Question

我在蜂巢中有两个表，如下所述在Hive中

表1：

<script src="/socket.io/socket.io.js"></script>
<div class="col">
  <h1>Sn00p Monitor Dashboard</h1>
  <ul id="messages"></ul>
</div>

表2：

id   name value
1    abc  stack
3    abc  overflow
4    abc  foo
6    abc  bar

我需要计算值列而不考虑id和name列。

预期输出为

id   name value       
5    xyz  overflow       
9    xyz  stackoverflow 
3    xyz  foo
23   xyz  bar

我试过了，可以在其他数据库中工作，但不能在蜂巢中工作

id name value
1  abc  stack
9  xyz  stackoverflow

Hive希望使用如下所述的group by子句。

select id,name,value from
 (SELECT id,name,value FROM table1  
   UNION ALL 
  SELECT id,name,value FROM table2) t 
 group by value having count(value) = 1;

并给出输出

select id,name,value from
  (SELECT id,name,value FROM table1  
    UNION ALL 
  SELECT id,name,value FROM table2) t 
 group by id,name,value having count(value) = 1;

我们将必须在select子句中使用分组的所有列。但是当我给它的时候考虑了所有的列，结果与预期的不同。

Answer 1

计算解析count(*) over(partition by value)。用数据示例进行测试：

with 

table1 as (
select stack (4,
              1,'abc','stack',
              3,'abc','overflow',
              4,'abc','foo',
              6,'abc','bar'
             ) as (id, name, value)
),

table2 as (
select stack (4,
              5,  'xyz','overflow',      
              9,  'xyz','stackoverflow',
              3,  'xyz','foo',
              23, 'xyz','bar'
             ) as (id, name, value)
)

select id, name, value
from(
select id, name, value, count(*) over(partition by value) value_cnt
 from
(SELECT id,name,value FROM table1  
  UNION ALL 
 SELECT id,name,value FROM table2) s
)s where value_cnt=1;

结果：

OK
id      name    value
1       abc     stack
9       xyz     stackoverflow
Time taken: 55.423 seconds, Fetched: 2 row(s)

Answer 2

您可以在下面尝试-

seELECT id,name,value FROM table1 a left join table2 b on a.value=b.value
where b.value is null
UNION ALL SELECT 
seELECT id,name,value FROM table2 a left join table1 b on a.value=b.value
where b.value is null

蜂巢：根据特定列中的唯一值填充其他列

2 个答案: