Frank Yellin created BEAM-4232:
----------------------------------
Summary: ValueProvider with default value fails using DirectRunner
Key: BEAM-4232
URL: https://issues.apache.org/jira/browse/BEAM-4232
Project: Beam
Issue Type: Bug
Components: runner-dataflow
Reporter: Frank Yellin
Assignee: Thomas Groh
The bug is shown by the following short piece of code.
{code:java}
public class ShowProblem implements Serializable {
public interface MyOptions
extends PipelineOptions, DataflowPipelineOptions {
@Description("OutputFile")
@Default.String("gs://xyzzy/helloWorld")
ValueProvider<String> getOutput();
void setOutput(ValueProvider<String> value);
}
public static void main(String args[]) {
MyOptions options = PipelineOptionsFactory.fromArgs(args)
.as(MyOptions.class);
Pipeline pipeline = Pipeline.create(options);
pipeline
.apply(Create.of("hello", "world", "!"))
.apply(TextIO.write().to(options.getOutput()));
pipeline.run();
}
}
{code}
If this pipeline is run with the "–output=gs://xyzzy/myfile" option, it runs
fine both locally and on Google. If this pipeline is run without the --option,
it runs fine on Google, but fails when run locally.
When given an output argument, the ValueProvider returned by getOutput() is a
StaticValueProvider; when using the default value returns a
RuntimeValueProvider, The latter seems to have a bug hinted at in the comment
just above RuntimeValueProvider.get(). The code expects that when running the
pipeline (and only when running the pipeline), someone has already called
{code:java}
RuntimeValueProvider.setRuntimeOptions(options){code}
There is apparently no call to this method when using a DirectRunner.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
