Teradata NORMALIZE选项

问题描述

我一直在使用td_normalize_overlap_meet缩短期间。我在论坛上看到过一些示例，这些示例使用CNT来标识已崩溃的时期数，并且我一直在寻找文档，以查看是否存在其他类似的功能但没有找到任何东西。我正在特别寻找一种可以保留上一个崩溃期间的开始日期的工具。例如，假设我有以下时期：

+-----------+------------+
|Start Date | End Date   |
+-----------+------------+
|2018-01-02 | 2018-01-04 |
|2018-01-05 | 2018-01-07 |
|2018-01-08 | 2018-01-10 |
+-----------+------------+

然后我将它们折叠成这个：

+-----------+------------+-----+
|Start Date | End Date   | CNT |
+-----------+------------+-----+
|2018-01-02 | 2018-01-10 | 3   |
+-----------+------------+-----+

有没有类似于CNT的功能可以给我呢？

+-----------+------------+-----+------------------------------------+
|Start Date | End Date   | CNT | Last Collapsed Period's Start Date |
+-----------+------------+-----+------------------------------------+
|2018-01-02 | 2018-01-10 | 3   | 2018-01-08                         |
+-----------+------------+-----+------------------------------------+

解决方法

SELECT NORMALIZE是一种非常简单的语法（特别是与那些td_normalize...函数相比），但是不能用于获取行数或 last行的开始日期。

获得期望结果的最简单方法是使用nPath表运算符。假设要归一化的行不止一组：

WITH cte AS 
 ( -- the base Select creating the not yet normalized rows
   SELECT *
   FROM mytab
 )
SELECT * 
FROM 
   NPath(ON cte
         PARTITION BY col -- grouping column(s)
         ORDER BY Start_date
         USING
           MODE (NonOverlapping)
           Symbols (start_date-1 > lag(end_date,1,date '0001-01-01') AS newgrp,-- starting row of a group of overlapping rows
                    start_date-1 <= lag(end_date,) as x)                      -- overlapping row
           Pattern ('newgrp.x*')                                                  -- start plus overlapping rows
           RESULT(First (col OF newgrp) AS col,-- grouping column(s)
                  first (start_date OF ANY(newgrp,x)) AS start_date,-- start date of group 
                  last  (end_date   OF ANY(newgrp,x)) AS end_date,-- end date of group
                  Count (*          OF ANY(newgrp,x)) AS Cnt,-- number of rows in group
                  last  (start_date OF ANY(newgrp,x)) AS last_start              -- start date of last row in group
                 )
        );