[
https://issues.apache.org/jira/browse/BEAM-7864?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16913794#comment-16913794
]
Kyle Weaver commented on BEAM-7864:
-----------------------------------
The underlying coder to the LengthPrefixCoder is ByteArrayCoder, which is the
fallback because we have unknown coder URN "beam:coder:pickled_python:v1". The
reshuffle transform is just receiving an array of bytes, which have been
presumably pickled somehow. We will need to unpickle them if we want to
separate keys and values. I'm not sure if that's possible.
> Portable spark Reshuffle coder cast exception
> ---------------------------------------------
>
> Key: BEAM-7864
> URL: https://issues.apache.org/jira/browse/BEAM-7864
> Project: Beam
> Issue Type: Bug
> Components: runner-spark
> Reporter: Kyle Weaver
> Assignee: Kyle Weaver
> Priority: Major
> Labels: portability-spark
>
> running :sdks:python:test-suites:portable:py35:portableWordCountBatch in
> either loopback or docker mode on master fails with exception:
>
> java.lang.ClassCastException: org.apache.beam.sdk.coders.LengthPrefixCoder
> cannot be cast to org.apache.beam.sdk.coders.KvCoder
> at
> org.apache.beam.runners.spark.translation.SparkBatchPortablePipelineTranslator.translateReshuffle(SparkBatchPortablePipelineTranslator.java:400)
> at
> org.apache.beam.runners.spark.translation.SparkBatchPortablePipelineTranslator.translate(SparkBatchPortablePipelineTranslator.java:147)
> at
> org.apache.beam.runners.spark.SparkPipelineRunner.lambda$run$1(SparkPipelineRunner.java:96)
> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
