[
https://issues.apache.org/jira/browse/BEAM-6743?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16776936#comment-16776936
]
Aditya Guru commented on BEAM-6743:
-----------------------------------
Both of the above cases seem like a bug to me. If not then my understanding of
windows and triggers is at fault. Can someone provide some clarifications?
> Triggers not working for bounded data
> -------------------------------------
>
> Key: BEAM-6743
> URL: https://issues.apache.org/jira/browse/BEAM-6743
> Project: Beam
> Issue Type: Bug
> Components: sdk-java-core
> Environment: Apache Beam 2.9.0 Java
> Google Cloud Dataflow Runner
> Reporter: Aditya Guru
> Priority: Major
>
> pCollection
> .apply(Window.<String>into(FixedWindows.of(Duration.millis(100)))
> .triggering(Repeatedly.forever(AfterPane.elementCountAtLeast(1000)))
> .discardingFiredPanes().withAllowedLateness(Duration.ZERO))
> .apply(TextIO.write().withWindowedWrites().withNumShards(1).to('gs-path'));
> Here pCollection is a *bounded* PCollection. I'm trying to break it into
> files of 1000 roughly, but all I get is 2 files one having 1000 other having
> the rest of the data.
> If instead I do:-
> pCollection
> .apply(new GlobalWindow())
> .triggering(Repeatedly.forever(AfterPane.elementCountAtLeast(1000)))
> .discardingFiredPanes().withAllowedLateness(Duration.ZERO))
> .apply(TextIO.write().withWindowedWrites().withNumShards(1).to('gs-path'));
> I get just one file.
> Both of the above cases should have conceptually divided the records into
> chucks of 1000 to be written in a file.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
