yt_dlp/downloader/hls.py

   1 import binascii
   2 import io
   3 import re
   4
   5 from .external import FFmpegFD
   6 from .fragment import FragmentFD
   7 from .. import webvtt
   8 from ..compat import compat_urlparse
   9 from ..dependencies import Cryptodome_AES
  10 from ..downloader import get_suitable_downloader
  11 from ..utils import bug_reports_message, parse_m3u8_attributes, update_url_query
  12
  13
  14 class HlsFD(FragmentFD):
  15     """
  16     Download segments in a m3u8 manifest. External downloaders can take over
  17     the fragment downloads by supporting the 'm3u8_frag_urls' protocol and
  18     re-defining 'supports_manifest' function
  19     """
  20
  21     FD_NAME = 'hlsnative'
  22
  23     @staticmethod
  24     def can_download(manifest, info_dict, allow_unplayable_formats=False):
  25         UNSUPPORTED_FEATURES = [
  26             # r'#EXT-X-BYTERANGE',  # playlists composed of byte ranges of media files [2]
  27
  28             # Live streams heuristic does not always work (e.g. geo restricted to Germany
  29             # http://hls-geo.daserste.de/i/videoportal/Film/c_620000/622873/format,716451,716457,716450,716458,716459,.mp4.csmil/index_4_av.m3u8?null=0)
  30             # r'#EXT-X-MEDIA-SEQUENCE:(?!0$)',  # live streams [3]
  31
  32             # This heuristic also is not correct since segments may not be appended as well.
  33             # Twitch vods of finished streams have EXT-X-PLAYLIST-TYPE:EVENT despite
  34             # no segments will definitely be appended to the end of the playlist.
  35             # r'#EXT-X-PLAYLIST-TYPE:EVENT',  # media segments may be appended to the end of
  36             #                                 # event media playlists [4]
  37             # r'#EXT-X-MAP:',  # media initialization [5]
  38             # 1. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.4
  39             # 2. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.2
  40             # 3. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.2
  41             # 4. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.5
  42             # 5. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.5
  43         ]
  44         if not allow_unplayable_formats:
  45             UNSUPPORTED_FEATURES += [
  46                 r'#EXT-X-KEY:METHOD=(?!NONE|AES-128)',  # encrypted streams [1]
  47             ]
  48
  49         def check_results():
  50             yield not info_dict.get('is_live')
  51             for feature in UNSUPPORTED_FEATURES:
  52                 yield not re.search(feature, manifest)
  53         return all(check_results())
  54
  55     def real_download(self, filename, info_dict):
  56         man_url = info_dict['url']
  57         self.to_screen('[%s] Downloading m3u8 manifest' % self.FD_NAME)
  58
  59         urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
  60         man_url = urlh.geturl()
  61         s = urlh.read().decode('utf-8', 'ignore')
  62
  63         can_download, message = self.can_download(s, info_dict, self.params.get('allow_unplayable_formats')), None
  64         if can_download and not Cryptodome_AES and '#EXT-X-KEY:METHOD=AES-128' in s:
  65             if FFmpegFD.available():
  66                 can_download, message = False, 'The stream has AES-128 encryption and pycryptodomex is not available'
  67             else:
  68                 message = ('The stream has AES-128 encryption and neither ffmpeg nor pycryptodomex are available; '
  69                            'Decryption will be performed natively, but will be extremely slow')
  70         if not can_download:
  71             has_drm = re.search('|'.join([
  72                 r'#EXT-X-FAXS-CM:',  # Adobe Flash Access
  73                 r'#EXT-X-(?:SESSION-)?KEY:.*?URI="skd://',  # Apple FairPlay
  74             ]), s)
  75             if has_drm and not self.params.get('allow_unplayable_formats'):
  76                 self.report_error(
  77                     'This video is DRM protected; Try selecting another format with --format or '
  78                     'add --check-formats to automatically fallback to the next best format')
  79                 return False
  80             message = message or 'Unsupported features have been detected'
  81             fd = FFmpegFD(self.ydl, self.params)
  82             self.report_warning(f'{message}; extraction will be delegated to {fd.get_basename()}')
  83             return fd.real_download(filename, info_dict)
  84         elif message:
  85             self.report_warning(message)
  86
  87         is_webvtt = info_dict['ext'] == 'vtt'
  88         if is_webvtt:
  89             real_downloader = None  # Packing the fragments is not currently supported for external downloader
  90         else:
  91             real_downloader = get_suitable_downloader(
  92                 info_dict, self.params, None, protocol='m3u8_frag_urls', to_stdout=(filename == '-'))
  93         if real_downloader and not real_downloader.supports_manifest(s):
  94             real_downloader = None
  95         if real_downloader:
  96             self.to_screen(f'[{self.FD_NAME}] Fragment downloads will be delegated to {real_downloader.get_basename()}')
  97
  98         def is_ad_fragment_start(s):
  99             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
 100                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',ad'))
 101
 102         def is_ad_fragment_end(s):
 103             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=master' in s
 104                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',segment'))
 105
 106         fragments = []
 107
 108         media_frags = 0
 109         ad_frags = 0
 110         ad_frag_next = False
 111         for line in s.splitlines():
 112             line = line.strip()
 113             if not line:
 114                 continue
 115             if line.startswith('#'):
 116                 if is_ad_fragment_start(line):
 117                     ad_frag_next = True
 118                 elif is_ad_fragment_end(line):
 119                     ad_frag_next = False
 120                 continue
 121             if ad_frag_next:
 122                 ad_frags += 1
 123                 continue
 124             media_frags += 1
 125
 126         ctx = {
 127             'filename': filename,
 128             'total_frags': media_frags,
 129             'ad_frags': ad_frags,
 130         }
 131
 132         if real_downloader:
 133             self._prepare_external_frag_download(ctx)
 134         else:
 135             self._prepare_and_start_frag_download(ctx, info_dict)
 136
 137         extra_state = ctx.setdefault('extra_state', {})
 138
 139         format_index = info_dict.get('format_index')
 140         extra_query = None
 141         extra_param_to_segment_url = info_dict.get('extra_param_to_segment_url')
 142         if extra_param_to_segment_url:
 143             extra_query = compat_urlparse.parse_qs(extra_param_to_segment_url)
 144         i = 0
 145         media_sequence = 0
 146         decrypt_info = {'METHOD': 'NONE'}
 147         byte_range = {}
 148         discontinuity_count = 0
 149         frag_index = 0
 150         ad_frag_next = False
 151         for line in s.splitlines():
 152             line = line.strip()
 153             if line:
 154                 if not line.startswith('#'):
 155                     if format_index and discontinuity_count != format_index:
 156                         continue
 157                     if ad_frag_next:
 158                         continue
 159                     frag_index += 1
 160                     if frag_index <= ctx['fragment_index']:
 161                         continue
 162                     frag_url = (
 163                         line
 164                         if re.match(r'^https?://', line)
 165                         else compat_urlparse.urljoin(man_url, line))
 166                     if extra_query:
 167                         frag_url = update_url_query(frag_url, extra_query)
 168
 169                     fragments.append({
 170                         'frag_index': frag_index,
 171                         'url': frag_url,
 172                         'decrypt_info': decrypt_info,
 173                         'byte_range': byte_range,
 174                         'media_sequence': media_sequence,
 175                     })
 176                     media_sequence += 1
 177
 178                 elif line.startswith('#EXT-X-MAP'):
 179                     if format_index and discontinuity_count != format_index:
 180                         continue
 181                     if frag_index > 0:
 182                         self.report_error(
 183                             'Initialization fragment found after media fragments, unable to download')
 184                         return False
 185                     frag_index += 1
 186                     map_info = parse_m3u8_attributes(line[11:])
 187                     frag_url = (
 188                         map_info.get('URI')
 189                         if re.match(r'^https?://', map_info.get('URI'))
 190                         else compat_urlparse.urljoin(man_url, map_info.get('URI')))
 191                     if extra_query:
 192                         frag_url = update_url_query(frag_url, extra_query)
 193
 194                     if map_info.get('BYTERANGE'):
 195                         splitted_byte_range = map_info.get('BYTERANGE').split('@')
 196                         sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 197                         byte_range = {
 198                             'start': sub_range_start,
 199                             'end': sub_range_start + int(splitted_byte_range[0]),
 200                         }
 201
 202                     fragments.append({
 203                         'frag_index': frag_index,
 204                         'url': frag_url,
 205                         'decrypt_info': decrypt_info,
 206                         'byte_range': byte_range,
 207                         'media_sequence': media_sequence
 208                     })
 209                     media_sequence += 1
 210
 211                 elif line.startswith('#EXT-X-KEY'):
 212                     decrypt_url = decrypt_info.get('URI')
 213                     decrypt_info = parse_m3u8_attributes(line[11:])
 214                     if decrypt_info['METHOD'] == 'AES-128':
 215                         if 'IV' in decrypt_info:
 216                             decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
 217                         if not re.match(r'^https?://', decrypt_info['URI']):
 218                             decrypt_info['URI'] = compat_urlparse.urljoin(
 219                                 man_url, decrypt_info['URI'])
 220                         if extra_query:
 221                             decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
 222                         if decrypt_url != decrypt_info['URI']:
 223                             decrypt_info['KEY'] = None
 224
 225                 elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
 226                     media_sequence = int(line[22:])
 227                 elif line.startswith('#EXT-X-BYTERANGE'):
 228                     splitted_byte_range = line[17:].split('@')
 229                     sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 230                     byte_range = {
 231                         'start': sub_range_start,
 232                         'end': sub_range_start + int(splitted_byte_range[0]),
 233                     }
 234                 elif is_ad_fragment_start(line):
 235                     ad_frag_next = True
 236                 elif is_ad_fragment_end(line):
 237                     ad_frag_next = False
 238                 elif line.startswith('#EXT-X-DISCONTINUITY'):
 239                     discontinuity_count += 1
 240                 i += 1
 241
 242         # We only download the first fragment during the test
 243         if self.params.get('test', False):
 244             fragments = [fragments[0] if fragments else None]
 245
 246         if real_downloader:
 247             info_dict['fragments'] = fragments
 248             fd = real_downloader(self.ydl, self.params)
 249             # TODO: Make progress updates work without hooking twice
 250             # for ph in self._progress_hooks:
 251             #     fd.add_progress_hook(ph)
 252             return fd.real_download(filename, info_dict)
 253
 254         if is_webvtt:
 255             def pack_fragment(frag_content, frag_index):
 256                 output = io.StringIO()
 257                 adjust = 0
 258                 overflow = False
 259                 mpegts_last = None
 260                 for block in webvtt.parse_fragment(frag_content):
 261                     if isinstance(block, webvtt.CueBlock):
 262                         extra_state['webvtt_mpegts_last'] = mpegts_last
 263                         if overflow:
 264                             extra_state['webvtt_mpegts_adjust'] += 1
 265                             overflow = False
 266                         block.start += adjust
 267                         block.end += adjust
 268
 269                         dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
 270
 271                         ready = []
 272
 273                         i = 0
 274                         is_new = True
 275                         while i < len(dedup_window):
 276                             wcue = dedup_window[i]
 277                             wblock = webvtt.CueBlock.from_json(wcue)
 278                             i += 1
 279                             if wblock.hinges(block):
 280                                 wcue['end'] = block.end
 281                                 is_new = False
 282                                 continue
 283                             if wblock == block:
 284                                 is_new = False
 285                                 continue
 286                             if wblock.end > block.start:
 287                                 continue
 288                             ready.append(wblock)
 289                             i -= 1
 290                             del dedup_window[i]
 291
 292                         if is_new:
 293                             dedup_window.append(block.as_json)
 294                         for block in ready:
 295                             block.write_into(output)
 296
 297                         # we only emit cues once they fall out of the duplicate window
 298                         continue
 299                     elif isinstance(block, webvtt.Magic):
 300                         # take care of MPEG PES timestamp overflow
 301                         if block.mpegts is None:
 302                             block.mpegts = 0
 303                         extra_state.setdefault('webvtt_mpegts_adjust', 0)
 304                         block.mpegts += extra_state['webvtt_mpegts_adjust'] << 33
 305                         if block.mpegts < extra_state.get('webvtt_mpegts_last', 0):
 306                             overflow = True
 307                             block.mpegts += 1 << 33
 308                         mpegts_last = block.mpegts
 309
 310                         if frag_index == 1:
 311                             extra_state['webvtt_mpegts'] = block.mpegts or 0
 312                             extra_state['webvtt_local'] = block.local or 0
 313                             # XXX: block.local = block.mpegts = None ?
 314                         else:
 315                             if block.mpegts is not None and block.local is not None:
 316                                 adjust = (
 317                                     (block.mpegts - extra_state.get('webvtt_mpegts', 0))
 318                                     - (block.local - extra_state.get('webvtt_local', 0))
 319                                 )
 320                             continue
 321                     elif isinstance(block, webvtt.HeaderBlock):
 322                         if frag_index != 1:
 323                             # XXX: this should probably be silent as well
 324                             # or verify that all segments contain the same data
 325                             self.report_warning(bug_reports_message(
 326                                 'Discarding a %s block found in the middle of the stream; '
 327                                 'if the subtitles display incorrectly,'
 328                                 % (type(block).__name__)))
 329                             continue
 330                     block.write_into(output)
 331
 332                 return output.getvalue().encode()
 333
 334             def fin_fragments():
 335                 dedup_window = extra_state.get('webvtt_dedup_window')
 336                 if not dedup_window:
 337                     return b''
 338
 339                 output = io.StringIO()
 340                 for cue in dedup_window:
 341                     webvtt.CueBlock.from_json(cue).write_into(output)
 342
 343                 return output.getvalue().encode()
 344
 345             self.download_and_append_fragments(
 346                 ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
 347         else:
 348             return self.download_and_append_fragments(ctx, fragments, info_dict)