tfx: {SPAN} doesn't work as expected with GCS

System information

• Have I specified the code to reproduce the issue (Yes, No): Yes
• Environment in which the code is executed (e.g., Local(Linux/MacOS/Windows), Interactive Notebook, Google Cloud, etc): Vertex AI Pipeline, Vertex AI Notebook, GCS storage
• TensorFlow version: 2.6
• TFX Version: 1.2.0
• Python version: 3.7
• Python dependencies (from pip freeze output): None

Describe the current behavior First of all, I have CIFAR10 dataset in the following location

• gs://cifar10-csp-public/cifar10/span-1/train/train.tfrecord
• gs://cifar10-csp-public/cifar10/span-1/test/test.tfrecord

With ImportExampleGen as defined below, it failed to get the dataset from the specified pattern paths.

data_path = "gs://cifar10-csp-public"

input_config = example_gen_pb2.Input(splits=[
              example_gen_pb2.Input.Split(name='train',
                                          pattern='cifar10/span-{SPAN}/train/*'),
              example_gen_pb2.Input.Split(name='val',
                                          pattern='cifar10/span-{SPAN}/test/*')
          ])

example_gen = tfx.components.ImportExampleGen(input_base=data_path, input_config=input_config)

As inspecting the logs, it complains the files don’t exist.

OSError: No files found based on the file pattern gs://cifar10-csp-public/cifar10/span-{SPAN}/train/*

Describe the expected behavior

The expected behaviour is that ImportExampleGen can correctly retrieve the data with {SPAN} specified. As it didn’t work as expected, I have tried out the code below

data_path = "gs://cifar10-csp-public"

splits = [
  example_gen_pb2.Input.Split(name='train',pattern='span-{SPAN}/train/*'),
  example_gen_pb2.Input.Split(name='val',pattern='span-{SPAN}/test/*')
]
_, span, version = utils.calculate_splits_fingerprint_span_and_version(data_path, splits)
  
input_config = example_gen_pb2.Input(splits=[
    example_gen_pb2.Input.Split(name='train', pattern=f'span-{span}/train/*'),
    example_gen_pb2.Input.Split(name='val', pattern=f'span-{span}/test/*')
])

example_gen = tfx.components.ImportExampleGen(input_base=data_path, input_config=input_config)

With the utility function calculate_splits_fingerprint_span_and_version, it works fine now. However, I just wonder why it didn’t work in the first place. Doesn’t ImportExampleGen use calculate_splits_fingerprint_span_and_version function internally?