SQL 返回列中值发生变化的记录

Question

我在 Hive 表中有数据，看起来像这样 -

VIN	Mode	event	Start	End
ABC123456789	Mode 1	Deauthorized	01/01/2010 00:00:00	05/05/2014 14:54:54
ABC123456789	Mode 1	Deauthorized	05/05/2014 14:54:54	05/13/2014 19:09:51
ABC123456789	Mode 1	Deauthorized	05/13/2014 19:09:51	11/13/2014 22:26:32
ABC123456789	Mode 1	Authorized	11/13/2014 22:26:32	11/13/2014 22:31:00
ABC123456789	Mode 1	Authorized	11/13/2014 22:31:00	11/14/2014 01:23:56
ABC123456789	Mode 2	Deauthorized	11/14/2014 01:23:56	11/18/2014 19:38:51
ABC123456789	Mode 2	Deauthorized	11/18/2014 19:38:51	11/18/2014 19:38:54
ABC123456789	Mode 2	Deauthorized	11/18/2014 19:38:54	11/18/2014 20:07:52
ABC123456789	Mode 2	Authorized	11/18/2014 20:07:52	12/17/2014 19:22:50
ABC123456789	Mode 2	Authorized	12/17/2014 19:22:50	02/25/2015 20:03:44
ABC123456789	Mode 2	Authorized	02/25/2015 20:03:44	02/25/2015 20:03:48
ABC123456789	Mode 3	Authorized	02/25/2015 20:03:48	02/25/2015 20:14:05
ABC123456789	Mode 3	Deauthorized	02/25/2015 20:14:05	02/25/2015 20:14:29
ABC123456789	Mode 3	Deauthorized	02/25/2015 20:14:29	02/25/2015 20:40:21

我想获得一个汇总数据，其中事件列中的值与之前的值相比发生了变化。数据点按开始时间戳的升序排列。我尝试了窗口功能，但它对我不起作用。结果看起来类似于我在下表中显示的内容。您能为此提出任何优化的解决方案吗？

VIN	Mode	event	Start	End
ABC123456789	Mode 1	Deauthorized	01/01/2010 00:00:00	05/05/2014 14:54:54
ABC123456789	Mode 1	Authorized	11/13/2014 22:26:32	11/13/2014 22:31:00
ABC123456789	Mode 2	Deauthorized	11/14/2014 01:23:56	11/18/2014 19:38:51
ABC123456789	Mode 2	Authorized	11/18/2014 20:07:52	12/17/2014 19:22:50
ABC123456789	Mode 3	Deauthorized	02/25/2015 20:14:05	02/25/2015 20:14:29

Answer 1

你可以使用lag():

select t.*
from (select t.*, 
             lag(event) over (partition by vin order by start) as prev_event
      from t
     ) t
where prev_event is null or prev_event <> event;

这会查看时间和vin 的变化。我不确定mode 是否也相关。如果是这样，请将其添加到partition by。