def process()

in basic_pitch/data/datasets/slakh.py [0:0]
59 lines of code
3 McCabe index (conditional complexity)

    def process(self, element: List[str]) -> List[Any]:
        import tempfile

        import numpy as np
        import ffmpeg

        from basic_pitch.constants import (
            AUDIO_N_CHANNELS,
            AUDIO_SAMPLE_RATE,
            FREQ_BINS_CONTOURS,
            FREQ_BINS_NOTES,
            ANNOTATION_HOP,
            N_FREQ_BINS_NOTES,
            N_FREQ_BINS_CONTOURS,
        )
        from basic_pitch.data import tf_example_serialization

        logging.info(f"Processing {element}")
        batch = []

        for track_id in element:
            track_remote = self.slakh_remote.track(track_id)

            with tempfile.TemporaryDirectory() as local_tmp_dir:
                slakh_local = mirdata.initialize("slakh", local_tmp_dir)
                track_local = slakh_local.track(track_id)

                for attr in self.DOWNLOAD_ATTRIBUTES:
                    source = getattr(track_remote, attr)
                    dest = getattr(track_local, attr)
                    logging.info(f"Downloading {attr} from {source} to {dest}")
                    os.makedirs(os.path.dirname(dest), exist_ok=True)
                    with self.filesystem.open(source) as s, open(dest, "wb") as d:
                        d.write(s.read())

                local_wav_path = "{}_tmp.wav".format(track_local.audio_path)
                ffmpeg.input(track_local.audio_path).output(
                    local_wav_path, ar=AUDIO_SAMPLE_RATE, ac=AUDIO_N_CHANNELS
                ).run()

                duration = float(ffmpeg.probe(local_wav_path)["format"]["duration"])
                time_scale = np.arange(0, duration + ANNOTATION_HOP, ANNOTATION_HOP)
                n_time_frames = len(time_scale)

                note_indices, note_values = track_local.notes.to_sparse_index(time_scale, "s", FREQ_BINS_NOTES, "hz")
                onset_indices, onset_values = track_local.notes.to_sparse_index(
                    time_scale, "s", FREQ_BINS_NOTES, "hz", onsets_only=True
                )
                contour_indices, contour_values = track_local.multif0.to_sparse_index(
                    time_scale, "s", FREQ_BINS_CONTOURS, "hz"
                )

                batch.append(
                    tf_example_serialization.to_transcription_tfexample(
                        track_id,
                        "slakh",
                        local_wav_path,
                        note_indices,
                        note_values,
                        onset_indices,
                        onset_values,
                        contour_indices,
                        contour_values,
                        (n_time_frames, N_FREQ_BINS_NOTES),
                        (n_time_frames, N_FREQ_BINS_CONTOURS),
                    )
                )

        logging.info(f"Finished processing batch of length {len(batch)}")
        return [batch]