2

我有一个包含表单事务数据的配置单元表(高度简化)

id  Status  Value   UpdatedTimeStamp

515 Open    1   2014-05-08T11:43:27
516 Open    NULL    2014-05-08T11:43:27
515 Answered    1   2014-05-09T11:43:27 
515 Closed  NULL    2014-05-10T11:43:27
517 Open    NULL    2014-05-09T11:43:27
516 Closed  NULL    2014-05-09T11:43:27

对于我的用例,我需要一个包含唯一 id 的输出表,其中列条目作为该列中该特定 id 按时间戳排序时的最后一个非 NULL 值。

因此,所需的输出是:

id  Status  Value   UpdatedTimeStamp

515 Closed  1   2014-05-10T11:43:27
516 Closed  NULL    2014-05-09T11:43:27
517 Open    NULL    2014-05-09T11:43:27

我试图通过在分区上使用窗口函数来实现这一点

INSERT OVERWRITE TABLE testSample2 SELECT id, FIRST_VALUE (Status) OVER SortedData, FIRST_VALUE (Value IGNORE NULLS) OVER SortedData, FIRST_VALUE (UpdatedTimeStamp) OVER SortedData FROM testSample WINDOW SortedData (PARTITION BY id ORDER BY UpdatedTimeStamp DESC ROWS BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING);

这给了我 NullPointerException(Hive 中可能还不支持 IGNORE NULLS)。请建议我如何取最后一个非 NULL 值。

4

1 回答 1

1

尝试使用这样的东西:

select ss.vl1, FIRST_VALUE(ss.ky) OVER (PARTITION BY ss.vl1 ORDER BY ss.nl, ss.vl rows unbounded preceding) 
FROM (
    SELECT IF(s.vl=123456, 'sasa', 'AAAAAA') as vl1, s.vl, s.ky, IF(s.ky IS NULL, 1, 0) as nl
    FROM (
        select explode(map(NULL, 123, 1, 1234, 2, 12345, 3, 123456)) as (ky, vl) from default.dual) s ) ss;
vl1     _wcol0
AAAAAA  1
AAAAAA  1
AAAAAA  1
sasa    3
于 2015-03-11T12:44:42.003 回答