sql窗口函数检测列值的变化答案

【问题标题】：sql window function to detect change in column valuessql窗口函数检测列值的变化
【发布时间】：2016-08-20 00:31:46
【问题描述】：

我想检测此（示例）db 中列值的变化

    WITH events(id, row,event) AS (
    VALUES
    (1,1, 0 )
   ,(1,2, 0 ) 
   ,(1,3, 1 )
   ,(1,4, 0 )
   ,(1,5, 1 )
   ,(2,1, 0 )
   ,(2,2, 1 )
   ,(3,1, 0 )
   ,(3,2, 0 )
   )
   select * from events

我正在寻找的是新列“代码”的代码，该列在之后切换为 1 de 事件列显示 1。在同一 id 内，代码保持 1。对于这个例子，这个新列将如下所示

    WITH events2(id, row,event, code) AS (
    VALUES
    (1,1, 0, 0 )
   ,(1,2, 0, 0 ) 
   ,(1,3, 1, 0 )
   ,(1,4, 0, 1 ) -- notice the switch here
   ,(1,5, 1, 1 ) -- 
   ,(2,1, 0, 0 )
   ,(2,2, 1, 0 )
   ,(3,1, 0, 0 )
   ,(3,2, 0, 0 )
   )
   select * from events2

我有预感，答案将与这个问题的答案有关：PostgreSQL window function: partition by comparison

不知怎的，我自己也弄不明白..

彼得

【问题讨论】：

标签： postgresql

【解决方案1】：

通过标量子查询进行 COALESCE：

WITH events(id, zrow, zevent) AS (
    VALUES
    (1,1, 0 ) ,(1,2, 0 ) ,(1,3, 1 ) ,(1,4, 0 ) ,(1,5, 1 )
   ,(2,1, 0 ) ,(2,2, 1 )
   ,(3,1, 0 ) ,(3,2, 0 )
   )
SELECT id, zrow, zevent
        , COALESCE((SELECT 1 FROM events ex WHERE ex.id = ev.id AND ex.zrow < ev.zrow AND ex.zevent> 0),0) AS oevent
 FROM events ev
        ;

或者，通过将布尔 EXISTS() 类型转换为 INTEGER 来避免 COALESCE()：

WITH events(id, zrow,event) AS (
    VALUES
    (1,1, 0 ) ,(1,2, 0 ) ,(1,3, 1 ) ,(1,4, 0 ) ,(1,5, 1 )
   ,(2,1, 0 ) ,(2,2, 1 )
   ,(3,1, 0 ) ,(3,2, 0 )
   )
SELECT id, zrow, event
        , EXISTS(SELECT 1 FROM events ex WHERE ex.id = ev.id AND ex.zrow < ev.zrow AND ex.event> 0)::integer AS oevent
FROM events ev
        ;

在同一组（帧）内的先前记录中查找 MAX() 值：

WITH events(id, zrow,event) AS (
    VALUES
    (1,1, 0 ) ,(1,2, 0 ) ,(1,3, 1 ) ,(1,4, 0 ) ,(1,5, 1 )
   ,(2,1, 0 ) ,(2,2, 1 )
   ,(3,1, 0 ) ,(3,2, 0 )
   )
, drag AS (
        SELECT id, zrow, event, MAX(event)
                OVER (PARTITION BY id
                        ORDER BY zrow
                        ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING
                        ) AS lagged
        FROM events ev
        )
SELECT id, zrow, event
        , COALESCE(lagged,0) AS oevent
        FROM drag dr
        ;

同样没有额外的 CTE：

WITH events(id, zrow,event) AS (
    VALUES
    (1,1, 0 ) ,(1,2, 0 ) ,(1,3, 1 ) ,(1,4, 0 ) ,(1,5, 1 )
   ,(2,1, 0 ) ,(2,2, 1 )
   ,(3,1, 0 ) ,(3,2, 0 )
   )
SELECT id, zrow, event, COALESCE(MAX(event) OVER (PARTITION BY id
                                                ORDER BY zrow
                                                ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING
                                ),0) AS lagged
        FROM events ev
        ;

另一种执行自联接的方法是使用递归查询。

【讨论】：