8a70e91ab9878edde4a4abece136f931d032f687,tensorflow_transform/beam/impl_test.py,BeamImplTest,testAnalyzeBeforeTransform,#BeamImplTest#,94

Before Change

        (input_data, input_metadata)
        | beam_impl.AnalyzeDataset(
            preprocessing_fn,
            os.path.join(self.get_temp_dir(), "analyze_before_transform_a")))
    transformed_eval_dataset = (
        ((eval_data, input_metadata), transform_fn)
        | beam_impl.TransformDataset())
    self.assertDatasetsEqual(
        transformed_eval_dataset,
        (expected_transformed_eval_data, expected_transformed_metadata))

After Change

    // Redo test with eval data, using AnalyzeDataset instead of
    // AnalyzeAndTransformDataset to genereate transform_fn.
    with beam_impl.Context(temp_dir=self.get_temp_dir()):
      transform_fn = (
          (input_data, input_metadata)
          | beam_impl.AnalyzeDataset(preprocessing_fn))
      transformed_eval_dataset = (
          ((eval_data, input_metadata), transform_fn)
          | beam_impl.TransformDataset())
    self.assertDatasetsEqual(
        transformed_eval_dataset,
        (expected_transformed_eval_data, expected_transformed_metadata))

Instances