列中相同值的最小时间戳分区

Question

我有一个下面的数据集，我试图在其中获取单个列中相同值的最小时间戳。

这是我的数据集。

我正在尝试创建一个列来查找与经销商的每次互动 user_first_comment。

喜欢下面

是否有任何技术可用于为列中的相同值添加增量值。例如列 RANK_Incremental，这样我就可以使用 Min 和 hpartition 来获得最终输出。

Answer 1

使用常见的 table 表达式（或派生的 tables）和 window 函数支持，请尝试以下操作：

WITH step1 AS (
        SELECT *, CASE WHEN LAG(note_type) OVER (PARTITION BY incident ORDER BY note_date) = note_type THEN 0 ELSE 1 END AS edge FROM incidents
     )
   , step2 AS (
        SELECT *, SUM(edge) OVER (PARTITION BY incident ORDER BY note_date) AS grp FROM step1
     )
SELECT *, FIRST_VALUE(note_date) OVER (PARTITION BY incident, grp ORDER BY note_date) AS first_date FROM step2
;

结果：

+----------+---------------------+----------------+------+------+---------------------+
| incident | note_date           | note_type      | edge | grp  | first_date          |
+----------+---------------------+----------------+------+------+---------------------+
|  5498091 | 2021-12-15 17:20:00 | USER_COMMENT   |    1 |    1 | 2021-12-15 17:20:00 |
|  5498091 | 2021-12-15 17:21:00 | USER_COMMENT   |    0 |    1 | 2021-12-15 17:20:00 |
|  5498091 | 2021-12-15 17:55:00 | DEALER_COMMENT |    1 |    2 | 2021-12-15 17:55:00 |
|  5498091 | 2021-12-15 17:59:00 | USER_COMMENT   |    1 |    3 | 2021-12-15 17:59:00 |
|  5498091 | 2021-12-16 11:02:00 | USER_COMMENT   |    0 |    3 | 2021-12-15 17:59:00 |
|  5498091 | 2021-12-16 16:46:00 | DEALER_COMMENT |    1 |    4 | 2021-12-16 16:46:00 |
+----------+---------------------+----------------+------+------+---------------------+

设置：

CREATE TABLE incidents (
   incident         int
 , note_date        datetime
 , note_type        varchar(20)
)
;

INSERT INTO incidents VALUES
   (5498091, '2021-12-15 17:20', 'USER_COMMENT')
 , (5498091, '2021-12-15 17:21', 'USER_COMMENT')
 , (5498091, '2021-12-15 17:55', 'DEALER_COMMENT')
 , (5498091, '2021-12-15 17:59', 'USER_COMMENT')
 , (5498091, '2021-12-16 11:02', 'USER_COMMENT')
 , (5498091, '2021-12-16 16:46', 'DEALER_COMMENT')
;

列中相同值的最小时间戳分区

Min time stamp partition by same value in column

sql

hadoop

hive

hiveql

impala