使用 CTE 消除具有空值的行

Question

我有一些数据

RowIdentifier   ID  RowID   Position    Data    Rn
1             1      1       a1          A1      1
2             1      2       a2          A2      1
3             1      3       a3          NULL    1
4             1      4       a3          A3      2
5             1      1       b1          B1      1
6             1      2       b2          NULL    1
7             1      3       b2          B2      2
8             1      4       b3          B3      1

想要的输出是

ID  RowID  Position  Data
1     1     a1       A1
1     1     b1       B1
1     2     a2       A2
1     2     b2       B2
1     3     a3       A3
1     3     b3       B3

我需要消除那些位置重复且数据为空的行。即在示例中，在 RowIdentifier 3 和 4 中，Position 列中的值为 a3，但第三条 RowIdentifier 记录不会出现在最终输出中，因为它在 Data 列中为 null。

ddl如下

Declare @t table(RowIdentifier int identity,ID int,RowID int,Position varchar(10),Data varchar(10),Rn int)
    Insert into @t
    Select 1,1,'a1','A1',1 union all
    Select 1,2,'a2','A2',1 union all
    Select 1,3,'a3',null,1 union all
    Select 1,4,'a3','A3',2 union all
    Select 1,1,'b1','B1',1 union all
    Select 1,2,'b2',null,1 union all
    Select 1,3,'b2','B2',2 union all
    Select 1,4,'b3','B3',1 

    Select * from @t

我的方法如下

;with cte as(
Select ID,RowID,Position,Position as p2,Data,RowIdentifier from @t
union all
select c4.ID,c4.RowID,c4.Position,c5.Position , c4.Data,c4.RowIdentifier
from cte c5
join @t c4 on c4.Position = c5.Position
where c5.RowIdentifier < c4.RowIdentifier
 )
 ,
 cte2 as(
 select * , rn = Row_Number() over(PARTITION by position order by RowIdentifier)
 from cte where Data is not null)

select ID,RowID,Position,Data from cte2 where rn =1

但未按预期输出工作。我的输出是

ID  RowID   Position    Data
1   1   a1  A1
1   2   a2  A2
1   4   a3  A3
1   1   b1  B1
1   3   b2  B2
1   4   b3  B3

需要帮助

谢谢

Answer 1

试试这个代码

Select 
    ID,
    dense_rank() over(order by substring(data,2,len(data))*1) as rowid,
    position,
    data 
from 
    @t 
where 
    data is not null
group by 
    ID,RowID,position,data

Answer 2

过滤后的简单聚合

SELECT
   ID, MIN(RowID) AS RowID, POSITION, Data
FROM
   @t
WHERE
   Data IS NOT NULL
GROUP BY
   ID, RowID, POSITION, Data
ORDER BY
   POSITION, RowID

这会保留原始数据集中的 RowID，这样会更正确

3             1      3       a3          NULL    1
4             1      4       a3          A3      2   --take this row