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_AES
  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:
  65             has_ffmpeg = FFmpegFD.available()
  66             no_crypto = not Cryptodome_AES and '#EXT-X-KEY:METHOD=AES-128' in s
  67             if no_crypto and has_ffmpeg:
  68                 can_download, message = False, 'The stream has AES-128 encryption and pycryptodomex is not available'
  69             elif no_crypto:
  70                 message = ('The stream has AES-128 encryption and neither ffmpeg nor pycryptodomex are available; '
  71                            'Decryption will be performed natively, but will be extremely slow')
  72             elif info_dict.get('extractor_key') == 'Generic' and re.search(r'(?m)#EXT-X-MEDIA-SEQUENCE:(?!0$)', s):
  73                 install_ffmpeg = '' if has_ffmpeg else 'install ffmpeg and '
  74                 message = ('Live HLS streams are not supported by the native downloader. If this is a livestream, '
  75                            f'please {install_ffmpeg}add "--downloader ffmpeg --hls-use-mpegts" to your command')
  76         if not can_download:
  77             has_drm = re.search('|'.join([
  78                 r'#EXT-X-FAXS-CM:',  # Adobe Flash Access
  79                 r'#EXT-X-(?:SESSION-)?KEY:.*?URI="skd://',  # Apple FairPlay
  80             ]), s)
  81             if has_drm and not self.params.get('allow_unplayable_formats'):
  82                 self.report_error(
  83                     'This video is DRM protected; Try selecting another format with --format or '
  84                     'add --check-formats to automatically fallback to the next best format')
  85                 return False
  86             message = message or 'Unsupported features have been detected'
  87             fd = FFmpegFD(self.ydl, self.params)
  88             self.report_warning(f'{message}; extraction will be delegated to {fd.get_basename()}')
  89             return fd.real_download(filename, info_dict)
  90         elif message:
  91             self.report_warning(message)
  92
  93         is_webvtt = info_dict['ext'] == 'vtt'
  94         if is_webvtt:
  95             real_downloader = None  # Packing the fragments is not currently supported for external downloader
  96         else:
  97             real_downloader = get_suitable_downloader(
  98                 info_dict, self.params, None, protocol='m3u8_frag_urls', to_stdout=(filename == '-'))
  99         if real_downloader and not real_downloader.supports_manifest(s):
 100             real_downloader = None
 101         if real_downloader:
 102             self.to_screen(f'[{self.FD_NAME}] Fragment downloads will be delegated to {real_downloader.get_basename()}')
 103
 104         def is_ad_fragment_start(s):
 105             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
 106                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',ad'))
 107
 108         def is_ad_fragment_end(s):
 109             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=master' in s
 110                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',segment'))
 111
 112         fragments = []
 113
 114         media_frags = 0
 115         ad_frags = 0
 116         ad_frag_next = False
 117         for line in s.splitlines():
 118             line = line.strip()
 119             if not line:
 120                 continue
 121             if line.startswith('#'):
 122                 if is_ad_fragment_start(line):
 123                     ad_frag_next = True
 124                 elif is_ad_fragment_end(line):
 125                     ad_frag_next = False
 126                 continue
 127             if ad_frag_next:
 128                 ad_frags += 1
 129                 continue
 130             media_frags += 1
 131
 132         ctx = {
 133             'filename': filename,
 134             'total_frags': media_frags,
 135             'ad_frags': ad_frags,
 136         }
 137
 138         if real_downloader:
 139             self._prepare_external_frag_download(ctx)
 140         else:
 141             self._prepare_and_start_frag_download(ctx, info_dict)
 142
 143         extra_state = ctx.setdefault('extra_state', {})
 144
 145         format_index = info_dict.get('format_index')
 146         extra_query = None
 147         extra_param_to_segment_url = info_dict.get('extra_param_to_segment_url')
 148         if extra_param_to_segment_url:
 149             extra_query = urllib.parse.parse_qs(extra_param_to_segment_url)
 150         i = 0
 151         media_sequence = 0
 152         decrypt_info = {'METHOD': 'NONE'}
 153         byte_range = {}
 154         discontinuity_count = 0
 155         frag_index = 0
 156         ad_frag_next = False
 157         for line in s.splitlines():
 158             line = line.strip()
 159             if line:
 160                 if not line.startswith('#'):
 161                     if format_index and discontinuity_count != format_index:
 162                         continue
 163                     if ad_frag_next:
 164                         continue
 165                     frag_index += 1
 166                     if frag_index <= ctx['fragment_index']:
 167                         continue
 168                     frag_url = (
 169                         line
 170                         if re.match(r'^https?://', line)
 171                         else urllib.parse.urljoin(man_url, line))
 172                     if extra_query:
 173                         frag_url = update_url_query(frag_url, extra_query)
 174
 175                     fragments.append({
 176                         'frag_index': frag_index,
 177                         'url': frag_url,
 178                         'decrypt_info': decrypt_info,
 179                         'byte_range': byte_range,
 180                         'media_sequence': media_sequence,
 181                     })
 182                     media_sequence += 1
 183
 184                 elif line.startswith('#EXT-X-MAP'):
 185                     if format_index and discontinuity_count != format_index:
 186                         continue
 187                     if frag_index > 0:
 188                         self.report_error(
 189                             'Initialization fragment found after media fragments, unable to download')
 190                         return False
 191                     frag_index += 1
 192                     map_info = parse_m3u8_attributes(line[11:])
 193                     frag_url = (
 194                         map_info.get('URI')
 195                         if re.match(r'^https?://', map_info.get('URI'))
 196                         else urllib.parse.urljoin(man_url, map_info.get('URI')))
 197                     if extra_query:
 198                         frag_url = update_url_query(frag_url, extra_query)
 199
 200                     if map_info.get('BYTERANGE'):
 201                         splitted_byte_range = map_info.get('BYTERANGE').split('@')
 202                         sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 203                         byte_range = {
 204                             'start': sub_range_start,
 205                             'end': sub_range_start + int(splitted_byte_range[0]),
 206                         }
 207
 208                     fragments.append({
 209                         'frag_index': frag_index,
 210                         'url': frag_url,
 211                         'decrypt_info': decrypt_info,
 212                         'byte_range': byte_range,
 213                         'media_sequence': media_sequence
 214                     })
 215                     media_sequence += 1
 216
 217                 elif line.startswith('#EXT-X-KEY'):
 218                     decrypt_url = decrypt_info.get('URI')
 219                     decrypt_info = parse_m3u8_attributes(line[11:])
 220                     if decrypt_info['METHOD'] == 'AES-128':
 221                         if 'IV' in decrypt_info:
 222                             decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
 223                         if not re.match(r'^https?://', decrypt_info['URI']):
 224                             decrypt_info['URI'] = urllib.parse.urljoin(
 225                                 man_url, decrypt_info['URI'])
 226                         if extra_query:
 227                             decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
 228                         if decrypt_url != decrypt_info['URI']:
 229                             decrypt_info['KEY'] = None
 230
 231                 elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
 232                     media_sequence = int(line[22:])
 233                 elif line.startswith('#EXT-X-BYTERANGE'):
 234                     splitted_byte_range = line[17:].split('@')
 235                     sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 236                     byte_range = {
 237                         'start': sub_range_start,
 238                         'end': sub_range_start + int(splitted_byte_range[0]),
 239                     }
 240                 elif is_ad_fragment_start(line):
 241                     ad_frag_next = True
 242                 elif is_ad_fragment_end(line):
 243                     ad_frag_next = False
 244                 elif line.startswith('#EXT-X-DISCONTINUITY'):
 245                     discontinuity_count += 1
 246                 i += 1
 247
 248         # We only download the first fragment during the test
 249         if self.params.get('test', False):
 250             fragments = [fragments[0] if fragments else None]
 251
 252         if real_downloader:
 253             info_dict['fragments'] = fragments
 254             fd = real_downloader(self.ydl, self.params)
 255             # TODO: Make progress updates work without hooking twice
 256             # for ph in self._progress_hooks:
 257             #     fd.add_progress_hook(ph)
 258             return fd.real_download(filename, info_dict)
 259
 260         if is_webvtt:
 261             def pack_fragment(frag_content, frag_index):
 262                 output = io.StringIO()
 263                 adjust = 0
 264                 overflow = False
 265                 mpegts_last = None
 266                 for block in webvtt.parse_fragment(frag_content):
 267                     if isinstance(block, webvtt.CueBlock):
 268                         extra_state['webvtt_mpegts_last'] = mpegts_last
 269                         if overflow:
 270                             extra_state['webvtt_mpegts_adjust'] += 1
 271                             overflow = False
 272                         block.start += adjust
 273                         block.end += adjust
 274
 275                         dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
 276
 277                         ready = []
 278
 279                         i = 0
 280                         is_new = True
 281                         while i < len(dedup_window):
 282                             wcue = dedup_window[i]
 283                             wblock = webvtt.CueBlock.from_json(wcue)
 284                             i += 1
 285                             if wblock.hinges(block):
 286                                 wcue['end'] = block.end
 287                                 is_new = False
 288                                 continue
 289                             if wblock == block:
 290                                 is_new = False
 291                                 continue
 292                             if wblock.end > block.start:
 293                                 continue
 294                             ready.append(wblock)
 295                             i -= 1
 296                             del dedup_window[i]
 297
 298                         if is_new:
 299                             dedup_window.append(block.as_json)
 300                         for block in ready:
 301                             block.write_into(output)
 302
 303                         # we only emit cues once they fall out of the duplicate window
 304                         continue
 305                     elif isinstance(block, webvtt.Magic):
 306                         # take care of MPEG PES timestamp overflow
 307                         if block.mpegts is None:
 308                             block.mpegts = 0
 309                         extra_state.setdefault('webvtt_mpegts_adjust', 0)
 310                         block.mpegts += extra_state['webvtt_mpegts_adjust'] << 33
 311                         if block.mpegts < extra_state.get('webvtt_mpegts_last', 0):
 312                             overflow = True
 313                             block.mpegts += 1 << 33
 314                         mpegts_last = block.mpegts
 315
 316                         if frag_index == 1:
 317                             extra_state['webvtt_mpegts'] = block.mpegts or 0
 318                             extra_state['webvtt_local'] = block.local or 0
 319                             # XXX: block.local = block.mpegts = None ?
 320                         else:
 321                             if block.mpegts is not None and block.local is not None:
 322                                 adjust = (
 323                                     (block.mpegts - extra_state.get('webvtt_mpegts', 0))
 324                                     - (block.local - extra_state.get('webvtt_local', 0))
 325                                 )
 326                             continue
 327                     elif isinstance(block, webvtt.HeaderBlock):
 328                         if frag_index != 1:
 329                             # XXX: this should probably be silent as well
 330                             # or verify that all segments contain the same data
 331                             self.report_warning(bug_reports_message(
 332                                 'Discarding a %s block found in the middle of the stream; '
 333                                 'if the subtitles display incorrectly,'
 334                                 % (type(block).__name__)))
 335                             continue
 336                     block.write_into(output)
 337
 338                 return output.getvalue().encode()
 339
 340             def fin_fragments():
 341                 dedup_window = extra_state.get('webvtt_dedup_window')
 342                 if not dedup_window:
 343                     return b''
 344
 345                 output = io.StringIO()
 346                 for cue in dedup_window:
 347                     webvtt.CueBlock.from_json(cue).write_into(output)
 348
 349                 return output.getvalue().encode()
 350
 351             self.download_and_append_fragments(
 352                 ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
 353         else:
 354             return self.download_and_append_fragments(ctx, fragments, info_dict)