yt_dlp/downloader/hls.py

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