You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@beam.apache.org by "huangxiaofeng (JIRA)" <ji...@apache.org> on 2017/04/07 09:42:41 UTC

[jira] [Created] (BEAM-1905) ParDoTranslation should translate a ParDo more simple?

huangxiaofeng created BEAM-1905:
-----------------------------------

             Summary: ParDoTranslation should translate a ParDo more simple?
                 Key: BEAM-1905
                 URL: https://issues.apache.org/jira/browse/BEAM-1905
             Project: Beam
          Issue Type: Improvement
          Components: runner-flink
            Reporter: huangxiaofeng
            Assignee: Aljoscha Krettek


now, each Pardo will translate to parDoOperate + select(tag) + flatmap, even if it do'st have sidelnput, and it just have one output.
it will make the transformations more complicated.

for (TaggedPValue output : outputs) {
        final int outputTag = tagsToLabels.get(output.getTag());

        TypeInformation outputTypeInfo = context.getTypeInfo((PCollection<?>) output.getValue());

        @SuppressWarnings("unchecked")
        DataStream unwrapped = splitStream.select(String.valueOf(outputTag))
          .flatMap(new FlatMapFunction<RawUnionValue, Object>() {
            @Override
            public void flatMap(RawUnionValue value, Collector<Object> out) throws Exception {
              out.collect(value.getValue());
            }
          }).returns(outputTypeInfo);

        context.setOutputDataStream(output.getValue(), unwrapped);
      }



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)