Performance Tweaks, Iterators, and Lazy Evaluations

[PyWoody]

A new function, dejavu.logic.decoder.find_files_g, has been created as an iterator replacement for find_files in the same file. This allows the updated Dejavu.fingerprint_directory method to utilize the concurrent.futures.ProcessPoolExecutor in conjunction with the concurrent.futures.as_completed function to submit files to be processed as they are yielded for immediate processing. Once all of the files have been submitted to the executor for processing, their respective results will be iterated over in as_completed...as they are completed.

These two modifications will allow considerable speed improvements over the existing methods. If anyone knows of where to find a large number of Creative Common openly licensed audio files to download and test on, I would be glad to post comparison results. I tried finding some today but every website was either broken, required creating accounts, or was so extensively rate-limited to be near useless.

Other minor improvements are

• Adding placeholder songs and songhashes_set in the __init__ of Dejavu to take advantage of __init__'s special dict
• Changing counts and songs_matches in Dejavu.align_matches to generator comprehensions
• Adding song_hash directly to self.songhashes_set instead of creating the variable first. Saves a lookup per iteration
• Waiting to call Dejavu.__load_fingerprinted_audio_hashes() until after all files have been processed
• Changing both channels in dejavu.logic.decoder.read to use list comprehensions