]> jfr.im git - yt-dlp.git/blob - yt_dlp/downloader/hls.py
f2868dc52bb10c6b5ad64a3fbf50266a8e00a89e
[yt-dlp.git] / 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 can_download(manifest, info_dict, allow_unplayable_formats=False):
32 UNSUPPORTED_FEATURES = [
33 # r'#EXT-X-BYTERANGE', # playlists composed of byte ranges of media files [2]
34
35 # Live streams heuristic does not always work (e.g. geo restricted to Germany
36 # 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)
37 # r'#EXT-X-MEDIA-SEQUENCE:(?!0$)', # live streams [3]
38
39 # This heuristic also is not correct since segments may not be appended as well.
40 # Twitch vods of finished streams have EXT-X-PLAYLIST-TYPE:EVENT despite
41 # no segments will definitely be appended to the end of the playlist.
42 # r'#EXT-X-PLAYLIST-TYPE:EVENT', # media segments may be appended to the end of
43 # # event media playlists [4]
44 # r'#EXT-X-MAP:', # media initialization [5]
45 # 1. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.4
46 # 2. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.2
47 # 3. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.2
48 # 4. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.5
49 # 5. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.5
50 ]
51 if not allow_unplayable_formats:
52 UNSUPPORTED_FEATURES += [
53 r'#EXT-X-KEY:METHOD=(?!NONE|AES-128)', # encrypted streams [1]
54 ]
55
56 def check_results():
57 yield not info_dict.get('is_live')
58 for feature in UNSUPPORTED_FEATURES:
59 yield not re.search(feature, manifest)
60 return all(check_results())
61
62 def real_download(self, filename, info_dict):
63 man_url = info_dict['url']
64 self.to_screen('[%s] Downloading m3u8 manifest' % self.FD_NAME)
65
66 urlh = self.ydl.urlopen(self._prepare_url(info_dict, man_url))
67 man_url = urlh.geturl()
68 s = urlh.read().decode('utf-8', 'ignore')
69
70 can_download, message = self.can_download(s, info_dict, self.params.get('allow_unplayable_formats')), None
71 if can_download:
72 has_ffmpeg = FFmpegFD.available()
73 no_crypto = not Cryptodome.AES and '#EXT-X-KEY:METHOD=AES-128' in s
74 if no_crypto and has_ffmpeg:
75 can_download, message = False, 'The stream has AES-128 encryption and pycryptodomex is not available'
76 elif no_crypto:
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 elif info_dict.get('extractor_key') == 'Generic' and re.search(r'(?m)#EXT-X-MEDIA-SEQUENCE:(?!0$)', s):
80 install_ffmpeg = '' if has_ffmpeg else 'install ffmpeg and '
81 message = ('Live HLS streams are not supported by the native downloader. If this is a livestream, '
82 f'please {install_ffmpeg}add "--downloader ffmpeg --hls-use-mpegts" to your command')
83 if not can_download:
84 has_drm = re.search('|'.join([
85 r'#EXT-X-FAXS-CM:', # Adobe Flash Access
86 r'#EXT-X-(?:SESSION-)?KEY:.*?URI="skd://', # Apple FairPlay
87 ]), s)
88 if has_drm and not self.params.get('allow_unplayable_formats'):
89 self.report_error(
90 'This video is DRM protected; Try selecting another format with --format or '
91 'add --check-formats to automatically fallback to the next best format')
92 return False
93 message = message or 'Unsupported features have been detected'
94 fd = FFmpegFD(self.ydl, self.params)
95 self.report_warning(f'{message}; extraction will be delegated to {fd.get_basename()}')
96 return fd.real_download(filename, info_dict)
97 elif message:
98 self.report_warning(message)
99
100 is_webvtt = info_dict['ext'] == 'vtt'
101 if is_webvtt:
102 real_downloader = None # Packing the fragments is not currently supported for external downloader
103 else:
104 real_downloader = get_suitable_downloader(
105 info_dict, self.params, None, protocol='m3u8_frag_urls', to_stdout=(filename == '-'))
106 if real_downloader and not real_downloader.supports_manifest(s):
107 real_downloader = None
108 if real_downloader:
109 self.to_screen(f'[{self.FD_NAME}] Fragment downloads will be delegated to {real_downloader.get_basename()}')
110
111 def is_ad_fragment_start(s):
112 return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
113 or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',ad'))
114
115 def is_ad_fragment_end(s):
116 return (s.startswith('#ANVATO-SEGMENT-INFO') and 'type=master' in s
117 or s.startswith('#UPLYNK-SEGMENT') and s.endswith(',segment'))
118
119 fragments = []
120
121 media_frags = 0
122 ad_frags = 0
123 ad_frag_next = False
124 for line in s.splitlines():
125 line = line.strip()
126 if not line:
127 continue
128 if line.startswith('#'):
129 if is_ad_fragment_start(line):
130 ad_frag_next = True
131 elif is_ad_fragment_end(line):
132 ad_frag_next = False
133 continue
134 if ad_frag_next:
135 ad_frags += 1
136 continue
137 media_frags += 1
138
139 ctx = {
140 'filename': filename,
141 'total_frags': media_frags,
142 'ad_frags': ad_frags,
143 }
144
145 if real_downloader:
146 self._prepare_external_frag_download(ctx)
147 else:
148 self._prepare_and_start_frag_download(ctx, info_dict)
149
150 extra_state = ctx.setdefault('extra_state', {})
151
152 format_index = info_dict.get('format_index')
153 extra_query = None
154 extra_param_to_segment_url = info_dict.get('extra_param_to_segment_url')
155 if extra_param_to_segment_url:
156 extra_query = urllib.parse.parse_qs(extra_param_to_segment_url)
157 i = 0
158 media_sequence = 0
159 decrypt_info = {'METHOD': 'NONE'}
160 external_aes_key = traverse_obj(info_dict, ('hls_aes', 'key'))
161 if external_aes_key:
162 external_aes_key = binascii.unhexlify(remove_start(external_aes_key, '0x'))
163 assert len(external_aes_key) in (16, 24, 32), 'Invalid length for HLS AES-128 key'
164 external_aes_iv = traverse_obj(info_dict, ('hls_aes', 'iv'))
165 if external_aes_iv:
166 external_aes_iv = binascii.unhexlify(remove_start(external_aes_iv, '0x').zfill(32))
167 byte_range = {}
168 discontinuity_count = 0
169 frag_index = 0
170 ad_frag_next = False
171 for line in s.splitlines():
172 line = line.strip()
173 if line:
174 if not line.startswith('#'):
175 if format_index and discontinuity_count != format_index:
176 continue
177 if ad_frag_next:
178 continue
179 frag_index += 1
180 if frag_index <= ctx['fragment_index']:
181 continue
182 frag_url = urljoin(man_url, line)
183 if extra_query:
184 frag_url = update_url_query(frag_url, extra_query)
185
186 fragments.append({
187 'frag_index': frag_index,
188 'url': frag_url,
189 'decrypt_info': decrypt_info,
190 'byte_range': byte_range,
191 'media_sequence': media_sequence,
192 })
193 media_sequence += 1
194
195 elif line.startswith('#EXT-X-MAP'):
196 if format_index and discontinuity_count != format_index:
197 continue
198 if frag_index > 0:
199 self.report_error(
200 'Initialization fragment found after media fragments, unable to download')
201 return False
202 frag_index += 1
203 map_info = parse_m3u8_attributes(line[11:])
204 frag_url = urljoin(man_url, map_info.get('URI'))
205 if extra_query:
206 frag_url = update_url_query(frag_url, extra_query)
207
208 if map_info.get('BYTERANGE'):
209 splitted_byte_range = map_info.get('BYTERANGE').split('@')
210 sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
211 byte_range = {
212 'start': sub_range_start,
213 'end': sub_range_start + int(splitted_byte_range[0]),
214 }
215
216 fragments.append({
217 'frag_index': frag_index,
218 'url': frag_url,
219 'decrypt_info': decrypt_info,
220 'byte_range': byte_range,
221 'media_sequence': media_sequence
222 })
223 media_sequence += 1
224
225 elif line.startswith('#EXT-X-KEY'):
226 decrypt_url = decrypt_info.get('URI')
227 decrypt_info = parse_m3u8_attributes(line[11:])
228 if decrypt_info['METHOD'] == 'AES-128':
229 if external_aes_iv:
230 decrypt_info['IV'] = external_aes_iv
231 elif 'IV' in decrypt_info:
232 decrypt_info['IV'] = binascii.unhexlify(decrypt_info['IV'][2:].zfill(32))
233 if external_aes_key:
234 decrypt_info['KEY'] = external_aes_key
235 else:
236 decrypt_info['URI'] = urljoin(man_url, decrypt_info['URI'])
237 if extra_query:
238 decrypt_info['URI'] = update_url_query(decrypt_info['URI'], extra_query)
239 if decrypt_url != decrypt_info['URI']:
240 decrypt_info['KEY'] = None
241
242 elif line.startswith('#EXT-X-MEDIA-SEQUENCE'):
243 media_sequence = int(line[22:])
244 elif line.startswith('#EXT-X-BYTERANGE'):
245 splitted_byte_range = line[17:].split('@')
246 sub_range_start = int(splitted_byte_range[1]) if len(splitted_byte_range) == 2 else byte_range['end']
247 byte_range = {
248 'start': sub_range_start,
249 'end': sub_range_start + int(splitted_byte_range[0]),
250 }
251 elif is_ad_fragment_start(line):
252 ad_frag_next = True
253 elif is_ad_fragment_end(line):
254 ad_frag_next = False
255 elif line.startswith('#EXT-X-DISCONTINUITY'):
256 discontinuity_count += 1
257 i += 1
258
259 # We only download the first fragment during the test
260 if self.params.get('test', False):
261 fragments = [fragments[0] if fragments else None]
262
263 if real_downloader:
264 info_dict['fragments'] = fragments
265 fd = real_downloader(self.ydl, self.params)
266 # TODO: Make progress updates work without hooking twice
267 # for ph in self._progress_hooks:
268 # fd.add_progress_hook(ph)
269 return fd.real_download(filename, info_dict)
270
271 if is_webvtt:
272 def pack_fragment(frag_content, frag_index):
273 output = io.StringIO()
274 adjust = 0
275 overflow = False
276 mpegts_last = None
277 for block in webvtt.parse_fragment(frag_content):
278 if isinstance(block, webvtt.CueBlock):
279 extra_state['webvtt_mpegts_last'] = mpegts_last
280 if overflow:
281 extra_state['webvtt_mpegts_adjust'] += 1
282 overflow = False
283 block.start += adjust
284 block.end += adjust
285
286 dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
287
288 ready = []
289
290 i = 0
291 is_new = True
292 while i < len(dedup_window):
293 wcue = dedup_window[i]
294 wblock = webvtt.CueBlock.from_json(wcue)
295 i += 1
296 if wblock.hinges(block):
297 wcue['end'] = block.end
298 is_new = False
299 continue
300 if wblock == block:
301 is_new = False
302 continue
303 if wblock.end > block.start:
304 continue
305 ready.append(wblock)
306 i -= 1
307 del dedup_window[i]
308
309 if is_new:
310 dedup_window.append(block.as_json)
311 for block in ready:
312 block.write_into(output)
313
314 # we only emit cues once they fall out of the duplicate window
315 continue
316 elif isinstance(block, webvtt.Magic):
317 # take care of MPEG PES timestamp overflow
318 if block.mpegts is None:
319 block.mpegts = 0
320 extra_state.setdefault('webvtt_mpegts_adjust', 0)
321 block.mpegts += extra_state['webvtt_mpegts_adjust'] << 33
322 if block.mpegts < extra_state.get('webvtt_mpegts_last', 0):
323 overflow = True
324 block.mpegts += 1 << 33
325 mpegts_last = block.mpegts
326
327 if frag_index == 1:
328 extra_state['webvtt_mpegts'] = block.mpegts or 0
329 extra_state['webvtt_local'] = block.local or 0
330 # XXX: block.local = block.mpegts = None ?
331 else:
332 if block.mpegts is not None and block.local is not None:
333 adjust = (
334 (block.mpegts - extra_state.get('webvtt_mpegts', 0))
335 - (block.local - extra_state.get('webvtt_local', 0))
336 )
337 continue
338 elif isinstance(block, webvtt.HeaderBlock):
339 if frag_index != 1:
340 # XXX: this should probably be silent as well
341 # or verify that all segments contain the same data
342 self.report_warning(bug_reports_message(
343 'Discarding a %s block found in the middle of the stream; '
344 'if the subtitles display incorrectly,'
345 % (type(block).__name__)))
346 continue
347 block.write_into(output)
348
349 return output.getvalue().encode()
350
351 def fin_fragments():
352 dedup_window = extra_state.get('webvtt_dedup_window')
353 if not dedup_window:
354 return b''
355
356 output = io.StringIO()
357 for cue in dedup_window:
358 webvtt.CueBlock.from_json(cue).write_into(output)
359
360 return output.getvalue().encode()
361
362 self.download_and_append_fragments(
363 ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
364 else:
365 return self.download_and_append_fragments(ctx, fragments, info_dict)