基于一列的Hive自联接

Question

我在Hive中有一个表，数据来自SAP系统。该表包含以下列的列和数据：

+======================================================================+
|document_number | year | cost_centre | vendor_account_number | amount | 
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |                       |  123.5 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  25.96 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |                       |  586   |
+----------------------------------------------------------------------+

如上所示，vendor_account_number列的值仅存在于一行中，我想将其带到其余所有行上。

预期输出如下：

+======================================================================+
|document_number | year | cost_centre | vendor_account_number | amount | 
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  123.5 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  25.96 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  586   |
+----------------------------------------------------------------------+

为实现这一目标，我在Hive中写了以下CTE

with non_blank_account_no as(
  select document_number, vendor_account_number
  from my_table
  where vendor_account_number != ''
)

然后自我左外连接如下：

select 
    a.document_number, a.year, 
    a.cost_centre, a.amount,
    b.vendor_account_number
from my_table a
left outer join non_blank_account_no b on a.document_number = b.document_number
where a.document_number = ' '

但是我得到了重复的输出，如下所示

+======================================================================+
|document_number | year | cost_centre | vendor_account_number | amount | 
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  123.5 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  25.96 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  586   |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  123.5 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  25.96 |
+----------------------------------------------------------------------+
|       1        | 2016 |     XZ10    |      1234567890       |  586   |
+----------------------------------------------------------------------+

有人可以帮我理解我的Hive查询有什么问题吗？

Answer 1

在许多用例中，自联接可以由Windows函数替换

select  document_number
       ,year
       ,cost_center

       ,max (case when vendor_account_number <> '' then vendor_account_number end) over 
        (
            partition by    document_number
        )                                       as vendor_account_number

       ,amount

from    my_table