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             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(
  97                 '[%s] Fragment downloads will be delegated to %s' % (self.FD_NAME, real_downloader.get_basename()))
  98
  99         def is_ad_fragment_start(s):
 100             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
 101                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',ad'))
 102
 103         def is_ad_fragment_end(s):
 104             return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=master' in s
 105                     or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',segment'))
 106
 107         fragments = []
 108
 109         media_frags = 0
 110         ad_frags = 0
 111         ad_frag_next = False
 112         for line in s.splitlines():
 113             line = line.strip()
 114             if not line:
 115                 continue
 116             if line.startswith('#'):
 117                 if is_ad_fragment_start(line):
 118                     ad_frag_next = True
 119                 elif is_ad_fragment_end(line):
 120                     ad_frag_next = False
 121                 continue
 122             if ad_frag_next:
 123                 ad_frags += 1
 124                 continue
 125             media_frags += 1
 126
 127         ctx = {
 128             'filename': filename,
 129             'total_frags': media_frags,
 130             'ad_frags': ad_frags,
 131         }
 132
 133         if real_downloader:
 134             self._prepare_external_frag_download(ctx)
 135         else:
 136             self._prepare_and_start_frag_download(ctx, info_dict)
 137
 138         extra_state = ctx.setdefault('extra_state', {})
 139
 140         format_index = info_dict.get('format_index')
 141         extra_query = None
 142         extra_param_to_segment_url = info_dict.get('extra_param_to_segment_url')
 143         if extra_param_to_segment_url:
 144             extra_query = compat_urlparse.parse_qs(extra_param_to_segment_url)
 145         i = 0
 146         media_sequence = 0
 147         decrypt_info = {'METHOD': 'NONE'}
 148         byte_range = {}
 149         discontinuity_count = 0
 150         frag_index = 0
 151         ad_frag_next = False
 152         for line in s.splitlines():
 153             line = line.strip()
 154             if line:
 155                 if not line.startswith('#'):
 156                     if format_index and discontinuity_count != format_index:
 157                         continue
 158                     if ad_frag_next:
 159                         continue
 160                     frag_index += 1
 161                     if frag_index <= ctx['fragment_index']:
 162                         continue
 163                     frag_url = (
 164                         line
 165                         if re.match(r'^https?://', line)
 166                         else compat_urlparse.urljoin(man_url, line))
 167                     if extra_query:
 168                         frag_url = update_url_query(frag_url, extra_query)
 169
 170                     fragments.append({
 171                         'frag_index': frag_index,
 172                         'url': frag_url,
 173                         'decrypt_info': decrypt_info,
 174                         'byte_range': byte_range,
 175                         'media_sequence': media_sequence,
 176                     })
 177                     media_sequence += 1
 178
 179                 elif line.startswith('#EXT-X-MAP'):
 180                     if format_index and discontinuity_count != format_index:
 181                         continue
 182                     if frag_index > 0:
 183                         self.report_error(
 184                             'Initialization fragment found after media fragments, unable to download')
 185                         return False
 186                     frag_index += 1
 187                     map_info = parse_m3u8_attributes(line[11:])
 188                     frag_url = (
 189                         map_info.get('URI')
 190                         if re.match(r'^https?://', map_info.get('URI'))
 191                         else compat_urlparse.urljoin(man_url, map_info.get('URI')))
 192                     if extra_query:
 193                         frag_url = update_url_query(frag_url, extra_query)
 194
 195                     fragments.append({
 196                         'frag_index': frag_index,
 197                         'url': frag_url,
 198                         'decrypt_info': decrypt_info,
 199                         'byte_range': byte_range,
 200                         'media_sequence': media_sequence
 201                     })
 202                     media_sequence += 1
 203
 204                     if map_info.get('BYTERANGE'):
 205                         splitted_byte_range = map_info.get('BYTERANGE').split('@')
 206                         sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 207                         byte_range = {
 208                             'start': sub_range_start,
 209                             'end': sub_range_start + int(splitted_byte_range[0]),
 210                         }
 211
 212                 elif line.startswith('#EXT-X-KEY'):
 213                     decrypt_url = decrypt_info.get('URI')
 214                     decrypt_info = parse_m3u8_attributes(line[11:])
 215                     if decrypt_info['METHOD'] == 'AES-128':
 216                         if 'IV' in decrypt_info:
 217                             decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
 218                         if not re.match(r'^https?://', decrypt_info['URI']):
 219                             decrypt_info['URI'] = compat_urlparse.urljoin(
 220                                 man_url, decrypt_info['URI'])
 221                         if extra_query:
 222                             decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
 223                         if decrypt_url != decrypt_info['URI']:
 224                             decrypt_info['KEY'] = None
 225
 226                 elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
 227                     media_sequence = int(line[22:])
 228                 elif line.startswith('#EXT-X-BYTERANGE'):
 229                     splitted_byte_range = line[17:].split('@')
 230                     sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
 231                     byte_range = {
 232                         'start': sub_range_start,
 233                         'end': sub_range_start + int(splitted_byte_range[0]),
 234                     }
 235                 elif is_ad_fragment_start(line):
 236                     ad_frag_next = True
 237                 elif is_ad_fragment_end(line):
 238                     ad_frag_next = False
 239                 elif line.startswith('#EXT-X-DISCONTINUITY'):
 240                     discontinuity_count += 1
 241                 i += 1
 242
 243         # We only download the first fragment during the test
 244         if self.params.get('test', False):
 245             fragments = [fragments[0] if fragments else None]
 246
 247         if real_downloader:
 248             info_copy = info_dict.copy()
 249             info_copy['fragments'] = fragments
 250             fd = real_downloader(self.ydl, self.params)
 251             # TODO: Make progress updates work without hooking twice
 252             # for ph in self._progress_hooks:
 253             #     fd.add_progress_hook(ph)
 254             return fd.real_download(filename, info_copy)
 255
 256         if is_webvtt:
 257             def pack_fragment(frag_content, frag_index):
 258                 output = io.StringIO()
 259                 adjust = 0
 260                 overflow = False
 261                 mpegts_last = None
 262                 for block in webvtt.parse_fragment(frag_content):
 263                     if isinstance(block, webvtt.CueBlock):
 264                         extra_state['webvtt_mpegts_last'] = mpegts_last
 265                         if overflow:
 266                             extra_state['webvtt_mpegts_adjust'] += 1
 267                             overflow = False
 268                         block.start += adjust
 269                         block.end += adjust
 270
 271                         dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
 272
 273                         ready = []
 274
 275                         i = 0
 276                         is_new = True
 277                         while i < len(dedup_window):
 278                             wcue = dedup_window[i]
 279                             wblock = webvtt.CueBlock.from_json(wcue)
 280                             i += 1
 281                             if wblock.hinges(block):
 282                                 wcue['end'] = block.end
 283                                 is_new = False
 284                                 continue
 285                             if wblock == block:
 286                                 is_new = False
 287                                 continue
 288                             if wblock.end > block.start:
 289                                 continue
 290                             ready.append(wblock)
 291                             i -= 1
 292                             del dedup_window[i]
 293
 294                         if is_new:
 295                             dedup_window.append(block.as_json)
 296                         for block in ready:
 297                             block.write_into(output)
 298
 299                         # we only emit cues once they fall out of the duplicate window
 300                         continue
 301                     elif isinstance(block, webvtt.Magic):
 302                         # take care of MPEG PES timestamp overflow
 303                         if block.mpegts is None:
 304                             block.mpegts = 0
 305                         extra_state.setdefault('webvtt_mpegts_adjust', 0)
 306                         block.mpegts += extra_state['webvtt_mpegts_adjust'] << 33
 307                         if block.mpegts < extra_state.get('webvtt_mpegts_last', 0):
 308                             overflow = True
 309                             block.mpegts += 1 << 33
 310                         mpegts_last = block.mpegts
 311
 312                         if frag_index == 1:
 313                             extra_state['webvtt_mpegts'] = block.mpegts or 0
 314                             extra_state['webvtt_local'] = block.local or 0
 315                             # XXX: block.local = block.mpegts = None ?
 316                         else:
 317                             if block.mpegts is not None and block.local is not None:
 318                                 adjust = (
 319                                     (block.mpegts - extra_state.get('webvtt_mpegts', 0))
 320                                     - (block.local - extra_state.get('webvtt_local', 0))
 321                                 )
 322                             continue
 323                     elif isinstance(block, webvtt.HeaderBlock):
 324                         if frag_index != 1:
 325                             # XXX: this should probably be silent as well
 326                             # or verify that all segments contain the same data
 327                             self.report_warning(bug_reports_message(
 328                                 'Discarding a %s block found in the middle of the stream; '
 329                                 'if the subtitles display incorrectly,'
 330                                 % (type(block).__name__)))
 331                             continue
 332                     block.write_into(output)
 333
 334                 return output.getvalue().encode('utf-8')
 335
 336             def fin_fragments():
 337                 dedup_window = extra_state.get('webvtt_dedup_window')
 338                 if not dedup_window:
 339                     return b''
 340
 341                 output = io.StringIO()
 342                 for cue in dedup_window:
 343                     webvtt.CueBlock.from_json(cue).write_into(output)
 344
 345                 return output.getvalue().encode('utf-8')
 346
 347             self.download_and_append_fragments(
 348                 ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
 349         else:
 350             return self.download_and_append_fragments(ctx, fragments, info_dict)