1 from __future__
import unicode_literals
6 from Crypto
.Cipher
import AES
7 can_decrypt_frag
= True
9 can_decrypt_frag
= False
11 import concurrent
.futures
12 can_threaded_download
= True
14 can_threaded_download
= False
16 from ..downloader
import _get_real_downloader
17 from .fragment
import FragmentFD
18 from .external
import FFmpegFD
20 from ..compat
import (
26 parse_m3u8_attributes
,
32 class HlsFD(FragmentFD
):
34 Download segments in a m3u8 manifest. External downloaders can take over
35 the fragment downloads by supporting the 'frag_urls' protocol and
36 re-defining 'supports_manifest' function
42 def can_download(manifest
, info_dict
, allow_unplayable_formats
=False, with_crypto
=can_decrypt_frag
):
43 UNSUPPORTED_FEATURES
= [
44 # r'#EXT-X-BYTERANGE', # playlists composed of byte ranges of media files [2]
46 # Live streams heuristic does not always work (e.g. geo restricted to Germany
47 # 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)
48 # r'#EXT-X-MEDIA-SEQUENCE:(?!0$)', # live streams [3]
50 # This heuristic also is not correct since segments may not be appended as well.
51 # Twitch vods of finished streams have EXT-X-PLAYLIST-TYPE:EVENT despite
52 # no segments will definitely be appended to the end of the playlist.
53 # r'#EXT-X-PLAYLIST-TYPE:EVENT', # media segments may be appended to the end of
54 # # event media playlists [4]
55 # r'#EXT-X-MAP:', # media initialization [5]
56 # 1. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.4
57 # 2. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.2
58 # 3. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.2
59 # 4. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.3.5
60 # 5. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.5
62 if not allow_unplayable_formats
:
63 UNSUPPORTED_FEATURES
+= [
64 r
'#EXT-X-KEY:METHOD=(?!NONE|AES-128)', # encrypted streams [1]
68 yield not info_dict
.get('is_live')
69 is_aes128_enc
= '#EXT-X-KEY:METHOD=AES-128' in manifest
70 yield with_crypto
or not is_aes128_enc
71 yield not (is_aes128_enc
and r
'#EXT-X-BYTERANGE' in manifest
)
72 for feature
in UNSUPPORTED_FEATURES
:
73 yield not re
.search(feature
, manifest
)
74 return all(check_results())
76 def real_download(self
, filename
, info_dict
):
77 man_url
= info_dict
['url']
78 self
.to_screen('[%s] Downloading m3u8 manifest' % self
.FD_NAME
)
80 urlh
= self
.ydl
.urlopen(self
._prepare
_url
(info_dict
, man_url
))
81 man_url
= urlh
.geturl()
82 s
= urlh
.read().decode('utf-8', 'ignore')
84 if not self
.can_download(s
, info_dict
, self
.params
.get('allow_unplayable_formats')):
85 if info_dict
.get('extra_param_to_segment_url') or info_dict
.get('_decryption_key_url'):
86 self
.report_error('pycryptodome not found. Please install it.')
88 if self
.can_download(s
, info_dict
, with_crypto
=True):
89 self
.report_warning('pycryptodome is needed to download this file with hlsnative')
91 'hlsnative has detected features it does not support, '
92 'extraction will be delegated to ffmpeg')
93 fd
= FFmpegFD(self
.ydl
, self
.params
)
94 # TODO: Make progress updates work without hooking twice
95 # for ph in self._progress_hooks:
96 # fd.add_progress_hook(ph)
97 return fd
.real_download(filename
, info_dict
)
99 real_downloader
= _get_real_downloader(info_dict
, 'frag_urls', self
.params
, None)
100 if real_downloader
and not real_downloader
.supports_manifest(s
):
101 real_downloader
= None
103 def is_ad_fragment_start(s
):
104 return (s
.startswith('#ANVATO-SEGMENT-INFO') and 'type=ad' in s
105 or s
.startswith('#UPLYNK-SEGMENT') and s
.endswith(',ad'))
107 def is_ad_fragment_end(s
):
108 return (s
.startswith('#ANVATO-SEGMENT-INFO') and 'type=master' in s
109 or s
.startswith('#UPLYNK-SEGMENT') and s
.endswith(',segment'))
116 for line
in s
.splitlines():
120 if line
.startswith('#'):
121 if is_ad_fragment_start(line
):
123 elif is_ad_fragment_end(line
):
132 'filename': filename
,
133 'total_frags': media_frags
,
134 'ad_frags': ad_frags
,
138 self
._prepare
_external
_frag
_download
(ctx
)
140 self
._prepare
_and
_start
_frag
_download
(ctx
)
142 fragment_retries
= self
.params
.get('fragment_retries', 0)
143 skip_unavailable_fragments
= self
.params
.get('skip_unavailable_fragments', True)
144 test
= self
.params
.get('test', False)
146 format_index
= info_dict
.get('format_index')
148 extra_param_to_segment_url
= info_dict
.get('extra_param_to_segment_url')
149 if extra_param_to_segment_url
:
150 extra_query
= compat_urlparse
.parse_qs(extra_param_to_segment_url
)
153 decrypt_info
= {'METHOD': 'NONE'}
155 discontinuity_count
= 0
158 for line
in s
.splitlines():
161 if not line
.startswith('#'):
162 if format_index
and discontinuity_count
!= format_index
:
167 if frag_index
<= ctx
['fragment_index']:
171 if re
.match(r
'^https?://', line
)
172 else compat_urlparse
.urljoin(man_url
, line
))
174 frag_url
= update_url_query(frag_url
, extra_query
)
177 'frag_index': frag_index
,
179 'decrypt_info': decrypt_info
,
180 'byte_range': byte_range
,
181 'media_sequence': media_sequence
,
184 elif line
.startswith('#EXT-X-MAP'):
185 if format_index
and discontinuity_count
!= format_index
:
189 'initialization fragment found after media fragments, unable to download')
192 map_info
= parse_m3u8_attributes(line
[11:])
195 if re
.match(r
'^https?://', map_info
.get('URI'))
196 else compat_urlparse
.urljoin(man_url
, map_info
.get('URI')))
198 frag_url
= update_url_query(frag_url
, extra_query
)
201 'frag_index': frag_index
,
203 'decrypt_info': decrypt_info
,
204 'byte_range': byte_range
,
205 'media_sequence': media_sequence
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']
212 'start': sub_range_start
,
213 'end': sub_range_start
+ int(splitted_byte_range
[0]),
216 elif line
.startswith('#EXT-X-KEY'):
217 decrypt_url
= decrypt_info
.get('URI')
218 decrypt_info
= parse_m3u8_attributes(line
[11:])
219 if decrypt_info
['METHOD'] == 'AES-128':
220 if 'IV' in decrypt_info
:
221 decrypt_info
['IV'] = binascii
.unhexlify(decrypt_info
['IV'][2:].zfill(32))
222 if not re
.match(r
'^https?://', decrypt_info
['URI']):
223 decrypt_info
['URI'] = compat_urlparse
.urljoin(
224 man_url
, decrypt_info
['URI'])
226 decrypt_info
['URI'] = update_url_query(decrypt_info
['URI'], extra_query
)
227 if decrypt_url
!= decrypt_info
['URI']:
228 decrypt_info
['KEY'] = None
230 elif line
.startswith('#EXT-X-MEDIA-SEQUENCE'):
231 media_sequence
= int(line
[22:])
232 elif line
.startswith('#EXT-X-BYTERANGE'):
233 splitted_byte_range
= line
[17:].split('@')
234 sub_range_start
= int(splitted_byte_range
[1]) if len(splitted_byte_range
) == 2 else byte_range
['end']
236 'start': sub_range_start
,
237 'end': sub_range_start
+ int(splitted_byte_range
[0]),
239 elif is_ad_fragment_start(line
):
241 elif is_ad_fragment_end(line
):
243 elif line
.startswith('#EXT-X-DISCONTINUITY'):
244 discontinuity_count
+= 1
248 # We only download the first fragment during the test
250 fragments
= [fragments
[0] if fragments
else None]
253 info_copy
= info_dict
.copy()
254 info_copy
['fragments'] = fragments
255 fd
= real_downloader(self
.ydl
, self
.params
)
256 # TODO: Make progress updates work without hooking twice
257 # for ph in self._progress_hooks:
258 # fd.add_progress_hook(ph)
259 success
= fd
.real_download(filename
, info_copy
)
263 def download_fragment(fragment
):
264 frag_index
= fragment
['frag_index']
265 frag_url
= fragment
['url']
266 decrypt_info
= fragment
['decrypt_info']
267 byte_range
= fragment
['byte_range']
268 media_sequence
= fragment
['media_sequence']
270 ctx
['fragment_index'] = frag_index
273 headers
= info_dict
.get('http_headers', {})
275 headers
['Range'] = 'bytes=%d-%d' % (byte_range
['start'], byte_range
['end'] - 1)
276 while count
<= fragment_retries
:
278 success
, frag_content
= self
._download
_fragment
(
279 ctx
, frag_url
, info_dict
, headers
)
281 return False, frag_index
283 except compat_urllib_error
.HTTPError
as err
:
284 # Unavailable (possibly temporary) fragments may be served.
285 # First we try to retry then either skip or abort.
286 # See https://github.com/ytdl-org/youtube-dl/issues/10165,
287 # https://github.com/ytdl-org/youtube-dl/issues/10448).
289 if count
<= fragment_retries
:
290 self
.report_retry_fragment(err
, frag_index
, count
, fragment_retries
)
291 if count
> fragment_retries
:
292 return False, frag_index
294 if decrypt_info
['METHOD'] == 'AES-128':
295 iv
= decrypt_info
.get('IV') or compat_struct_pack('>8xq', media_sequence
)
296 decrypt_info
['KEY'] = decrypt_info
.get('KEY') or self
.ydl
.urlopen(
297 self
._prepare
_url
(info_dict
, info_dict
.get('_decryption_key_url') or decrypt_info
['URI'])).read()
298 # Don't decrypt the content in tests since the data is explicitly truncated and it's not to a valid block
299 # size (see https://github.com/ytdl-org/youtube-dl/pull/27660). Tests only care that the correct data downloaded,
300 # not what it decrypts to.
302 frag_content
= AES
.new(
303 decrypt_info
['KEY'], AES
.MODE_CBC
, iv
).decrypt(frag_content
)
305 return frag_content
, frag_index
307 def append_fragment(frag_content
, frag_index
):
309 fragment_filename
= '%s-Frag%d' % (ctx
['tmpfilename'], frag_index
)
311 file, frag_sanitized
= sanitize_open(fragment_filename
, 'rb')
312 ctx
['fragment_filename_sanitized'] = frag_sanitized
314 self
._append
_fragment
(ctx
, frag_content
)
316 except FileNotFoundError
:
317 if skip_unavailable_fragments
:
318 self
.report_skip_fragment(frag_index
)
322 'fragment %s not found, unable to continue' % frag_index
)
325 if skip_unavailable_fragments
:
326 self
.report_skip_fragment(frag_index
)
330 'fragment %s not found, unable to continue' % frag_index
)
333 max_workers
= self
.params
.get('concurrent_fragment_downloads', 1)
334 if can_threaded_download
and max_workers
> 1:
335 self
.report_warning('The download speed shown is only of one thread. This is a known issue')
336 with concurrent
.futures
.ThreadPoolExecutor(max_workers
) as pool
:
337 futures
= [pool
.submit(download_fragment
, fragment
) for fragment
in fragments
]
338 # timeout must be 0 to return instantly
339 done
, not_done
= concurrent
.futures
.wait(futures
, timeout
=0)
342 # Check every 1 second for KeyboardInterrupt
343 freshly_done
, not_done
= concurrent
.futures
.wait(not_done
, timeout
=1)
345 except KeyboardInterrupt:
346 for future
in not_done
:
348 # timeout must be none to cancel
349 concurrent
.futures
.wait(not_done
, timeout
=None)
350 raise KeyboardInterrupt
351 results
= [future
.result() for future
in futures
]
353 for frag_content
, frag_index
in results
:
354 result
= append_fragment(frag_content
, frag_index
)
358 for fragment
in fragments
:
359 frag_content
, frag_index
= download_fragment(fragment
)
360 result
= append_fragment(frag_content
, frag_index
)
364 self
._finish
_frag
_download
(ctx
)