""" Created 10-13-19 by Matt C. McCallum """ # Local imports from audio_utils import mp3_to_wav # Third party imports # None. # Python standard library imports from multiprocessing import Pool import os import tempfile import traceback from functools import wraps import logging def estimator(func): """ Simple wrapper function around a function that analyizes a file. The wrapper logs the function that is analyzing the file and the file that is being analyzed. Args: func: function - A file analysis function that takes the filename as the first argument. Return: function - The wrapped function (with logging). """ @wraps(func) def est_func(fname, *args, **kwargs): logging.info('Analyzing "{}" estimator for track: {}'.format(func.__name__, fname)) try: with tempfile.NamedTemporaryFile(mode='wb', suffix='.wav', prefix='tmp') as temp_audio_file: with open(fname, 'rb') as mp3_file: temp_audio_file.write(mp3_to_wav(mp3_file).read()) result = func(temp_audio_file.name, *args, **kwargs) return result[0], fname except Exception: logging.error('Failed to analyze "{}" for track: {}'.format(func.__name__, fname), exc_info=True) return [[], fname] return est_func def process_estimator(args, estimator, output_dir, num_threads): """ Process all files provided by a given algorithm and places the results as new-line separated values in a text file. Args: args: list(tuple(str, *)) - A list of sets of arguments to pass to the estimator function, iteratively. The first in each argument tuple should be the filename of the mp3 audio to be analyzed. The remaining arguments may be additional metadata used in estimation. estimator: function - A function that takes in an audio filename and produces estimates of beat positions as list of float values (in seconds). output_dir: str - The path to a directory within which to save the beat position estimates as individual text files - one per file specified in `filenames`. num_threads: int - The number of threads to use to analyze the set of files. each file for analysis is assigned to a single one of these threads, while the files themselves are split between threads. """ # Analyze beats if num_threads > 1: the_pool = Pool(num_threads, maxtasksperchild=1) estimates = the_pool.starmap(estimator, args) the_pool.close() the_pool.join() else: estimates = [estimator(*arg) for arg in args] logging.info('Saving results for estimator: "{}"'.format(estimator.__name__)) # Save beats for est in estimates: output_fname = os.path.join(output_dir, os.path.splitext(os.path.basename(est[1]))[0] + '.txt') with open(output_fname, 'w') as f: f.write('\n'.join([str(time_marker) for time_marker in est[0]]))