Esper 聚合结果

Question

假设我有一个事件。该事件有 5 列。类型栏、姓名栏、来电号码、检测时间戳栏和收费栏。 Esper 产生以下 9 个元组：

[Event, CallPOPDWH, 385984MBSNE, time_1, 10]  
[Event, CallPOPDWH, 385984MBSNE, time_2, 30]
[Event, CallPOPDWH, 3859157UQHNI,time_3, 40]
[Event, CallPOPDWH, 3859157UQHNI,time_4, 40]
[Event, CallPOPDWH, 385984MBSNE, time_5, 40]
[Event, CallPOPDWH, 3859157UQHNI,time_6, 45]
[Event, CallPOPDWH, 385984MBSNE, time_7, 30]
[Event, CallPOPDWH, 385984MBSNE, time_8, 90]
[Event, CallPOPDWH, 385984MBSNE, time_9, 70]

我想编写一个查询，它将作为结果返回，对于每个呼叫号码，所有检测时间戳，只要费用总和大于 100。只要它找到例如呼叫号码的元组：

385984MBSNE,3859157UQHNI with sum(charge)>100 

它会返回：

[385984MBSNE,[time_1, time_2, time_5,time_7],120]
[3859157UQHNI,[time_3, time_4, time_6],125]
[385984MBSNE,[time_8, time_9],160]

重要的是存储每个呼叫号码的时间戳的窗口，当费用总和大于 100 时关闭。然后它将再次打开，直到找到与 sum>100 条件匹配的元组。I尝试过类似的东西：

select calling_number,
       window(timestamp).distinctOf(),
       sum(charge)
from Event.std:groupwin(calling_number).win:expr_batch(sum(total_call_charge_amount) > 100)
group by calling_number

或

select calling_number,
       window(timestamp).distinctOf(),
       sum(charge)
from Event.std:groupwin(calling_number).win:time(120 second))
group by calling_number
having sum(e.total_call_charge_amount) > 100 

它们似乎都不能正常工作。我观察到时间戳窗口不会删除以前的时间戳。任何帮助都会非常感激。

Answer 1

select calling_number,
       window(timestamp).distinctOf(),
       sum(charge)
from Event.win:time(120 second))
group by calling_number
having sum(total_call_charge_amount) > 100 

group-by 子句不正确。否则你所拥有的非常接近。

Windows 不会关闭。我认为您想说的是，您只想每 120 秒输出第一行（首先使用输出），或者您想丢弃呼叫号码的所有累积数据并重新开始（参见上下文或命名窗口为此）或其他什么？

对于给定的呼叫号码发生输出后从零开始，这是一种使用命名窗口的解决方案。

create window EventWindow#time(120) as Event;

insert into EventWindow select * from Event;

insert into ThresholdDetected select calling_number,  window(timestamp).distinctOf(), sum(charge)
from EventWindow
group by calling_number
having sum(total_call_charge_amount) > 100;

on ThresholdDetected as td delete from EventWindow as ew where ew.calling_number = td.calling_number;