如何在Esper的match_recognize中匹配分区首个达标事件？

针对分区首个事件即为B（capacity>=100）无法触发匹配的情况，可通过以下两种方式调整模式：

方式一：用可选前缀覆盖起始场景 #

扩展原有模式，允许A出现0次或多次后接B，同时保留B→A的切换逻辑，示例代码：

match_recognize (
    partition by device_id
    measures
        first(B.capacity) as first_b_capacity,
        timestamp as trigger_time
    pattern (A* B | B A)
    define
        A as capacity < 100,
        B as capacity >= 100
)

• A* B 同时匹配两种场景：分区首个事件就是B（A出现0次）、A之后切换到B（A出现1次及以上）
• B A 保留原有B→A的切换匹配逻辑

方式二：显式匹配起始B场景 #

如果需要更精准控制，可拆分模式为三个分支，专门匹配分区首个B的情况，示例：

match_recognize (
    partition by device_id
    measures
        case
            when matched as START_B then first(B.capacity)
            when matched as A_TO_B then first(B.capacity)
            when matched as B_TO_A then first(A.capacity)
        end as trigger_value,
        timestamp as trigger_time
    pattern (START_B | A_TO_B | B_TO_A)
    define
        A as capacity < 100,
        B as capacity >= 100,
        START_B as B and prev(count(*)) = 0, -- 判断为分区首个事件
        A_TO_B as A followed by B,
        B_TO_A as B followed by A
)

关键语法说明 #

Esper中A*表示A出现0次或多次，这是合法语法；若要严格匹配0次或1次A，可改用A{0,1} B。你之前尝试的A{0} B写法不符合Esper语法规范，因此无法生效。

内容的提问来源于stack exchange，提问作者Dakotah North