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