Question

我尝试编写查询以根据一个字段选择连续块，将其转换为状态更改记录。基本上我只需要选择一个＆＃34; v1＆＃34;如果在基于id排序时多次出现相同的值。让我们假设下表：

| id | v1 | v2 | v3 |
---------------------
|1   | a  |  b | b  |
|2   | a  |  e | b  |
|3   | z  |  b | b  |
|4   | z  |  o | p  |
|5   | a  |  h | t  |
|6   | a  |  i | k  |
|7   | b  |  g | p  |

期望的结果：

|time sequence    | state  |
----------------------------
|1                | a      |
|2                | z      |
|3                | a      |
|4                | b      |

任何人都有什么想法？

Answer 1

在这种情况下使用window function，lag()。

此查询查找系列的开头：

select id, v1, v1 is distinct from lag(v1) over (order by id) as start
from my_table

 id | v1 | start 
----+----+-------
  1 | a  | t
  2 | a  | f
  3 | z  | t
  4 | z  | f
  5 | a  | t
  6 | a  | f
  7 | b  | t
(7 rows)

仅选择start = true行并添加行号：

select row_number() over (order by id) as sequence, v1 as state
from (
    select id, v1, v1 is distinct from lag(v1) over (order by id) as start
    from my_table
    ) s
where start

 sequence | state 
----------+-------
        1 | a
        2 | z
        3 | a
        4 | b
(4 rows)

从表格数据中提取顺序状态变化

1 个答案: