Question

我在包含多个日期的表格中重复了一行：

 ID      STATE       DATE
----------------------------
id01   connected  2015-04-04
id01   connected  2015-04-05
id01   connected  2015-04-08
id01   disconect  2015-04-11
id01   disconect  2015-04-12
id01   connected  2015-04-13

我想要一个带有“开始日期”和“结束日期”的查询，结果如下：

 ID      STATE    START DATE   END DATE
----------------------------------------
id01   connected  2015-04-04  2015-04-10
id01   disconect  2015-04-11  2015-04-12
id01   connected  2015-04-13  XXXXXXXXXX

最后一个“结束日期”并不重要（最后一个值，null，now（）...）

最重要的是检测更改日期（在此示例中，2015-04-10没有行，2015-04-13发生相同的状态）。

可行的解决方案？（无效）

SELECT ID, STATE, MIN(date), MAX(date) 
   FROM TABLE
   GROUP BY ID, STATE;

无效，因为合并间隔：

 ID      STATE    START DATE   END DATE
----------------------------------------
id01   connected  2015-04-04  XXXXXXXXXX
id01   disconect  2015-04-11  2015-04-12

查询已在Impala（类似SQL92）中运行

Answer 1

Impala支持窗口功能。这个问题是＆＃34;缺口和岛屿＆＃34;问题，所以可以使用行号的差异来解决：

select id, state, min(date) as start_date, max(date) as end_date
from (select t.*,
             row_number() over (partition by id order by date) as seqnum_id,
             row_number() over (partition by id, state order by date) as seqnum_isd
      from table t
     ) t
group by id, state, (seqnum_id - seqnum_isd);

差异的逻辑并不困难，但是当你第一次学习它时却很棘手。它有助于运行子查询并查看行号值是什么 - 以及差异定义每个组的原因。

Answer 2

（代表OP发布）。

从Gordon Linoff's answer，将“差距和岛屿”问题转化为我的研究案例，有解决方案：

select 
    id,
    state,
    start_date,
    date_add(lag(start_date, 1) over (partition by id order by start_date desc), -1) as end_date
from 
    (select id, state, min(date) as start_date, max(date) as end_date
        from (select t.*,
                row_number() over (partition by id order by date) as seqnum_id,
                row_number() over (partition by id, state order by date) as seqnum_isd
            from test t
        ) t
    group by id, state, (seqnum_id - seqnum_isd)) t_range
order by start_date;

SQL将日期缩小到“开始 - 结束”范围

2 个答案: