yt_dlp/downloader/hls.py

   1 import binascii
   2 import io
   3 import re
   4 import urllib.parse
   5
   6 from . import get_suitable_downloader
   7 from .external import FFmpegFD
   8 from .fragment import FragmentFD
   9 from .. import webvtt
  10 from ..dependencies import Cryptodome
  11 from ..utils import (
  12     bug_reports_message,
  13     parse_m3u8_attributes,
  14     remove_start,
  15     traverse_obj,
  16     update_url_query,
  17     urljoin,
  18 )
  19
  20
  21 class HlsFD(FragmentFD):
  22     """
  23     Download segments in a m3u8 manifest. External downloaders can take over
  24     the fragment downloads by supporting the 'm3u8_frag_urls' protocol and
  25     re-defining 'supports_manifest' function
  26     """
  27
  28     FD_NAME = 'hlsnative'
  29
  30     @staticmethod
  31     def _has_drm(manifest):  # TODO: https://github.com/yt-dlp/yt-dlp/pull/5039
  32         return bool(re.search('|'.join((
  33             r'#EXT-X-(?:SESSION-)?KEY:.*?URI="skd://',  # Apple FairPlay
  34             r'#EXT-X-(?:SESSION-)?KEY:.*?KEYFORMAT="com\.apple\.streamingkeydelivery"',  # Apple FairPlay
  35             r'#EXT-X-(?:SESSION-)?KEY:.*?KEYFORMAT="com\.microsoft\.playready"',  # Microsoft PlayReady
  36             r'#EXT-X-FAXS-CM:',  # Adobe Flash Access
  37         )), manifest))
  38
  39     @classmethod
  40     def can_download(cls, manifest, info_dict, allow_unplayable_formats=False):
  41         UNSUPPORTED_FEATURES = [
  42             # r'#EXT-X-BYTERANGE',  # playlists composed of byte ranges of media files [2]
  43
  44             # Live streams heuristic does not always work (e.g. geo restricted to Germany
  45             # 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)
  46             # r'#EXT-X-MEDIA-SEQUENCE:(?!0$)',  # live streams [3]
  47
  48             # This heuristic also is not correct since segments may not be appended as well.
  49             # Twitch vods of finished streams have EXT-X-PLAYLIST-TYPE:EVENT despite
  50             # no segments will definitely be appended to the end of the playlist.
  51             # r'#EXT-X-PLAYLIST-TYPE:EVENT',  # media segments may be appended to the end of
  52             #                                 # event media playlists [4]
  53             # r'#EXT-X-MAP:',  # media initialization [5]
  54             # 1. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.4
  55             # 2. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.2
  56             # 3. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.2
  57             # 4. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.5
  58             # 5. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.5
  59         ]
  60         if not allow_unplayable_formats:
  61             UNSUPPORTED_FEATURES += [
  62                 r'#EXT-X-KEY:METHOD=(?!NONE|AES-128)',  # encrypted streams [1], but not necessarily DRM
  63             ]
  64
  65         def check_results():
  66             yield not info_dict.get('is_live')
  67             for feature in UNSUPPORTED_FEATURES:
  68                 yield not re.search(feature, manifest)
  69             if not allow_unplayable_formats:
  70                 yield not cls._has_drm(manifest)
  71         return all(check_results())
  72
  73     def real_download(self, filename, info_dict):
  74         man_url = info_dict['url']
  75         self.to_screen(f'[{self.FD_NAME}] Downloading m3u8 manifest')
  76
  77         urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
  78         man_url = urlh.url
  79         s = urlh.read().decode('utf-8', 'ignore')
  80
  81         can_download, message = self.can_download(s, info_dict, self.params.get('allow_unplayable_formats')), None
  82         if can_download:
  83             has_ffmpeg = FFmpegFD.available()
  84             no_crypto = not Cryptodome.AES and '#EXT-X-KEY:METHOD=AES-128' in s
  85             if no_crypto and has_ffmpeg:
  86                 can_download, message = False, 'The stream has AES-128 encryption and pycryptodomex is not available'
  87             elif no_crypto:
  88                 message = ('The stream has AES-128 encryption and neither ffmpeg nor pycryptodomex are available; '
  89                            'Decryption will be performed natively, but will be extremely slow')
  90             elif info_dict.get('extractor_key') == 'Generic' and re.search(r'(?m)#EXT-X-MEDIA-SEQUENCE:(?!0$)', s):
  91                 install_ffmpeg = '' if has_ffmpeg else 'install ffmpeg and '
  92                 message = ('Live HLS streams are not supported by the native downloader. If this is a livestream, '
  93                            f'please {install_ffmpeg}add "--downloader ffmpeg --hls-use-mpegts" to your command')
  94         if not can_download:
  95             if self._has_drm(s) and not self.params.get('allow_unplayable_formats'):
  96                 if info_dict.get('has_drm') and self.params.get('test'):
  97                     self.to_screen(f'[{self.FD_NAME}] This format is DRM protected', skip_eol=True)
  98                 else:
  99                     self.report_error(
 100                         'This format is DRM protected; Try selecting another format with --format or '
 101                         'add --check-formats to automatically fallback to the next best format', tb=False)
 102                 return False
 103             message = message or 'Unsupported features have been detected'
 104             fd = FFmpegFD(self.ydl, self.params)
 105             self.report_warning(f'{message}; extraction will be delegated to {fd.get_basename()}')
 106             return fd.real_download(filename, info_dict)
 107         elif message:
 108             self.report_warning(message)
 109
 110         is_webvtt = info_dict['ext'] == 'vtt'
 111         if is_webvtt:
 112             real_downloader = None  # Packing the fragments is not currently supported for external downloader
 113         else:
 114             real_downloader = get_suitable_downloader(
 115                 info_dict, self.params, None, protocol='m3u8_frag_urls', to_stdout=(filename == '-'))
 116         if real_downloader and not real_downloader.supports_manifest(s):
 117             real_downloader = None
 118         if real_downloader:
 119             self.to_screen(f'[{self.FD_NAME}] Fragment downloads will be delegated to {real_downloader.get_basename()}')
 120
 121         def is_ad_fragment_start(s):
 122             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
 123                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',ad'))
 124
 125         def is_ad_fragment_end(s):
 126             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=master' in s
 127                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',segment'))
 128
 129         fragments = []
 130
 131         media_frags = 0
 132         ad_frags = 0
 133         ad_frag_next = False
 134         for line in s.splitlines():
 135             line = line.strip()
 136             if not line:
 137                 continue
 138             if line.startswith('#'):
 139                 if is_ad_fragment_start(line):
 140                     ad_frag_next = True
 141                 elif is_ad_fragment_end(line):
 142                     ad_frag_next = False
 143                 continue
 144             if ad_frag_next:
 145                 ad_frags += 1
 146                 continue
 147             media_frags += 1
 148
 149         ctx = {
 150             'filename': filename,
 151             'total_frags': media_frags,
 152             'ad_frags': ad_frags,
 153         }
 154
 155         if real_downloader:
 156             self._prepare_external_frag_download(ctx)
 157         else:
 158             self._prepare_and_start_frag_download(ctx, info_dict)
 159
 160         extra_state = ctx.setdefault('extra_state', {})
 161
 162         format_index = info_dict.get('format_index')
 163         extra_segment_query = None
 164         if extra_param_to_segment_url := info_dict.get('extra_param_to_segment_url'):
 165             extra_segment_query = urllib.parse.parse_qs(extra_param_to_segment_url)
 166         extra_key_query = None
 167         if extra_param_to_key_url := info_dict.get('extra_param_to_key_url'):
 168             extra_key_query = urllib.parse.parse_qs(extra_param_to_key_url)
 169         i = 0
 170         media_sequence = 0
 171         decrypt_info = {'METHOD': 'NONE'}
 172         external_aes_key = traverse_obj(info_dict, ('hls_aes', 'key'))
 173         if external_aes_key:
 174             external_aes_key = binascii.unhexlify(remove_start(external_aes_key, '0x'))
 175             assert len(external_aes_key) in (16, 24, 32), 'Invalid length for HLS AES-128 key'
 176         external_aes_iv = traverse_obj(info_dict, ('hls_aes', 'iv'))
 177         if external_aes_iv:
 178             external_aes_iv = binascii.unhexlify(remove_start(external_aes_iv, '0x').zfill(32))
 179         byte_range = {}
 180         discontinuity_count = 0
 181         frag_index = 0
 182         ad_frag_next = False
 183         for line in s.splitlines():
 184             line = line.strip()
 185             if line:
 186                 if not line.startswith('#'):
 187                     if format_index and discontinuity_count != format_index:
 188                         continue
 189                     if ad_frag_next:
 190                         continue
 191                     frag_index += 1
 192                     if frag_index <= ctx['fragment_index']:
 193                         continue
 194                     frag_url = urljoin(man_url, line)
 195                     if extra_segment_query:
 196                         frag_url = update_url_query(frag_url, extra_segment_query)
 197
 198                     fragments.append({
 199                         'frag_index': frag_index,
 200                         'url': frag_url,
 201                         'decrypt_info': decrypt_info,
 202                         'byte_range': byte_range,
 203                         'media_sequence': media_sequence,
 204                     })
 205                     media_sequence += 1
 206
 207                 elif line.startswith('#EXT-X-MAP'):
 208                     if format_index and discontinuity_count != format_index:
 209                         continue
 210                     if frag_index > 0:
 211                         self.report_error(
 212                             'Initialization fragment found after media fragments, unable to download')
 213                         return False
 214                     frag_index += 1
 215                     map_info = parse_m3u8_attributes(line[11:])
 216                     frag_url = urljoin(man_url, map_info.get('URI'))
 217                     if extra_segment_query:
 218                         frag_url = update_url_query(frag_url, extra_segment_query)
 219
 220                     if map_info.get('BYTERANGE'):
 221                         splitted_byte_range = map_info.get('BYTERANGE').split('@')
 222                         sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 223                         byte_range = {
 224                             'start': sub_range_start,
 225                             'end': sub_range_start + int(splitted_byte_range[0]),
 226                         }
 227
 228                     fragments.append({
 229                         'frag_index': frag_index,
 230                         'url': frag_url,
 231                         'decrypt_info': decrypt_info,
 232                         'byte_range': byte_range,
 233                         'media_sequence': media_sequence,
 234                     })
 235                     media_sequence += 1
 236
 237                 elif line.startswith('#EXT-X-KEY'):
 238                     decrypt_url = decrypt_info.get('URI')
 239                     decrypt_info = parse_m3u8_attributes(line[11:])
 240                     if decrypt_info['METHOD'] == 'AES-128':
 241                         if external_aes_iv:
 242                             decrypt_info['IV'] = external_aes_iv
 243                         elif 'IV' in decrypt_info:
 244                             decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
 245                         if external_aes_key:
 246                             decrypt_info['KEY'] = external_aes_key
 247                         else:
 248                             decrypt_info['URI'] = urljoin(man_url, decrypt_info['URI'])
 249                             if extra_key_query or extra_segment_query:
 250                                 # Fall back to extra_segment_query to key for backwards compat
 251                                 decrypt_info['URI'] = update_url_query(
 252                                     decrypt_info['URI'], extra_key_query or extra_segment_query)
 253                             if decrypt_url != decrypt_info['URI']:
 254                                 decrypt_info['KEY'] = None
 255
 256                 elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
 257                     media_sequence = int(line[22:])
 258                 elif line.startswith('#EXT-X-BYTERANGE'):
 259                     splitted_byte_range = line[17:].split('@')
 260                     sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 261                     byte_range = {
 262                         'start': sub_range_start,
 263                         'end': sub_range_start + int(splitted_byte_range[0]),
 264                     }
 265                 elif is_ad_fragment_start(line):
 266                     ad_frag_next = True
 267                 elif is_ad_fragment_end(line):
 268                     ad_frag_next = False
 269                 elif line.startswith('#EXT-X-DISCONTINUITY'):
 270                     discontinuity_count += 1
 271                 i += 1
 272
 273         # We only download the first fragment during the test
 274         if self.params.get('test', False):
 275             fragments = [fragments[0] if fragments else None]
 276
 277         if real_downloader:
 278             info_dict['fragments'] = fragments
 279             fd = real_downloader(self.ydl, self.params)
 280             # TODO: Make progress updates work without hooking twice
 281             # for ph in self._progress_hooks:
 282             #     fd.add_progress_hook(ph)
 283             return fd.real_download(filename, info_dict)
 284
 285         if is_webvtt:
 286             def pack_fragment(frag_content, frag_index):
 287                 output = io.StringIO()
 288                 adjust = 0
 289                 overflow = False
 290                 mpegts_last = None
 291                 for block in webvtt.parse_fragment(frag_content):
 292                     if isinstance(block, webvtt.CueBlock):
 293                         extra_state['webvtt_mpegts_last'] = mpegts_last
 294                         if overflow:
 295                             extra_state['webvtt_mpegts_adjust'] += 1
 296                             overflow = False
 297                         block.start += adjust
 298                         block.end += adjust
 299
 300                         dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
 301
 302                         ready = []
 303
 304                         i = 0
 305                         is_new = True
 306                         while i < len(dedup_window):
 307                             wcue = dedup_window[i]
 308                             wblock = webvtt.CueBlock.from_json(wcue)
 309                             i += 1
 310                             if wblock.hinges(block):
 311                                 wcue['end'] = block.end
 312                                 is_new = False
 313                                 continue
 314                             if wblock == block:
 315                                 is_new = False
 316                                 continue
 317                             if wblock.end > block.start:
 318                                 continue
 319                             ready.append(wblock)
 320                             i -= 1
 321                             del dedup_window[i]
 322
 323                         if is_new:
 324                             dedup_window.append(block.as_json)
 325                         for block in ready:
 326                             block.write_into(output)
 327
 328                         # we only emit cues once they fall out of the duplicate window
 329                         continue
 330                     elif isinstance(block, webvtt.Magic):
 331                         # take care of MPEG PES timestamp overflow
 332                         if block.mpegts is None:
 333                             block.mpegts = 0
 334                         extra_state.setdefault('webvtt_mpegts_adjust', 0)
 335                         block.mpegts += extra_state['webvtt_mpegts_adjust'] << 33
 336                         if block.mpegts < extra_state.get('webvtt_mpegts_last', 0):
 337                             overflow = True
 338                             block.mpegts += 1 << 33
 339                         mpegts_last = block.mpegts
 340
 341                         if frag_index == 1:
 342                             extra_state['webvtt_mpegts'] = block.mpegts or 0
 343                             extra_state['webvtt_local'] = block.local or 0
 344                             # XXX: block.local = block.mpegts = None ?
 345                         else:
 346                             if block.mpegts is not None and block.local is not None:
 347                                 adjust = (
 348                                     (block.mpegts - extra_state.get('webvtt_mpegts', 0))
 349                                     - (block.local - extra_state.get('webvtt_local', 0))
 350                                 )
 351                             continue
 352                     elif isinstance(block, webvtt.HeaderBlock):
 353                         if frag_index != 1:
 354                             # XXX: this should probably be silent as well
 355                             # or verify that all segments contain the same data
 356                             self.report_warning(bug_reports_message(
 357                                 f'Discarding a {type(block).__name__} block found in the middle of the stream; '
 358                                 'if the subtitles display incorrectly,'))
 359                             continue
 360                     block.write_into(output)
 361
 362                 return output.getvalue().encode()
 363
 364             def fin_fragments():
 365                 dedup_window = extra_state.get('webvtt_dedup_window')
 366                 if not dedup_window:
 367                     return b''
 368
 369                 output = io.StringIO()
 370                 for cue in dedup_window:
 371                     webvtt.CueBlock.from_json(cue).write_into(output)
 372
 373                 return output.getvalue().encode()
 374
 375             if len(fragments) == 1:
 376                 self.download_and_append_fragments(ctx, fragments, info_dict)
 377             else:
 378                 self.download_and_append_fragments(
 379                     ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
 380         else:
 381             return self.download_and_append_fragments(ctx, fragments, info_dict)