rohdesamuel commented on a change in pull request #11335: [BEAM-9692]: Make CombineValues portable URL: https://github.com/apache/beam/pull/11335#discussion_r407783064
########## File path: sdks/python/apache_beam/runners/dataflow/dataflow_runner.py ########## @@ -110,22 +110,27 @@ class DataflowRunner(PipelineRunner): # Imported here to avoid circular dependencies. # TODO: Remove the apache_beam.pipeline dependency in CreatePTransformOverride + from apache_beam.runners.dataflow.ptransform_overrides import CombineValuesPTransformOverride from apache_beam.runners.dataflow.ptransform_overrides import CreatePTransformOverride from apache_beam.runners.dataflow.ptransform_overrides import ReadPTransformOverride from apache_beam.runners.dataflow.ptransform_overrides import JrhReadPTransformOverride - _PTRANSFORM_OVERRIDES = [] # type: List[PTransformOverride] + # Thesse overrides should be applied before the proto representation of the + # graph is created. + _PTRANSFORM_OVERRIDES = [ + CombineValuesPTransformOverride() Review comment: This override should place the pipeline object into the same state as if the runner had defined an apply_CombineValues, what am I missing? Looking at the code, is it because other overrides might also use a CombineValues transform so it might needed to be replaced again? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
