hive row_number()

问题背景：想对字段A分组，然后根据字段B排序

最笨的方法：

SELECT b.*
FROM (
  SELECT 
    A
  , MAX(B) AS B
  FROM mytable
  WHERE 
     SOME CONDITIONS
  GROUP BY  A
) a
JOIN (
  SELECT 
      A
    , C
    , B
  FROM mytable
  WHERE 
      SOME CONDITIONS
    GROUP BY A, B, C
)b 
ON a.B = b.B AND a.A = b.A
;

后来学会了row_num(),是hive自带的函数，它的作用是按指定的列进行分组生成行序列。在ROW_NUMBER(a,b) 时，若两条记录的a，b列相同，则行序列+1，否则重新计数。

一开始还是写出来很笨很丑的代码：

SELECT *, rank
FROM(
  SELECT 
    *
  , row_number() over (partition by A order by B desc ) rank
  FROM (
    SELECT 
        A
      , C
      , B
    FROM mytable
    WHERE 
        SOME CONDITIONS
    GROUP BY 
      A
    , C
    , B
  ) tb_1
) tb_2
WHERE rank < 2

其实正确的应该是如下方法：

select * from(
 	SELECT  
              A
            , C
            , B
            , row_number() over (partition by A order by B desc ) rank
    FROM mytable
    WHERE 
        SOME CONDITION
 ) a where a.rank=1