-from __future__ import unicode_literals
-
-import re
import binascii
-try:
- from Crypto.Cipher import AES
- can_decrypt_frag = True
-except ImportError:
- can_decrypt_frag = False
-try:
- import concurrent.futures
- can_threaded_download = True
-except ImportError:
- can_threaded_download = False
-
-from ..downloader import _get_real_downloader
-from .fragment import FragmentFD
-from .external import FFmpegFD
+import io
+import re
+import urllib.parse
-from ..compat import (
- compat_urllib_error,
- compat_urlparse,
- compat_struct_pack,
-)
+from . import get_suitable_downloader
+from .external import FFmpegFD
+from .fragment import FragmentFD
+from .. import webvtt
+from ..dependencies import Cryptodome
from ..utils import (
+ bug_reports_message,
parse_m3u8_attributes,
- sanitize_open,
+ remove_start,
+ traverse_obj,
update_url_query,
+ urljoin,
)
class HlsFD(FragmentFD):
"""
Download segments in a m3u8 manifest. External downloaders can take over
- the fragment downloads by supporting the 'frag_urls' protocol and
+ the fragment downloads by supporting the 'm3u8_frag_urls' protocol and
re-defining 'supports_manifest' function
"""
FD_NAME = 'hlsnative'
@staticmethod
- def can_download(manifest, info_dict, allow_unplayable_formats=False, with_crypto=can_decrypt_frag):
+ def _has_drm(manifest): # TODO: https://github.com/yt-dlp/yt-dlp/pull/5039
+ return bool(re.search('|'.join((
+ r'#EXT-X-(?:SESSION-)?KEY:.*?URI="skd://', # Apple FairPlay
+ r'#EXT-X-(?:SESSION-)?KEY:.*?KEYFORMAT="com\.apple\.streamingkeydelivery"', # Apple FairPlay
+ r'#EXT-X-(?:SESSION-)?KEY:.*?KEYFORMAT="com\.microsoft\.playready"', # Microsoft PlayReady
+ r'#EXT-X-FAXS-CM:', # Adobe Flash Access
+ )), manifest))
+
+ @classmethod
+ def can_download(cls, manifest, info_dict, allow_unplayable_formats=False):
UNSUPPORTED_FEATURES = [
# r'#EXT-X-BYTERANGE', # playlists composed of byte ranges of media files [2]
]
if not allow_unplayable_formats:
UNSUPPORTED_FEATURES += [
- r'#EXT-X-KEY:METHOD=(?!NONE|AES-128)', # encrypted streams [1]
+ r'#EXT-X-KEY:METHOD=(?!NONE|AES-128)', # encrypted streams [1], but not necessarily DRM
]
def check_results():
yield not info_dict.get('is_live')
- is_aes128_enc = '#EXT-X-KEY:METHOD=AES-128' in manifest
- yield with_crypto or not is_aes128_enc
- yield not (is_aes128_enc and r'#EXT-X-BYTERANGE' in manifest)
for feature in UNSUPPORTED_FEATURES:
yield not re.search(feature, manifest)
+ if not allow_unplayable_formats:
+ yield not cls._has_drm(manifest)
return all(check_results())
def real_download(self, filename, info_dict):
self.to_screen('[%s] Downloading m3u8 manifest' % self.FD_NAME)
urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
- man_url = urlh.geturl()
+ man_url = urlh.url
s = urlh.read().decode('utf-8', 'ignore')
- if not self.can_download(s, info_dict, self.params.get('allow_unplayable_formats')):
- if info_dict.get('extra_param_to_segment_url') or info_dict.get('_decryption_key_url'):
- self.report_error('pycryptodome not found. Please install it.')
+ can_download, message = self.can_download(s, info_dict, self.params.get('allow_unplayable_formats')), None
+ if can_download:
+ has_ffmpeg = FFmpegFD.available()
+ no_crypto = not Cryptodome.AES and '#EXT-X-KEY:METHOD=AES-128' in s
+ if no_crypto and has_ffmpeg:
+ can_download, message = False, 'The stream has AES-128 encryption and pycryptodomex is not available'
+ elif no_crypto:
+ message = ('The stream has AES-128 encryption and neither ffmpeg nor pycryptodomex are available; '
+ 'Decryption will be performed natively, but will be extremely slow')
+ elif info_dict.get('extractor_key') == 'Generic' and re.search(r'(?m)#EXT-X-MEDIA-SEQUENCE:(?!0$)', s):
+ install_ffmpeg = '' if has_ffmpeg else 'install ffmpeg and '
+ message = ('Live HLS streams are not supported by the native downloader. If this is a livestream, '
+ f'please {install_ffmpeg}add "--downloader ffmpeg --hls-use-mpegts" to your command')
+ if not can_download:
+ if self._has_drm(s) and not self.params.get('allow_unplayable_formats'):
+ if info_dict.get('has_drm') and self.params.get('test'):
+ self.to_screen(f'[{self.FD_NAME}] This format is DRM protected', skip_eol=True)
+ else:
+ self.report_error(
+ 'This format is DRM protected; Try selecting another format with --format or '
+ 'add --check-formats to automatically fallback to the next best format', tb=False)
return False
- if self.can_download(s, info_dict, with_crypto=True):
- self.report_warning('pycryptodome is needed to download this file with hlsnative')
- self.report_warning(
- 'hlsnative has detected features it does not support, '
- 'extraction will be delegated to ffmpeg')
+ message = message or 'Unsupported features have been detected'
fd = FFmpegFD(self.ydl, self.params)
- # TODO: Make progress updates work without hooking twice
- # for ph in self._progress_hooks:
- # fd.add_progress_hook(ph)
+ self.report_warning(f'{message}; extraction will be delegated to {fd.get_basename()}')
return fd.real_download(filename, info_dict)
+ elif message:
+ self.report_warning(message)
- real_downloader = _get_real_downloader(info_dict, 'frag_urls', self.params, None)
+ is_webvtt = info_dict['ext'] == 'vtt'
+ if is_webvtt:
+ real_downloader = None # Packing the fragments is not currently supported for external downloader
+ else:
+ real_downloader = get_suitable_downloader(
+ info_dict, self.params, None, protocol='m3u8_frag_urls', to_stdout=(filename == '-'))
if real_downloader and not real_downloader.supports_manifest(s):
real_downloader = None
+ if real_downloader:
+ self.to_screen(f'[{self.FD_NAME}] Fragment downloads will be delegated to {real_downloader.get_basename()}')
def is_ad_fragment_start(s):
return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
if real_downloader:
self._prepare_external_frag_download(ctx)
else:
- self._prepare_and_start_frag_download(ctx)
+ self._prepare_and_start_frag_download(ctx, info_dict)
- fragment_retries = self.params.get('fragment_retries', 0)
- skip_unavailable_fragments = self.params.get('skip_unavailable_fragments', True)
- test = self.params.get('test', False)
+ extra_state = ctx.setdefault('extra_state', {})
format_index = info_dict.get('format_index')
extra_query = None
extra_param_to_segment_url = info_dict.get('extra_param_to_segment_url')
if extra_param_to_segment_url:
- extra_query = compat_urlparse.parse_qs(extra_param_to_segment_url)
+ extra_query = urllib.parse.parse_qs(extra_param_to_segment_url)
i = 0
media_sequence = 0
decrypt_info = {'METHOD': 'NONE'}
+ external_aes_key = traverse_obj(info_dict, ('hls_aes', 'key'))
+ if external_aes_key:
+ external_aes_key = binascii.unhexlify(remove_start(external_aes_key, '0x'))
+ assert len(external_aes_key) in (16, 24, 32), 'Invalid length for HLS AES-128 key'
+ external_aes_iv = traverse_obj(info_dict, ('hls_aes', 'iv'))
+ if external_aes_iv:
+ external_aes_iv = binascii.unhexlify(remove_start(external_aes_iv, '0x').zfill(32))
byte_range = {}
discontinuity_count = 0
frag_index = 0
frag_index += 1
if frag_index <= ctx['fragment_index']:
continue
- frag_url = (
- line
- if re.match(r'^https?://', line)
- else compat_urlparse.urljoin(man_url, line))
+ frag_url = urljoin(man_url, line)
if extra_query:
frag_url = update_url_query(frag_url, extra_query)
'byte_range': byte_range,
'media_sequence': media_sequence,
})
+ media_sequence += 1
elif line.startswith('#EXT-X-MAP'):
if format_index and discontinuity_count != format_index:
continue
if frag_index > 0:
self.report_error(
- 'initialization fragment found after media fragments, unable to download')
+ 'Initialization fragment found after media fragments, unable to download')
return False
frag_index += 1
map_info = parse_m3u8_attributes(line[11:])
- frag_url = (
- map_info.get('URI')
- if re.match(r'^https?://', map_info.get('URI'))
- else compat_urlparse.urljoin(man_url, map_info.get('URI')))
+ frag_url = urljoin(man_url, map_info.get('URI'))
if extra_query:
frag_url = update_url_query(frag_url, extra_query)
- fragments.append({
- 'frag_index': frag_index,
- 'url': frag_url,
- 'decrypt_info': decrypt_info,
- 'byte_range': byte_range,
- 'media_sequence': media_sequence
- })
-
if map_info.get('BYTERANGE'):
splitted_byte_range = map_info.get('BYTERANGE').split('@')
sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
'end': sub_range_start + int(splitted_byte_range[0]),
}
+ fragments.append({
+ 'frag_index': frag_index,
+ 'url': frag_url,
+ 'decrypt_info': decrypt_info,
+ 'byte_range': byte_range,
+ 'media_sequence': media_sequence
+ })
+ media_sequence += 1
+
elif line.startswith('#EXT-X-KEY'):
decrypt_url = decrypt_info.get('URI')
decrypt_info = parse_m3u8_attributes(line[11:])
if decrypt_info['METHOD'] == 'AES-128':
- if 'IV' in decrypt_info:
+ if external_aes_iv:
+ decrypt_info['IV'] = external_aes_iv
+ elif 'IV' in decrypt_info:
decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
- if not re.match(r'^https?://', decrypt_info['URI']):
- decrypt_info['URI'] = compat_urlparse.urljoin(
- man_url, decrypt_info['URI'])
- if extra_query:
- decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
- if decrypt_url != decrypt_info['URI']:
- decrypt_info['KEY'] = None
+ if external_aes_key:
+ decrypt_info['KEY'] = external_aes_key
+ else:
+ decrypt_info['URI'] = urljoin(man_url, decrypt_info['URI'])
+ if extra_query:
+ decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
+ if decrypt_url != decrypt_info['URI']:
+ decrypt_info['KEY'] = None
elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
media_sequence = int(line[22:])
elif line.startswith('#EXT-X-DISCONTINUITY'):
discontinuity_count += 1
i += 1
- media_sequence += 1
# We only download the first fragment during the test
- if test:
+ if self.params.get('test', False):
fragments = [fragments[0] if fragments else None]
if real_downloader:
- info_copy = info_dict.copy()
- info_copy['fragments'] = fragments
+ info_dict['fragments'] = fragments
fd = real_downloader(self.ydl, self.params)
# TODO: Make progress updates work without hooking twice
# for ph in self._progress_hooks:
# fd.add_progress_hook(ph)
- success = fd.real_download(filename, info_copy)
- if not success:
- return False
- else:
- def download_fragment(fragment):
- frag_index = fragment['frag_index']
- frag_url = fragment['url']
- decrypt_info = fragment['decrypt_info']
- byte_range = fragment['byte_range']
- media_sequence = fragment['media_sequence']
-
- ctx['fragment_index'] = frag_index
-
- count = 0
- headers = info_dict.get('http_headers', {})
- if byte_range:
- headers['Range'] = 'bytes=%d-%d' % (byte_range['start'], byte_range['end'] - 1)
- while count <= fragment_retries:
- try:
- success, frag_content = self._download_fragment(
- ctx, frag_url, info_dict, headers)
- if not success:
- return False, frag_index
- break
- except compat_urllib_error.HTTPError as err:
- # Unavailable (possibly temporary) fragments may be served.
- # First we try to retry then either skip or abort.
- # See https://github.com/ytdl-org/youtube-dl/issues/10165,
- # https://github.com/ytdl-org/youtube-dl/issues/10448).
- count += 1
- if count <= fragment_retries:
- self.report_retry_fragment(err, frag_index, count, fragment_retries)
- if count > fragment_retries:
- return False, frag_index
-
- if decrypt_info['METHOD'] == 'AES-128':
- iv = decrypt_info.get('IV') or compat_struct_pack('>8xq', media_sequence)
- decrypt_info['KEY'] = decrypt_info.get('KEY') or self.ydl.urlopen(
- self._prepare_url(info_dict, info_dict.get('_decryption_key_url') or decrypt_info['URI'])).read()
- # Don't decrypt the content in tests since the data is explicitly truncated and it's not to a valid block
- # size (see https://github.com/ytdl-org/youtube-dl/pull/27660). Tests only care that the correct data downloaded,
- # not what it decrypts to.
- if not test:
- frag_content = AES.new(
- decrypt_info['KEY'], AES.MODE_CBC, iv).decrypt(frag_content)
-
- return frag_content, frag_index
-
- def append_fragment(frag_content, frag_index):
- if frag_content:
- fragment_filename = '%s-Frag%d' % (ctx['tmpfilename'], frag_index)
- try:
- file, frag_sanitized = sanitize_open(fragment_filename, 'rb')
- ctx['fragment_filename_sanitized'] = frag_sanitized
- file.close()
- self._append_fragment(ctx, frag_content)
- return True
- except FileNotFoundError:
- if skip_unavailable_fragments:
- self.report_skip_fragment(frag_index)
- return True
- else:
- self.report_error(
- 'fragment %s not found, unable to continue' % frag_index)
- return False
- else:
- if skip_unavailable_fragments:
- self.report_skip_fragment(frag_index)
- return True
- else:
- self.report_error(
- 'fragment %s not found, unable to continue' % frag_index)
- return False
+ return fd.real_download(filename, info_dict)
- max_workers = self.params.get('concurrent_fragment_downloads', 1)
- if can_threaded_download and max_workers > 1:
- self.report_warning('The download speed shown is only of one thread. This is a known issue')
- with concurrent.futures.ThreadPoolExecutor(max_workers) as pool:
- futures = [pool.submit(download_fragment, fragment) for fragment in fragments]
- # timeout must be 0 to return instantly
- done, not_done = concurrent.futures.wait(futures, timeout=0)
- try:
- while not_done:
- # Check every 1 second for KeyboardInterrupt
- freshly_done, not_done = concurrent.futures.wait(not_done, timeout=1)
- done |= freshly_done
- except KeyboardInterrupt:
- for future in not_done:
- future.cancel()
- # timeout must be none to cancel
- concurrent.futures.wait(not_done, timeout=None)
- raise KeyboardInterrupt
- results = [future.result() for future in futures]
-
- for frag_content, frag_index in results:
- result = append_fragment(frag_content, frag_index)
- if not result:
- return False
+ if is_webvtt:
+ def pack_fragment(frag_content, frag_index):
+ output = io.StringIO()
+ adjust = 0
+ overflow = False
+ mpegts_last = None
+ for block in webvtt.parse_fragment(frag_content):
+ if isinstance(block, webvtt.CueBlock):
+ extra_state['webvtt_mpegts_last'] = mpegts_last
+ if overflow:
+ extra_state['webvtt_mpegts_adjust'] += 1
+ overflow = False
+ block.start += adjust
+ block.end += adjust
+
+ dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
+
+ ready = []
+
+ i = 0
+ is_new = True
+ while i < len(dedup_window):
+ wcue = dedup_window[i]
+ wblock = webvtt.CueBlock.from_json(wcue)
+ i += 1
+ if wblock.hinges(block):
+ wcue['end'] = block.end
+ is_new = False
+ continue
+ if wblock == block:
+ is_new = False
+ continue
+ if wblock.end > block.start:
+ continue
+ ready.append(wblock)
+ i -= 1
+ del dedup_window[i]
+
+ if is_new:
+ dedup_window.append(block.as_json)
+ for block in ready:
+ block.write_into(output)
+
+ # we only emit cues once they fall out of the duplicate window
+ continue
+ elif isinstance(block, webvtt.Magic):
+ # take care of MPEG PES timestamp overflow
+ if block.mpegts is None:
+ block.mpegts = 0
+ extra_state.setdefault('webvtt_mpegts_adjust', 0)
+ block.mpegts += extra_state['webvtt_mpegts_adjust'] << 33
+ if block.mpegts < extra_state.get('webvtt_mpegts_last', 0):
+ overflow = True
+ block.mpegts += 1 << 33
+ mpegts_last = block.mpegts
+
+ if frag_index == 1:
+ extra_state['webvtt_mpegts'] = block.mpegts or 0
+ extra_state['webvtt_local'] = block.local or 0
+ # XXX: block.local = block.mpegts = None ?
+ else:
+ if block.mpegts is not None and block.local is not None:
+ adjust = (
+ (block.mpegts - extra_state.get('webvtt_mpegts', 0))
+ - (block.local - extra_state.get('webvtt_local', 0))
+ )
+ continue
+ elif isinstance(block, webvtt.HeaderBlock):
+ if frag_index != 1:
+ # XXX: this should probably be silent as well
+ # or verify that all segments contain the same data
+ self.report_warning(bug_reports_message(
+ 'Discarding a %s block found in the middle of the stream; '
+ 'if the subtitles display incorrectly,'
+ % (type(block).__name__)))
+ continue
+ block.write_into(output)
+
+ return output.getvalue().encode()
+
+ def fin_fragments():
+ dedup_window = extra_state.get('webvtt_dedup_window')
+ if not dedup_window:
+ return b''
+
+ output = io.StringIO()
+ for cue in dedup_window:
+ webvtt.CueBlock.from_json(cue).write_into(output)
+
+ return output.getvalue().encode()
+
+ if len(fragments) == 1:
+ self.download_and_append_fragments(ctx, fragments, info_dict)
else:
- for fragment in fragments:
- frag_content, frag_index = download_fragment(fragment)
- result = append_fragment(frag_content, frag_index)
- if not result:
- return False
-
- self._finish_frag_download(ctx)
- return True
+ self.download_and_append_fragments(
+ ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
+ else:
+ return self.download_and_append_fragments(ctx, fragments, info_dict)