Mary和4 / Jane）。我尝试使用row_number() over (partition by id order by person)，然后通过row_number = 1进行过滤，但结果如下：

id  person  description
1   John    The cat sat
2   Jane    The dog barked
3   James   The parrot
4   Jane    The dog barked

玛丽没有代表，因为简是2和4的第一行。

Answer 1

如果您是由于Description上的联接而到达这里的，那么解决此问题的一种方法是获取两个row_numbers（）。

两个分区都在描述上。

一个按ID排序，另一个按Person排序。

然后选择ID与Row_Numbers匹配的个人的描述。

以伪形式，应该看起来像这样：

with cte_ID AS (SELECT Description, ID, {RowNumber ordered by ID} AS ID_RN...)
,    cte_Person AS (SELECT Description, Person, {RowNumber ordered by Person} AS Person_RN...)
SELECT ID, Person, Description
FROM cte_ID JOIN cte_Person ON Description=Description AND ID_RN=Person_RN

您可以通过在原始联接中添加row_numbers（按Description进行分区）到每个原始表中，然后在Description和Row_Number上联接来完成此操作。

Answer 2

使用row_number窗口功能

    with t as
(
select 1 as  id,'John' as person,'The cat sat' as des
union all
select 2 as  id,'Jane' as person,'The dog barked' 
union all
select 2 as  id,'Mary' ,'The dog barked' 
union all
select 3 , 'James' ,'The parrot'  
union all
select 4 , 'Jane' ,'The dog barked'  
union all
select 4 , 'Mary' ,'The dog barked'  

) select * from
(
select *,row_number() over(partition by des  order by id ) rn 
 from t
) as t1 where t1.rn=1
union
select * from
(
select *,row_number() over(partition by des  order by person ) rn 
 from t
) as t1 where t1.rn=1

id  person  des             rn
1   John    The cat sat     1
2   Mary    The dog barked  1
3   James   The parrot      1
4   Jane    The dog barked  1

T-SQL从包含一对值的列表中获取一个

2 个答案: