[ 
https://issues.apache.org/jira/browse/FLINK-4239?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15392029#comment-15392029
 ] 

ASF GitHub Bot commented on FLINK-4239:
---------------------------------------

Github user kl0u commented on a diff in the pull request:

    https://github.com/apache/flink/pull/2278#discussion_r72077206
  
    --- Diff: 
flink-tests/src/test/java/org/apache/flink/test/windowing/sessionwindows/SessionWindowITCase.java
 ---
    @@ -114,8 +115,8 @@ private void runTest(
                        windowedStream = 
windowedStream.allowedLateness(Time.milliseconds(ALLOWED_LATENESS_MS));
                }
     
    -           if (!PURGE_WINDOW_ON_FIRE) {
    -                   windowedStream = windowedStream.trigger(new 
NonPurgingEventTimeTriggerWrapper());
    +           if (PURGE_WINDOW_ON_FIRE) {
    +                   windowedStream = 
windowedStream.trigger(PurgingTrigger.of(EventTimeTrigger.create()));
                }
    --- End diff --
    
    Thanks @aljoscha 


> Set Default Allowed Lateness to Zero and Make Triggers Non-Purging
> ------------------------------------------------------------------
>
>                 Key: FLINK-4239
>                 URL: https://issues.apache.org/jira/browse/FLINK-4239
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Streaming
>    Affects Versions: 1.1.0
>            Reporter: Aljoscha Krettek
>            Assignee: Aljoscha Krettek
>
> In preparation for upcoming changes in 1.2 we should set the default allowed 
> lateness to zero and make all built-in triggers non-purging by default. 
> Currently, {{EventTimeTrigger}} and {{ProcessingTimeTrigger}} purge when they 
> fire. This leads to unexpected behavior when a user sets some meaningful 
> allowed lateness. The behavior will be that the window is purged when firing 
> and the state will not actually be kept within the allowed lateness.
> Changing the behavior to non-purging requires changing the default allowed 
> lateness from {{Long.MAX_VALUE}} to {{0}}. Otherwise we would have memory 
> leaks.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to