CI tests are broken: SchemaInferenceError

[albertvillanova]

CI test (unit, ubuntu-latest, deps-minimum) is broken, raising a SchemaInferenceError: see https://github.com/huggingface/datasets/actions/runs/3930901593/jobs/6721492004

FAILED tests/test_beam.py::BeamBuilderTest::test_download_and_prepare_sharded - datasets.arrow_writer.SchemaInferenceError: Please pass `features` or at least one example when writing data

Stack trace:

______________ BeamBuilderTest.test_download_and_prepare_sharded _______________
[gw1] linux -- Python 3.7.15 /opt/hostedtoolcache/Python/3.7.15/x64/bin/python

self = <tests.test_beam.BeamBuilderTest testMethod=test_download_and_prepare_sharded>

    @require_beam
    def test_download_and_prepare_sharded(self):
        import apache_beam as beam
    
        original_write_parquet = beam.io.parquetio.WriteToParquet
    
        expected_num_examples = len(get_test_dummy_examples())
        with tempfile.TemporaryDirectory() as tmp_cache_dir:
            builder = DummyBeamDataset(cache_dir=tmp_cache_dir, beam_runner="DirectRunner")
            with patch("apache_beam.io.parquetio.WriteToParquet") as write_parquet_mock:
                write_parquet_mock.side_effect = partial(original_write_parquet, num_shards=2)
>               builder.download_and_prepare()

tests/test_beam.py:97: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
/opt/hostedtoolcache/Python/3.7.15/x64/lib/python3.7/site-packages/datasets/builder.py:864: in download_and_prepare
    **download_and_prepare_kwargs,
/opt/hostedtoolcache/Python/3.7.15/x64/lib/python3.7/site-packages/datasets/builder.py:1976: in _download_and_prepare
    num_examples, num_bytes = beam_writer.finalize(metrics.query(m_filter))
/opt/hostedtoolcache/Python/3.7.15/x64/lib/python3.7/site-packages/datasets/arrow_writer.py:694: in finalize
    shard_num_bytes, _ = parquet_to_arrow(source, destination)
/opt/hostedtoolcache/Python/3.7.15/x64/lib/python3.7/site-packages/datasets/arrow_writer.py:740: in parquet_to_arrow
    num_bytes, num_examples = writer.finalize()
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = <datasets.arrow_writer.ArrowWriter object at 0x7f6dcbb3e810>
close_stream = True

    def finalize(self, close_stream=True):
        self.write_rows_on_file()
        # In case current_examples < writer_batch_size, but user uses finalize()
        if self._check_duplicates:
            self.check_duplicate_keys()
            # Re-intializing to empty list for next batch
            self.hkey_record = []
        self.write_examples_on_file()
        # If schema is known, infer features even if no examples were written
        if self.pa_writer is None and self.schema:
            self._build_writer(self.schema)
        if self.pa_writer is not None:
            self.pa_writer.close()
            self.pa_writer = None
            if close_stream:
                self.stream.close()
        else:
            if close_stream:
                self.stream.close()
>           raise SchemaInferenceError("Please pass `features` or at least one example when writing data")
E           datasets.arrow_writer.SchemaInferenceError: Please pass `features` or at least one example when writing data

/opt/hostedtoolcache/Python/3.7.15/x64/lib/python3.7/site-packages/datasets/arrow_writer.py:593: SchemaInferenceError