class DownloadManager():
#! Big pool sizes on slow connections will lead to more incomplete downloads
poolSize = 4
def __init__(self):
# start a server for objects shared across processes
self.displayManager = DisplayManager()
self.downloadTracker = DownloadTracker()
self.displayManager.clear()
if sys.platform == "win32":
#! ProactorEventLoop is required on Windows to run subprocess asynchronously
#! it is default since Python 3.8 but has to be changed for previous versions
loop = asyncio.ProactorEventLoop()
asyncio.set_event_loop(loop)
self.loop = asyncio.get_event_loop()
#! semaphore is required to limit concurrent asyncio executions
self.semaphore = asyncio.Semaphore(self.poolSize)
#! thread pool executor is used to run blocking (CPU-bound) code from a thread
self.thread_executor = concurrent.futures.ThreadPoolExecutor(
max_workers=self.poolSize)
def download_single_song(self, songObj: SongObj) -> None:
'''
`songObj` `song` : song to be downloaded
RETURNS `~`
downloads the given song
'''
self.downloadTracker.clear()
self.downloadTracker.load_song_list([songObj])
self.displayManager.reset()
self.displayManager.set_song_count_to(1)
self._download_asynchronously([songObj])
def download_multiple_songs(self, songObjList: List[SongObj]) -> None:
'''
`list<songObj>` `songObjList` : list of songs to be downloaded
RETURNS `~`
downloads the given songs in parallel
'''
self.downloadTracker.clear()
self.downloadTracker.load_song_list(songObjList)
self.displayManager.reset()
self.displayManager.set_song_count_to(len(songObjList))
self._download_asynchronously(songObjList)
def resume_download_from_tracking_file(self, trackingFilePath: str) -> None:
'''
`str` `trackingFilePath` : path to a .spotdlTrackingFile
RETURNS `~`
downloads songs present on the .spotdlTrackingFile in parallel
'''
self.downloadTracker.clear()
self.downloadTracker.load_tracking_file(trackingFilePath)
songObjList = self.downloadTracker.get_song_list()
self.displayManager.reset()
self.displayManager.set_song_count_to(len(songObjList))
self._download_asynchronously(songObjList)
async def download_song(self, songObj: SongObj) -> None:
'''
`songObj` `songObj` : song to be downloaded
RETURNS `~`
Downloads, Converts, Normalizes song & embeds metadata as ID3 tags.
'''
#! all YouTube downloads are to .\Temp; they are then converted and put into .\ and
#! finally followed up with ID3 metadata tags
#! we explicitly use the os.path.join function here to ensure download is
#! platform agnostic
# Create a .\Temp folder if not present
tempFolder = Path('.', 'Temp')
if not tempFolder.exists():
tempFolder.mkdir()
# build file name of converted file
artistStr = ''
#! we eliminate contributing artist names that are also in the song name, else we
#! would end up with things like 'Jetta, Mastubs - I'd love to change the world
#! (Mastubs REMIX).mp3' which is kinda an odd file name.
for artist in songObj.get_contributing_artists():
if artist.lower() not in songObj.get_song_name().lower():
artistStr += artist + ', '
#! the ...[:-2] is to avoid the last ', ' appended to artistStr
convertedFileName = artistStr[:-2] + ' - ' + songObj.get_song_name()
#! this is windows specific (disallowed chars)
for disallowedChar in ['/', '?', '\\', '*', '|', '<', '>']:
if disallowedChar in convertedFileName:
convertedFileName = convertedFileName.replace(
disallowedChar, '')
#! double quotes (") and semi-colons (:) are also disallowed characters but we would
#! like to retain their equivalents, so they aren't removed in the prior loop
convertedFileName = convertedFileName.replace(
'"', "'").replace(': ', ' - ')
convertedFilePath = Path(".", f"{convertedFileName}.opus")
# if a song is already downloaded skip it
if convertedFilePath.is_file():
if self.displayManager:
self.displayManager.notify_download_skip()
if self.downloadTracker:
self.downloadTracker.notify_download_completion(songObj)
#! None is the default return value of all functions, we just explicitly define
#! it here as a continent way to avoid executing the rest of the function.
return None
# download Audio from YouTube
if self.displayManager:
youtubeHandler = YouTube(
url=songObj.get_youtube_link(),
on_progress_callback=self.displayManager.pytube_progress_hook
)
else:
youtubeHandler = YouTube(songObj.get_youtube_link())
trackAudioStream = youtubeHandler.streams.filter(
only_audio=True).order_by('bitrate').last()
if not trackAudioStream:
print(f"Unable to get audio stream for \"{songObj.get_song_name()}\" "
f"by \"{songObj.get_contributing_artists()[0]}\" "
f"from video \"{songObj.get_youtube_link()}\"")
return None
downloadedFilePathString = await self._download_from_youtube(convertedFileName, tempFolder,
trackAudioStream)
if downloadedFilePathString is None:
return None
downloadedFilePath = Path(downloadedFilePathString)
# convert downloaded file
command = 'ffmpeg -v quiet -y -i "%s" -acodec libopus -b:a 80K -vbr on -compression_level 10 "%s"'
#! bash/ffmpeg on Unix systems need to have excape char (\) for special characters: \$
#! alternatively the quotes could be reversed (single <-> double) in the command then
#! the windows special characters needs escaping (^): ^\ ^& ^| ^> ^< ^^
if sys.platform == 'win32':
formattedCommand = command % (
str(downloadedFilePath),
str(convertedFilePath)
)
else:
formattedCommand = command % (
str(downloadedFilePath).replace('$', '\$'),
str(convertedFilePath).replace('$', '\$')
)
process = await asyncio.subprocess.create_subprocess_shell(formattedCommand)
_ = await process.communicate()
#! Wait till converted file is actually created
while True:
if convertedFilePath.is_file():
break
if self.displayManager:
self.displayManager.notify_conversion_completion()
# embed song details
audioFile = OggOpus(convertedFilePath)
#! Get rid of all existing tags (if any exist)
audioFile.delete()
#! song name
audioFile['title'] = songObj.get_song_name()
audioFile['titlesort'] = songObj.get_song_name()
#! track number
audioFile['tracknumber'] = str(songObj.get_track_number())
#! genres
audioFile['genre'] = songObj.get_genres()
#! all involved artists
audioFile['artist'] = songObj.get_contributing_artists()
#! album name
audioFile['album'] = songObj.get_album_name()
#! album artist (all of 'em)
audioFile['albumartist'] = songObj.get_album_artists()
#! album release date (to what ever precision available)
audioFile['date'] = songObj.get_album_release()
audioFile['originaldate'] = songObj.get_album_release()
audioFile.save()
#! setting the album art
audioFile = OggOpus(convertedFilePath)
rawAlbumArt = urlopen(songObj.get_album_cover_url()).read()
picture = Picture()
picture.data = rawAlbumArt
picture.type = 3
picture.desc = u"Cover"
picture.mime = u"image/jpeg"
picture_data = picture.write()
encoded_data = base64.b64encode(picture_data)
vcomment_value = encoded_data.decode("ascii")
audioFile["metadata_block_picture"] = [vcomment_value]
audioFile.save()
# Do the necessary cleanup
if self.displayManager:
self.displayManager.notify_download_completion()
if self.downloadTracker:
self.downloadTracker.notify_download_completion(songObj)
# delete the unnecessary YouTube download File
if downloadedFilePath and downloadedFilePath.is_file():
downloadedFilePath.unlink()
def close(self) -> None:
'''
RETURNS `~`
cleans up across all processes
'''
self.displayManager.close()
async def _download_from_youtube(self, convertedFileName, tempFolder, trackAudioStream):
#! The following function calls blocking code, which would block whole event loop.
#! Therefore it has to be called in a separate thread via ThreadPoolExecutor. This
#! is not a problem, since GIL is released for the I/O operations, so it shouldn't
#! hurt performance.
return await self.loop.run_in_executor(
self.thread_executor,
self._perform_audio_download,
convertedFileName,
tempFolder,
trackAudioStream
)
def _perform_audio_download(self, convertedFileName, tempFolder, trackAudioStream):
#! The actual download, if there is any error, it'll be here,
try:
#! pyTube will save the song in .\Temp\$songName.mp4 or .webm, it doesn't save as '.mp3'
downloadedFilePath = trackAudioStream.download(
output_path=tempFolder,
filename=convertedFileName,
skip_existing=False
)
return downloadedFilePath
except:
#! This is equivalent to a failed download, we do nothing, the song remains on
#! downloadTrackers download queue and all is well...
#!
#! None is again used as a convenient exit
tempFiles = Path(tempFolder).glob(f'{convertedFileName}.*')
for tempFile in tempFiles:
tempFile.unlink()
return None
async def _pool_download(self, song_obj: SongObj):
#! Run asynchronous task in a pool to make sure that all processes
#! don't run at once.
# tasks that cannot acquire semaphore will wait here until it's free
# only certain amount of tasks can acquire the semaphore at the same time
async with self.semaphore:
return await self.download_song(song_obj)
def _download_asynchronously(self, song_obj_list):
tasks = [self._pool_download(song) for song in song_obj_list]
# call all task asynchronously, and wait until all are finished
self.loop.run_until_complete(asyncio.gather(*tasks))
def download_song(songObj: SongObj,
displayManager: DisplayManager = None,
downloadTracker: DownloadTracker = None) -> None:
'''
`songObj` `songObj` : song to be downloaded
`AutoProxy` `displayManager` : autoproxy reference to a `DisplayManager`
`AutoProxy` `downloadTracker`: autoproxy reference to a `DownloadTracker`
RETURNS `~`
Downloads, Converts, Normalizes song & embeds metadata as ID3 tags.
'''
#! all YouTube downloads are to .\Temp; they are then converted and put into .\ and
#! finally followed up with ID3 metadata tags
#! we explicitly use the os.path.join function here to ensure download is
#! platform agnostic
# Create a .\Temp folder if not present
tempFolder = join('.', 'Temp')
if not exists(tempFolder):
mkdir(tempFolder)
# build file name of converted file
artistStr = ''
#! we eliminate contributing artist names that are also in the song name, else we
#! would end up with things like 'Jetta, Mastubs - I'd love to change the world
#! (Mastubs REMIX).mp3' which is kinda an odd file name.
for artist in songObj.get_contributing_artists():
if artist.lower() not in songObj.get_song_name().lower():
artistStr += artist + ', '
#! the ...[:-2] is to avoid the last ', ' appended to artistStr
convertedFileName = artistStr[:-2] + ' - ' + songObj.get_song_name()
#! this is windows specific (disallowed chars)
for disallowedChar in ['/', '?', '\\', '*', '|', '<', '>']:
if disallowedChar in convertedFileName:
convertedFileName = convertedFileName.replace(disallowedChar, '')
#! double quotes (") and semi-colons (:) are also disallowed characters but we would
#! like to retain their equivalents, so they aren't removed in the prior loop
convertedFileName = convertedFileName.replace('"',
"'").replace(': ', ' - ')
#! if a songObj's playlistIndex is not None then we prepend it to keep the playlist/album order
playlistIndex = songObj.get_playlist_index()
if playlistIndex is not None:
convertedFileName = '{:04d}'.format(
playlistIndex) + ' - ' + convertedFileName
convertedFilePath = join('.', convertedFileName) + '.mp3'
# if a song is already downloaded skip it
if exists(convertedFilePath):
if displayManager:
displayManager.notify_download_skip()
if downloadTracker:
downloadTracker.notify_download_completion(songObj)
#! None is the default return value of all functions, we just explicitly define
#! it here as a continent way to avoid executing the rest of the function.
return None
# download Audio from YouTube
if displayManager:
youtubeHandler = YouTube(
url=songObj.get_youtube_link(),
on_progress_callback=displayManager.pytube_progress_hook)
else:
youtubeHandler = YouTube(songObj.get_youtube_link())
trackAudioStream = youtubeHandler.streams.get_audio_only()
#! The actual download, if there is any error, it'll be here,
try:
#! pyTube will save the song in .\Temp\$songName.mp4, it doesn't save as '.mp3'
downloadedFilePath = trackAudioStream.download(
output_path=tempFolder,
filename=convertedFileName,
skip_existing=False)
except:
#! This is equivalent to a failed download, we do nothing, the song remains on
#! downloadTrackers download queue and all is well...
#!
#! None is again used as a convenient exit
remove(join(tempFolder, convertedFileName) + '.mp4')
return None
# convert downloaded file to MP3 with normalization
#! -af loudnorm=I=-7:LRA applies EBR 128 loudness normalization algorithm with
#! intergrated loudness target (I) set to -17, using values lower than -15
#! causes 'pumping' i.e. rhythmic variation in loudness that should not
#! exist -loud parts exaggerate, soft parts left alone.
#!
#! apad=pad_dur=2 adds 2 seconds of silence toward the end of the track, this is
#! done because the loudnorm filter clips/cuts/deletes the last 1-2 seconds on
#! occasion especially if the song is EDM-like, so we add a few extra seconds to
#! combat that.
#!
#! -acodec libmp3lame sets the encoded to 'libmp3lame' which is far better
#! than the default 'mp3_mf', '-abr true' automatically determines and passes the
#! audio encoding bitrate to the filters and encoder. This ensures that the
#! sampled length of songs matches the actual length (i.e. a 5 min song won't display
#! as 47 seconds long in your music player, yeah that was an issue earlier.)
command = 'ffmpeg -v quiet -y -i "%s" -acodec libmp3lame -abr true -af loudnorm=I=-17 "%s"'
formattedCommand = command % (downloadedFilePath, convertedFilePath)
run_in_shell(formattedCommand)
#! Wait till converted file is actually created
while True:
if exists(convertedFilePath):
break
if displayManager:
displayManager.notify_conversion_completion()
# embed song details
#! we save tags as both ID3 v2.3 and v2.4
#! The simple ID3 tags
audioFile = EasyID3(convertedFilePath)
#! Get rid of all existing ID3 tags (if any exist)
audioFile.delete()
#! song name
audioFile['title'] = songObj.get_song_name()
audioFile['titlesort'] = songObj.get_song_name()
#! track number
audioFile['tracknumber'] = str(songObj.get_track_number())
#! genres (pretty pointless if you ask me)
#! we only apply the first available genre as ID3 v2.3 doesn't support multiple
#! genres and ~80% of the world PC's run Windows - an OS with no ID3 v2.4 support
genres = songObj.get_genres()
if len(genres) > 0:
audioFile['genre'] = genres[0]
#! all involved artists
audioFile['artist'] = songObj.get_contributing_artists()
#! album name
audioFile['album'] = songObj.get_album_name()
#! album artist (all of 'em)
audioFile['albumartist'] = songObj.get_album_artists()
#! album release date (to what ever precision available)
audioFile['date'] = songObj.get_album_release()
audioFile['originaldate'] = songObj.get_album_release()
#! spotify link: in case you wanna re-download your whole offline library,
#! you can just read the links from the tags and redownload the songs.
audioFile['website'] = songObj.get_spotify_link()
#! save as both ID3 v2.3 & v2.4 as v2.3 isn't fully features and
#! windows doesn't support v2.4 until later versions of Win10
audioFile.save(v2_version=3)
#! setting the album art
audioFile = ID3(convertedFilePath)
rawAlbumArt = urlopen(songObj.get_album_cover_url()).read()
audioFile['APIC'] = AlbumCover(encoding=3,
mime='image/jpeg',
type=3,
desc='Cover',
data=rawAlbumArt)
#! adding lyrics
try:
lyrics = songObj.get_song_lyrics()
USLTOutput = USLT(encoding=3, lang=u'eng', desc=u'desc', text=lyrics)
audioFile["USLT::'eng'"] = USLTOutput
except:
pass
audioFile.save(v2_version=3)
# Do the necessary cleanup
if displayManager:
displayManager.notify_download_completion()
if downloadTracker:
downloadTracker.notify_download_completion(songObj)
# delete the unnecessary YouTube download File
remove(downloadedFilePath)
class DownloadManager():
#! Big pool sizes on slow connections will lead to more incomplete downloads
poolSize = 4
def __init__(self):
# start a server for objects shared across processes
self.displayManager = DisplayManager()
self.downloadTracker = DownloadTracker()
self.displayManager.clear()
if sys.platform == "win32":
#! ProactorEventLoop is required on Windows to run subprocess asynchronously
#! it is default since Python 3.8 but has to be changed for previous versions
loop = asyncio.ProactorEventLoop()
asyncio.set_event_loop(loop)
self.loop = asyncio.get_event_loop()
#! semaphore is required to limit concurrent asyncio executions
self.semaphore = asyncio.Semaphore(self.poolSize)
#! thread pool executor is used to run blocking (CPU-bound) code from a thread
self.thread_executor = concurrent.futures.ThreadPoolExecutor(
max_workers=self.poolSize)
def download_single_song(self, songObj: SongObj) -> None:
'''
`songObj` `song` : song to be downloaded
RETURNS `~`
downloads the given song
'''
self.downloadTracker.clear()
self.downloadTracker.load_song_list([songObj])
self.displayManager.reset()
self.displayManager.set_song_count_to(1)
self._download_asynchronously([songObj])
def download_multiple_songs(self, songObjList: List[SongObj]) -> None:
'''
`list<songObj>` `songObjList` : list of songs to be downloaded
RETURNS `~`
downloads the given songs in parallel
'''
self.downloadTracker.clear()
self.downloadTracker.load_song_list(songObjList)
self.displayManager.reset()
self.displayManager.set_song_count_to(len(songObjList))
self._download_asynchronously(songObjList)
def resume_download_from_tracking_file(self,
trackingFilePath: str) -> None:
'''
`str` `trackingFilePath` : path to a .spotdlTrackingFile
RETURNS `~`
downloads songs present on the .spotdlTrackingFile in parallel
'''
self.downloadTracker.clear()
self.downloadTracker.load_tracking_file(trackingFilePath)
songObjList = self.downloadTracker.get_song_list()
self.displayManager.reset()
self.displayManager.set_song_count_to(len(songObjList))
self._download_asynchronously(songObjList)
async def download_song(self, songObj: SongObj) -> None:
'''
`songObj` `songObj` : song to be downloaded
RETURNS `~`
Downloads, Converts, Normalizes song & embeds metadata as ID3 tags.
'''
#! all YouTube downloads are to .\Temp; they are then converted and put into .\ and
#! finally followed up with ID3 metadata tags
#! we explicitly use the os.path.join function here to ensure download is
#! platform agnostic
# Create a .\Temp folder if not present
tempFolder = Path('.', 'Temp')
if not tempFolder.exists():
tempFolder.mkdir()
# build file name of converted file
artistStr = ''
#! we eliminate contributing artist names that are also in the song name, else we
#! would end up with things like 'Jetta, Mastubs - I'd love to change the world
#! (Mastubs REMIX).mp3' which is kinda an odd file name.
for artist in songObj.get_contributing_artists():
if artist.lower() not in songObj.get_song_name().lower():
artistStr += artist + ', '
#! the ...[:-2] is to avoid the last ', ' appended to artistStr
convertedFileName = artistStr[:-2] + ' - ' + songObj.get_song_name()
#! this is windows specific (disallowed chars)
for disallowedChar in ['/', '?', '\\', '*', '|', '<', '>']:
if disallowedChar in convertedFileName:
convertedFileName = convertedFileName.replace(
disallowedChar, '')
#! double quotes (") and semi-colons (:) are also disallowed characters but we would
#! like to retain their equivalents, so they aren't removed in the prior loop
convertedFileName = convertedFileName.replace('"', "'").replace(
': ', ' - ')
convertedFilePath = Path(".", f"{convertedFileName}.mp3")
# if a song is already downloaded skip it
if convertedFilePath.is_file():
if self.displayManager:
self.displayManager.notify_download_skip()
if self.downloadTracker:
self.downloadTracker.notify_download_completion(songObj)
#! None is the default return value of all functions, we just explicitly define
#! it here as a continent way to avoid executing the rest of the function.
return None
# download Audio from YouTube
if self.displayManager:
youtubeHandler = YouTube(
url=songObj.get_youtube_link(),
on_progress_callback=self.displayManager.pytube_progress_hook)
else:
youtubeHandler = YouTube(songObj.get_youtube_link())
trackAudioStream = youtubeHandler.streams.filter(
only_audio=True).order_by('bitrate').last()
if not trackAudioStream:
print(
f"Unable to get audio stream for \"{songObj.get_song_name()}\" "
f"by \"{songObj.get_contributing_artists()[0]}\" "
f"from video \"{songObj.get_youtube_link()}\"")
return None
downloadedFilePathString = await self._download_from_youtube(
convertedFileName, tempFolder, trackAudioStream)
if downloadedFilePathString is None:
return None
downloadedFilePath = Path(downloadedFilePathString)
# convert downloaded file to MP3 with normalization
#! -af loudnorm=I=-7:LRA applies EBR 128 loudness normalization algorithm with
#! intergrated loudness target (I) set to -17, using values lower than -15
#! causes 'pumping' i.e. rhythmic variation in loudness that should not
#! exist -loud parts exaggerate, soft parts left alone.
#!
#! dynaudnorm applies dynamic non-linear RMS based normalization, this is what
#! actually normalized the audio. The loudnorm filter just makes the apparent
#! loudness constant
#!
#! apad=pad_dur=2 adds 2 seconds of silence toward the end of the track, this is
#! done because the loudnorm filter clips/cuts/deletes the last 1-2 seconds on
#! occasion especially if the song is EDM-like, so we add a few extra seconds to
#! combat that.
#!
#! -acodec libmp3lame sets the encoded to 'libmp3lame' which is far better
#! than the default 'mp3_mf', '-abr true' automatically determines and passes the
#! audio encoding bitrate to the filters and encoder. This ensures that the
#! sampled length of songs matches the actual length (i.e. a 5 min song won't display
#! as 47 seconds long in your music player, yeah that was an issue earlier.)
command = 'ffmpeg -v quiet -y -i "%s" -acodec libmp3lame -abr true ' \
f'-b:a {trackAudioStream.bitrate} ' \
'-af "apad=pad_dur=2, dynaudnorm, loudnorm=I=-17" "%s"'
#! bash/ffmpeg on Unix systems need to have excape char (\) for special characters: \$
#! alternatively the quotes could be reversed (single <-> double) in the command then
#! the windows special characters needs escaping (^): ^\ ^& ^| ^> ^< ^^
if sys.platform == 'win32':
formattedCommand = command % (str(downloadedFilePath),
str(convertedFilePath))
else:
formattedCommand = command % (str(downloadedFilePath).replace(
'$', '\$'), str(convertedFilePath).replace('$', '\$'))
process = await asyncio.subprocess.create_subprocess_shell(
formattedCommand)
_ = await process.communicate()
#! Wait till converted file is actually created
while True:
if convertedFilePath.is_file():
break
if self.displayManager:
self.displayManager.notify_conversion_completion()
# embed song details
#! we save tags as both ID3 v2.3 and v2.4
#! The simple ID3 tags
audioFile = EasyID3(convertedFilePath)
#! Get rid of all existing ID3 tags (if any exist)
audioFile.delete()
#! song name
audioFile['title'] = songObj.get_song_name()
audioFile['titlesort'] = songObj.get_song_name()
#! track number
audioFile['tracknumber'] = str(songObj.get_track_number())
#! genres (pretty pointless if you ask me)
#! we only apply the first available genre as ID3 v2.3 doesn't support multiple
#! genres and ~80% of the world PC's run Windows - an OS with no ID3 v2.4 support
genres = songObj.get_genres()
if len(genres) > 0:
audioFile['genre'] = genres[0]
#! all involved artists
audioFile['artist'] = songObj.get_contributing_artists()
#! album name
audioFile['album'] = songObj.get_album_name()
#! album artist (all of 'em)
audioFile['albumartist'] = songObj.get_album_artists()
#! album release date (to what ever precision available)
audioFile['date'] = songObj.get_album_release()
audioFile['originaldate'] = songObj.get_album_release()
#! save as both ID3 v2.3 & v2.4 as v2.3 isn't fully features and
#! windows doesn't support v2.4 until later versions of Win10
audioFile.save(v2_version=3)
#! setting the album art
audioFile = ID3(convertedFilePath)
rawAlbumArt = urlopen(songObj.get_album_cover_url()).read()
audioFile['APIC'] = AlbumCover(encoding=3,
mime='image/jpeg',
type=3,
desc='Cover',
data=rawAlbumArt)
audioFile.save(v2_version=3)
# Do the necessary cleanup
if self.displayManager:
self.displayManager.notify_download_completion()
if self.downloadTracker:
self.downloadTracker.notify_download_completion(songObj)
# delete the unnecessary YouTube download File
if downloadedFilePath and downloadedFilePath.is_file():
downloadedFilePath.unlink()
def close(self) -> None:
'''
RETURNS `~`
cleans up across all processes
'''
self.displayManager.close()
async def _download_from_youtube(self, convertedFileName, tempFolder,
trackAudioStream):
#! The following function calls blocking code, which would block whole event loop.
#! Therefore it has to be called in a separate thread via ThreadPoolExecutor. This
#! is not a problem, since GIL is released for the I/O operations, so it shouldn't
#! hurt performance.
return await self.loop.run_in_executor(self.thread_executor,
self._perform_audio_download,
convertedFileName, tempFolder,
trackAudioStream)
def _perform_audio_download(self, convertedFileName, tempFolder,
trackAudioStream):
#! The actual download, if there is any error, it'll be here,
try:
#! pyTube will save the song in .\Temp\$songName.mp4 or .webm, it doesn't save as '.mp3'
downloadedFilePath = trackAudioStream.download(
output_path=tempFolder,
filename=convertedFileName,
skip_existing=False)
return downloadedFilePath
except:
#! This is equivalent to a failed download, we do nothing, the song remains on
#! downloadTrackers download queue and all is well...
#!
#! None is again used as a convenient exit
tempFiles = Path(tempFolder).glob(f'{convertedFileName}.*')
for tempFile in tempFiles:
tempFile.unlink()
return None
async def _pool_download(self, song_obj: SongObj):
#! Run asynchronous task in a pool to make sure that all processes
#! don't run at once.
# tasks that cannot acquire semaphore will wait here until it's free
# only certain amount of tasks can acquire the semaphore at the same time
async with self.semaphore:
return await self.download_song(song_obj)
def _download_asynchronously(self, song_obj_list):
tasks = [self._pool_download(song) for song in song_obj_list]
# call all task asynchronously, and wait until all are finished
self.loop.run_until_complete(asyncio.gather(*tasks))
def download_song(
songObj: SongObj,
displayManager: DisplayManager = None,
downloadTracker: DownloadTracker = None,
) -> None:
"""
`songObj` `songObj` : song to be downloaded
`AutoProxy` `displayManager` : autoproxy reference to a `DisplayManager`
`AutoProxy` `downloadTracker`: autoproxy reference to a `DownloadTracker`
RETURNS `~`
Downloads, Converts, Normalizes song & embeds metadata as ID3 tags.
"""
#! all YouTube downloads are to .\Temp; they are then converted and put into .\ and
#! finally followed up with ID3 metadata tags
#! we explicitly use the os.path.join function here to ensure download is
#! platform agnostic
# Create a .\Temp folder if not present
tempFolder = join(".", "Temp")
if not exists(tempFolder):
mkdir(tempFolder)
# build file name of converted file
artistStr = ""
#! we eliminate contributing artist names that are also in the song name, else we
#! would end up with things like 'Jetta, Mastubs - I'd love to change the world
#! (Mastubs REMIX).mp3' which is kinda an odd file name.
for artist in songObj.get_contributing_artists():
if artist.lower() not in songObj.get_song_name().lower():
artistStr += artist + ", "
#! the ...[:-2] is to avoid the last ', ' appended to artistStr
convertedFileName = artistStr[:-2] + " - " + songObj.get_song_name()
#! this is windows specific (disallowed chars)
for disallowedChar in ["/", "?", "\\", "*", "|", "<", ">"]:
if disallowedChar in convertedFileName:
convertedFileName = convertedFileName.replace(disallowedChar, "")
#! double quotes (") and semi-colons (:) are also disallowed characters but we would
#! like to retain their equivalents, so they aren't removed in the prior loop
convertedFileName = convertedFileName.replace('"',
"'").replace(": ", " - ")
convertedFilePath = join(".", convertedFileName) + ".mp3"
# if a song is already downloaded skip it
if exists(convertedFilePath):
if displayManager:
displayManager.notify_download_skip()
if downloadTracker:
downloadTracker.notify_download_completion(songObj)
#! None is the default return value of all functions, we just explicitly define
#! it here as a continent way to avoid executing the rest of the function.
return None
Lyrics = songObj.get_lyrics()
# download Audio from YouTube
if displayManager:
youtubeHandler = YouTube(
url=songObj.get_youtube_link(),
on_progress_callback=displayManager.pytube_progress_hook,
)
else:
youtubeHandler = YouTube(songObj.get_youtube_link())
trackAudioStream = youtubeHandler.streams.get_audio_only()
#! The actual download, if there is any error, it'll be here,
try:
#! pyTube will save the song in .\Temp\$songName.mp4, it doesn't save as '.mp3'
downloadedFilePath = trackAudioStream.download(
output_path=tempFolder,
filename=convertedFileName,
skip_existing=False)
except:
#! This is equivalent to a failed download, we do nothing, the song remains on
#! downloadTrackers download queue and all is well...
#!
#! None is again used as a convenient exit
remove(join(tempFolder, convertedFileName) + ".mp4")
return None
# convert downloaded file to MP3 with normalization
#! -af loudnorm=I=-7:LRA applies EBR 128 loudness normalization algorithm with
#! intergrated loudness target (I) set to -17, using values lower than -15
#! causes 'pumping' i.e. rhythmic variation in loudness that should not
#! exist -loud parts exaggerate, soft parts left alone.
#!
#! dynaudnorm applies dynamic non-linear RMS based normalization, this is what
#! actually normalized the audio. The loudnorm filter just makes the apparent
#! loudness constant
#!
#! apad=pad_dur=2 adds 2 seconds of silence toward the end of the track, this is
#! done because the loudnorm filter clips/cuts/deletes the last 1-2 seconds on
#! occasion especially if the song is EDM-like, so we add a few extra seconds to
#! combat that.
#!
#! -acodec libmp3lame sets the encoded to 'libmp3lame' which is far better
#! than the default 'mp3_mf', '-abr true' automatically determines and passes the
#! audio encoding bitrate to the filters and encoder. This ensures that the
#! sampled length of songs matches the actual length (i.e. a 5 min song won't display
#! as 47 seconds long in your music player, yeah that was an issue earlier.)
command = 'ffmpeg -v quiet -hwaccel_output_format cuda -y -i "%s" -acodec libmp3lame -abr true "%s"'
formattedCommand = command % (downloadedFilePath, convertedFilePath)
# run_in_shell(formattedCommand)
return_code = None
return_code = call(formattedCommand)
if return_code != 0:
raise ("Error occurred during conversion, ffmpeg issue probably")
if displayManager:
displayManager.notify_conversion_completion()
# embed song details
#! we save tags as both ID3 v2.3 and v2.4
#! The simple ID3 tags
audioFile = EasyID3(convertedFilePath)
#! Get rid of all existing ID3 tags (if any exist)
audioFile.delete()
#! song name
audioFile["title"] = songObj.get_song_name()
audioFile["titlesort"] = songObj.get_song_name()
#! track number
audioFile["tracknumber"] = str(songObj.get_track_number())
#! genres (pretty pointless if you ask me)
#! we only apply the first available genre as ID3 v2.3 doesn't support multiple
#! genres and ~80% of the world PC's run Windows - an OS with no ID3 v2.4 support
genres = songObj.get_genres()
if len(genres) > 0:
audioFile["genre"] = genres[0]
#! all involved artists
audioFile["artist"] = songObj.get_contributing_artists()
#! album name
audioFile["album"] = songObj.get_album_name()
#! album artist (all of 'em)
audioFile["albumartist"] = songObj.get_album_artists()
#! album release date (to what ever precision available)
audioFile["date"] = songObj.get_album_release()
audioFile["originaldate"] = songObj.get_album_release()
#! save as both ID3 v2.3 & v2.4 as v2.3 isn't fully features and
#! windows doesn't support v2.4 until later versions of Win10
audioFile.save(v2_version=3)
#! setting the album art
audioFile = ID3(convertedFilePath)
rawAlbumArt = get(songObj.get_album_cover_url()).content
audioFile["APIC"] = AlbumCover(encoding=3,
mime="image/jpeg",
type=3,
desc="Cover",
data=rawAlbumArt)
audioFile["USLT"] = USLT(encoding=3, desc=u"Lyrics", text=Lyrics)
audioFile.save(v2_version=3)
# Do the necessary cleanup
if displayManager:
displayManager.notify_download_completion()
if downloadTracker:
downloadTracker.notify_download_completion(songObj)
# delete the unnecessary YouTube download File
remove(downloadedFilePath)