[
https://issues.apache.org/jira/browse/BEAM-6407?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Niel Markwick updated BEAM-6407:
--------------------------------
Description:
When FileIO.writeDynamic is used with automatic sharding and a Contextful.Fn
that uses side inputs for the file naming, DirectRunner (and TestPipeline) fail
with:
{{java.lang.IllegalStateException: All PCollectionViews that are consumed must
be written by some WriteView PTransform: Missing [<unnamed>
[RunnerPCollectionView]]}}
Example code:
{code:java}
PCollectionView<String> outputDirectoryname =
pipeline.apply(
"outputDir",
Create.of("/tmp/testout")).apply(View.asSingleton());
Contextful.Fn<String, FileIO.Write.FileNaming> manifestNaming =
(element, c) ->
(window, pane, numShards, shardIndex, compression) ->
c.sideInput(outputDirectoryname);
pipeline.apply(FileIO.<String, String>writeDynamic()
.by(SerializableFunctions.constant(""))
.withDestinationCoder(StringUtf8Coder.of())
.via(TextIO.sink())
.withTempDirectory("/tmp")
.withNaming(Contextful.of(
manifestNaming,
Requirements.requiresSideInputs(outputDirectoryname))));
{code}
This does not occur in Dataflow-runner
It does not occur if the ContextFul.Fn is not given side inputs.
It does not occur if withNumShards(1) is set.
It did not occur in 2.8.0, and does in 2.9.0
The cause appears to be due to the DirectRunner using TransformOverrides
re-writing FileIO sinks to use runner-determined-sharding
( see [DirectRunner.java line
226|https://github.com/apache/beam/blob/master/runners/direct-java/src/main/java/org/apache/beam/runners/direct/DirectRunner.java#L226]
)
but I do not know why this started occuring in 2.9.0...
was:
When FileIO.writeDynamic is used with automatic sharding and a Contextful.Fn
that uses side inputs for the file naming, DirectRunner (and TestPipeline) fail
with:
{{java.lang.IllegalStateException: All PCollectionViews that are consumed must
be written by some WriteView PTransform: Missing [<unnamed>
[RunnerPCollectionView]]}}
Example code:
{\{PCollectionView<String> outputDirectoryname = }}
{{ pipeline.apply(}}
{{ "outputDir",}}
{{ Create.of("/tmp/testout")).apply(View.asSingleton());}}
{{Contextful.Fn<String, FileIO.Write.FileNaming> manifestNaming =}}
{{ (element, c) ->}}
\{{ (window, pane, numShards, shardIndex, compression) -> }}
{{ c.sideInput(outputDirectoryname);}}
{{pipeline.apply(FileIO.<String, String>writeDynamic()}}
{{ .by(SerializableFunctions.constant(""))}}
{{ .withDestinationCoder(StringUtf8Coder.of())}}
{{ .via(TextIO.sink())}}
{{ .withTempDirectory("/tmp")}}
{{ .withNaming(Contextful.of(}}
{{ manifestNaming,}}
{{ Requirements.requiresSideInputs(outputDirectoryname))));}}
This does not occur in Dataflow-runner
It does not occur if the ContextFul.Fn is not given side inputs.
It does not occur if withNumShards(1) is set.
It did not occur in 2.8.0, and does in 2.9.0
The cause appears to be due to the DirectRunner using TransformOverrides
re-writing FileIO sinks to use runner-determined-sharding
( see [DirectRunner.java line
226|https://github.com/apache/beam/blob/master/runners/direct-java/src/main/java/org/apache/beam/runners/direct/DirectRunner.java#L226]
)
but I do not know why this started occuring in 2.9.0...
> regression: FileIO.writeDynamic() with side inputs fails in DirectRunner
> ------------------------------------------------------------------------
>
> Key: BEAM-6407
> URL: https://issues.apache.org/jira/browse/BEAM-6407
> Project: Beam
> Issue Type: Bug
> Components: beam-model
> Affects Versions: 2.9.0
> Reporter: Niel Markwick
> Assignee: Kenneth Knowles
> Priority: Major
> Labels: regression
> Attachments: beam-filewriter-demo.tgz
>
>
> When FileIO.writeDynamic is used with automatic sharding and a Contextful.Fn
> that uses side inputs for the file naming, DirectRunner (and TestPipeline)
> fail with:
> {{java.lang.IllegalStateException: All PCollectionViews that are consumed
> must be written by some WriteView PTransform: Missing [<unnamed>
> [RunnerPCollectionView]]}}
>
> Example code:
> {code:java}
> PCollectionView<String> outputDirectoryname =
> pipeline.apply(
> "outputDir",
> Create.of("/tmp/testout")).apply(View.asSingleton());
> Contextful.Fn<String, FileIO.Write.FileNaming> manifestNaming =
> (element, c) ->
> (window, pane, numShards, shardIndex, compression) ->
> c.sideInput(outputDirectoryname);
> pipeline.apply(FileIO.<String, String>writeDynamic()
> .by(SerializableFunctions.constant(""))
> .withDestinationCoder(StringUtf8Coder.of())
> .via(TextIO.sink())
> .withTempDirectory("/tmp")
> .withNaming(Contextful.of(
> manifestNaming,
> Requirements.requiresSideInputs(outputDirectoryname))));
> {code}
>
> This does not occur in Dataflow-runner
> It does not occur if the ContextFul.Fn is not given side inputs.
> It does not occur if withNumShards(1) is set.
> It did not occur in 2.8.0, and does in 2.9.0
>
> The cause appears to be due to the DirectRunner using TransformOverrides
> re-writing FileIO sinks to use runner-determined-sharding
> ( see [DirectRunner.java line
> 226|https://github.com/apache/beam/blob/master/runners/direct-java/src/main/java/org/apache/beam/runners/direct/DirectRunner.java#L226]
> )
> but I do not know why this started occuring in 2.9.0...
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)