1 import concurrent
.futures
11 from .common
import FileDownloader
12 from .http
import HttpFD
13 from ..aes
import aes_cbc_decrypt_bytes
, unpad_pkcs7
14 from ..compat
import compat_os_name
24 class HttpQuietDownloader(HttpFD
):
25 def to_screen(self
, *args
, **kargs
):
28 to_console_title
= to_screen
31 class FragmentFD(FileDownloader
):
33 A base file downloader class for fragmented media (e.g. f4m/m3u8 manifests).
37 fragment_retries: Number of times to retry a fragment for HTTP error
38 (DASH and hlsnative only). Default is 0 for API, but 10 for CLI
39 skip_unavailable_fragments:
40 Skip unavailable fragments (DASH and hlsnative only)
41 keep_fragments: Keep downloaded fragments on disk after downloading is
43 concurrent_fragment_downloads: The number of threads to use for native hls and dash downloads
44 _no_ytdl_file: Don't use .ytdl file
46 For each incomplete fragment download yt-dlp keeps on disk a special
47 bookkeeping file with download state and metadata (in future such files will
48 be used for any incomplete download handled by yt-dlp). This file is
49 used to properly handle resuming, check download file consistency and detect
50 potential errors. The file has a .ytdl extension and represents a standard
51 JSON file of the following format:
54 Dictionary of extractor related data. TBD.
57 Dictionary of downloader related data. May contain following data:
59 Dictionary with current (being downloaded) fragment data:
60 index: 0-based index of current fragment among all fragments
62 Total count of fragments
64 This feature is experimental and file format may change in future.
67 def report_retry_fragment(self
, err
, frag_index
, count
, retries
):
68 self
.deprecation_warning('yt_dlp.downloader.FragmentFD.report_retry_fragment is deprecated. '
69 'Use yt_dlp.downloader.FileDownloader.report_retry instead')
70 return self
.report_retry(err
, count
, retries
, frag_index
)
72 def report_skip_fragment(self
, frag_index
, err
=None):
73 err
= f
' {err};' if err
else ''
74 self
.to_screen(f
'[download]{err} Skipping fragment {frag_index:d} ...')
76 def _prepare_url(self
, info_dict
, url
):
77 headers
= info_dict
.get('http_headers')
78 return sanitized_Request(url
, None, headers
) if headers
else url
80 def _prepare_and_start_frag_download(self
, ctx
, info_dict
):
81 self
._prepare
_frag
_download
(ctx
)
82 self
._start
_frag
_download
(ctx
, info_dict
)
84 def __do_ytdl_file(self
, ctx
):
85 return ctx
['live'] is not True and ctx
['tmpfilename'] != '-' and not self
.params
.get('_no_ytdl_file')
87 def _read_ytdl_file(self
, ctx
):
88 assert 'ytdl_corrupt' not in ctx
89 stream
, _
= self
.sanitize_open(self
.ytdl_filename(ctx
['filename']), 'r')
91 ytdl_data
= json
.loads(stream
.read())
92 ctx
['fragment_index'] = ytdl_data
['downloader']['current_fragment']['index']
93 if 'extra_state' in ytdl_data
['downloader']:
94 ctx
['extra_state'] = ytdl_data
['downloader']['extra_state']
96 ctx
['ytdl_corrupt'] = True
100 def _write_ytdl_file(self
, ctx
):
101 frag_index_stream
, _
= self
.sanitize_open(self
.ytdl_filename(ctx
['filename']), 'w')
104 'current_fragment': {
105 'index': ctx
['fragment_index'],
108 if 'extra_state' in ctx
:
109 downloader
['extra_state'] = ctx
['extra_state']
110 if ctx
.get('fragment_count') is not None:
111 downloader
['fragment_count'] = ctx
['fragment_count']
112 frag_index_stream
.write(json
.dumps({'downloader': downloader}
))
114 frag_index_stream
.close()
116 def _download_fragment(self
, ctx
, frag_url
, info_dict
, headers
=None, request_data
=None):
117 fragment_filename
= '%s-Frag%d' % (ctx
['tmpfilename'], ctx
['fragment_index'])
118 fragment_info_dict
= {
120 'http_headers': headers
or info_dict
.get('http_headers'),
121 'request_data': request_data
,
122 'ctx_id': ctx
.get('ctx_id'),
125 if ctx
['dl'].params
.get('continuedl', True):
126 frag_resume_len
= self
.filesize_or_none(self
.temp_name(fragment_filename
))
127 fragment_info_dict
['frag_resume_len'] = ctx
['frag_resume_len'] = frag_resume_len
129 success
, _
= ctx
['dl'].download(fragment_filename
, fragment_info_dict
)
132 if fragment_info_dict
.get('filetime'):
133 ctx
['fragment_filetime'] = fragment_info_dict
.get('filetime')
134 ctx
['fragment_filename_sanitized'] = fragment_filename
137 def _read_fragment(self
, ctx
):
138 if not ctx
.get('fragment_filename_sanitized'):
141 down
, frag_sanitized
= self
.sanitize_open(ctx
['fragment_filename_sanitized'], 'rb')
142 except FileNotFoundError
:
146 ctx
['fragment_filename_sanitized'] = frag_sanitized
147 frag_content
= down
.read()
151 def _append_fragment(self
, ctx
, frag_content
):
153 ctx
['dest_stream'].write(frag_content
)
154 ctx
['dest_stream'].flush()
156 if self
.__do
_ytdl
_file
(ctx
):
157 self
._write
_ytdl
_file
(ctx
)
158 if not self
.params
.get('keep_fragments', False):
159 self
.try_remove(encodeFilename(ctx
['fragment_filename_sanitized']))
160 del ctx
['fragment_filename_sanitized']
162 def _prepare_frag_download(self
, ctx
):
163 if not ctx
.setdefault('live', False):
164 total_frags_str
= '%d' % ctx
['total_frags']
165 ad_frags
= ctx
.get('ad_frags', 0)
167 total_frags_str
+= ' (not including %d ad)' % ad_frags
169 total_frags_str
= 'unknown (live)'
170 self
.to_screen(f
'[{self.FD_NAME}] Total fragments: {total_frags_str}')
171 self
.report_destination(ctx
['filename'])
172 dl
= HttpQuietDownloader(self
.ydl
, {
177 tmpfilename
= self
.temp_name(ctx
['filename'])
180 # Establish possible resume length
181 resume_len
= self
.filesize_or_none(tmpfilename
)
185 # Should be initialized before ytdl file check
187 'tmpfilename': tmpfilename
,
191 if self
.__do
_ytdl
_file
(ctx
):
192 ytdl_file_exists
= os
.path
.isfile(encodeFilename(self
.ytdl_filename(ctx
['filename'])))
193 continuedl
= self
.params
.get('continuedl', True)
194 if continuedl
and ytdl_file_exists
:
195 self
._read
_ytdl
_file
(ctx
)
196 is_corrupt
= ctx
.get('ytdl_corrupt') is True
197 is_inconsistent
= ctx
['fragment_index'] > 0 and resume_len
== 0
198 if is_corrupt
or is_inconsistent
:
200 '.ytdl file is corrupt' if is_corrupt
else
201 'Inconsistent state of incomplete fragment download')
203 '%s. Restarting from the beginning ...' % message
)
204 ctx
['fragment_index'] = resume_len
= 0
205 if 'ytdl_corrupt' in ctx
:
206 del ctx
['ytdl_corrupt']
207 self
._write
_ytdl
_file
(ctx
)
212 self
._read
_ytdl
_file
(ctx
)
213 ctx
['fragment_index'] = resume_len
= 0
214 self
._write
_ytdl
_file
(ctx
)
215 assert ctx
['fragment_index'] == 0
217 dest_stream
, tmpfilename
= self
.sanitize_open(tmpfilename
, open_mode
)
221 'dest_stream': dest_stream
,
222 'tmpfilename': tmpfilename
,
223 # Total complete fragments downloaded so far in bytes
224 'complete_frags_downloaded_bytes': resume_len
,
227 def _start_frag_download(self
, ctx
, info_dict
):
228 resume_len
= ctx
['complete_frags_downloaded_bytes']
229 total_frags
= ctx
['total_frags']
230 ctx_id
= ctx
.get('ctx_id')
231 # This dict stores the download progress, it's updated by the progress
234 'status': 'downloading',
235 'downloaded_bytes': resume_len
,
236 'fragment_index': ctx
['fragment_index'],
237 'fragment_count': total_frags
,
238 'filename': ctx
['filename'],
239 'tmpfilename': ctx
['tmpfilename'],
245 'fragment_started': start
,
246 # Amount of fragment's bytes downloaded by the time of the previous
247 # frag progress hook invocation
248 'prev_frag_downloaded_bytes': 0,
251 def frag_progress_hook(s
):
252 if s
['status'] not in ('downloading', 'finished'):
255 if not total_frags
and ctx
.get('fragment_count'):
256 state
['fragment_count'] = ctx
['fragment_count']
258 if ctx_id
is not None and s
.get('ctx_id') != ctx_id
:
261 state
['max_progress'] = ctx
.get('max_progress')
262 state
['progress_idx'] = ctx
.get('progress_idx')
264 time_now
= time
.time()
265 state
['elapsed'] = time_now
- start
266 frag_total_bytes
= s
.get('total_bytes') or 0
267 s
['fragment_info_dict'] = s
.pop('info_dict', {})
270 (ctx
['complete_frags_downloaded_bytes'] + frag_total_bytes
)
271 / (state
['fragment_index'] + 1) * total_frags
)
272 state
['total_bytes_estimate'] = estimated_size
274 if s
['status'] == 'finished':
275 state
['fragment_index'] += 1
276 ctx
['fragment_index'] = state
['fragment_index']
277 state
['downloaded_bytes'] += frag_total_bytes
- ctx
['prev_frag_downloaded_bytes']
278 ctx
['complete_frags_downloaded_bytes'] = state
['downloaded_bytes']
279 ctx
['speed'] = state
['speed'] = self
.calc_speed(
280 ctx
['fragment_started'], time_now
, frag_total_bytes
)
281 ctx
['fragment_started'] = time
.time()
282 ctx
['prev_frag_downloaded_bytes'] = 0
284 frag_downloaded_bytes
= s
['downloaded_bytes']
285 state
['downloaded_bytes'] += frag_downloaded_bytes
- ctx
['prev_frag_downloaded_bytes']
286 ctx
['speed'] = state
['speed'] = self
.calc_speed(
287 ctx
['fragment_started'], time_now
, frag_downloaded_bytes
- ctx
.get('frag_resume_len', 0))
289 state
['eta'] = self
.calc_eta(state
['speed'], estimated_size
- state
['downloaded_bytes'])
290 ctx
['prev_frag_downloaded_bytes'] = frag_downloaded_bytes
291 self
._hook
_progress
(state
, info_dict
)
293 ctx
['dl'].add_progress_hook(frag_progress_hook
)
297 def _finish_frag_download(self
, ctx
, info_dict
):
298 ctx
['dest_stream'].close()
299 if self
.__do
_ytdl
_file
(ctx
):
300 ytdl_filename
= encodeFilename(self
.ytdl_filename(ctx
['filename']))
301 if os
.path
.isfile(ytdl_filename
):
302 self
.try_remove(ytdl_filename
)
303 elapsed
= time
.time() - ctx
['started']
305 to_file
= ctx
['tmpfilename'] != '-'
307 downloaded_bytes
= self
.filesize_or_none(ctx
['tmpfilename'])
309 downloaded_bytes
= ctx
['complete_frags_downloaded_bytes']
311 if not downloaded_bytes
:
313 self
.try_remove(ctx
['tmpfilename'])
314 self
.report_error('The downloaded file is empty')
317 self
.try_rename(ctx
['tmpfilename'], ctx
['filename'])
318 filetime
= ctx
.get('fragment_filetime')
319 if self
.params
.get('updatetime', True) and filetime
:
320 with contextlib
.suppress(Exception):
321 os
.utime(ctx
['filename'], (time
.time(), filetime
))
323 self
._hook
_progress
({
324 'downloaded_bytes': downloaded_bytes
,
325 'total_bytes': downloaded_bytes
,
326 'filename': ctx
['filename'],
327 'status': 'finished',
329 'ctx_id': ctx
.get('ctx_id'),
330 'max_progress': ctx
.get('max_progress'),
331 'progress_idx': ctx
.get('progress_idx'),
335 def _prepare_external_frag_download(self
, ctx
):
336 if 'live' not in ctx
:
339 total_frags_str
= '%d' % ctx
['total_frags']
340 ad_frags
= ctx
.get('ad_frags', 0)
342 total_frags_str
+= ' (not including %d ad)' % ad_frags
344 total_frags_str
= 'unknown (live)'
345 self
.to_screen(f
'[{self.FD_NAME}] Total fragments: {total_frags_str}')
347 tmpfilename
= self
.temp_name(ctx
['filename'])
349 # Should be initialized before ytdl file check
351 'tmpfilename': tmpfilename
,
355 def decrypter(self
, info_dict
):
359 if url
not in _key_cache
:
360 _key_cache
[url
] = self
.ydl
.urlopen(self
._prepare
_url
(info_dict
, url
)).read()
361 return _key_cache
[url
]
363 def decrypt_fragment(fragment
, frag_content
):
364 if frag_content
is None:
366 decrypt_info
= fragment
.get('decrypt_info')
367 if not decrypt_info
or decrypt_info
['METHOD'] != 'AES-128':
369 iv
= decrypt_info
.get('IV') or struct
.pack('>8xq', fragment
['media_sequence'])
370 decrypt_info
['KEY'] = (decrypt_info
.get('KEY')
371 or _get_key(traverse_obj(info_dict
, ('hls_aes', 'uri')) or decrypt_info
['URI']))
372 # Don't decrypt the content in tests since the data is explicitly truncated and it's not to a valid block
373 # size (see https://github.com/ytdl-org/youtube-dl/pull/27660). Tests only care that the correct data downloaded,
374 # not what it decrypts to.
375 if self
.params
.get('test', False):
377 return unpad_pkcs7(aes_cbc_decrypt_bytes(frag_content
, decrypt_info
['KEY'], iv
))
379 return decrypt_fragment
381 def download_and_append_fragments_multiple(self
, *args
, **kwargs
):
383 @params (ctx1, fragments1, info_dict1), (ctx2, fragments2, info_dict2), ...
384 all args must be either tuple or list
386 interrupt_trigger
= [True]
387 max_progress
= len(args
)
388 if max_progress
== 1:
389 return self
.download_and_append_fragments(*args
[0], **kwargs
)
390 max_workers
= self
.params
.get('concurrent_fragment_downloads', 1)
392 self
._prepare
_multiline
_status
(max_progress
)
393 is_live
= any(traverse_obj(args
, (..., 2, 'is_live')))
395 def thread_func(idx
, ctx
, fragments
, info_dict
, tpe
):
396 ctx
['max_progress'] = max_progress
397 ctx
['progress_idx'] = idx
398 return self
.download_and_append_fragments(
399 ctx
, fragments
, info_dict
, **kwargs
, tpe
=tpe
, interrupt_trigger
=interrupt_trigger
)
401 class FTPE(concurrent
.futures
.ThreadPoolExecutor
):
402 # has to stop this or it's going to wait on the worker thread itself
403 def __exit__(self
, exc_type
, exc_val
, exc_tb
):
406 if compat_os_name
== 'nt':
407 def future_result(future
):
410 return future
.result(0.1)
411 except KeyboardInterrupt:
413 except concurrent
.futures
.TimeoutError
:
416 def future_result(future
):
417 return future
.result()
419 def interrupt_trigger_iter(fg
):
421 if not interrupt_trigger
[0]:
426 for idx
, (ctx
, fragments
, info_dict
) in enumerate(args
):
427 tpe
= FTPE(math
.ceil(max_workers
/ max_progress
))
428 job
= tpe
.submit(thread_func
, idx
, ctx
, interrupt_trigger_iter(fragments
), info_dict
, tpe
)
429 spins
.append((tpe
, job
))
432 for tpe
, job
in spins
:
434 result
= result
and future_result(job
)
435 except KeyboardInterrupt:
436 interrupt_trigger
[0] = False
438 tpe
.shutdown(wait
=True)
439 if not interrupt_trigger
[0] and not is_live
:
440 raise KeyboardInterrupt()
441 # we expect the user wants to stop and DO WANT the preceding postprocessors to run;
442 # so returning a intermediate result here instead of KeyboardInterrupt on live
445 def download_and_append_fragments(
446 self
, ctx
, fragments
, info_dict
, *, is_fatal
=(lambda idx
: False),
447 pack_func
=(lambda content
, idx
: content
), finish_func
=None,
448 tpe
=None, interrupt_trigger
=(True, )):
450 if not self
.params
.get('skip_unavailable_fragments', True):
451 is_fatal
= lambda _
: True
453 def download_fragment(fragment
, ctx
):
454 if not interrupt_trigger
[0]:
457 frag_index
= ctx
['fragment_index'] = fragment
['frag_index']
458 ctx
['last_error'] = None
459 headers
= info_dict
.get('http_headers', {}).copy()
460 byte_range
= fragment
.get('byte_range')
462 headers
['Range'] = 'bytes=%d-%d' % (byte_range
['start'], byte_range
['end'] - 1)
464 # Never skip the first fragment
465 fatal
= is_fatal(fragment
.get('index') or (frag_index
- 1))
467 def error_callback(err
, count
, retries
):
468 if fatal
and count
> retries
:
469 ctx
['dest_stream'].close()
470 self
.report_retry(err
, count
, retries
, frag_index
, fatal
)
471 ctx
['last_error'] = err
473 for retry
in RetryManager(self
.params
.get('fragment_retries'), error_callback
):
475 ctx
['fragment_count'] = fragment
.get('fragment_count')
476 if not self
._download
_fragment
(
477 ctx
, fragment
['url'], info_dict
, headers
, info_dict
.get('request_data')):
479 except (urllib
.error
.HTTPError
, http
.client
.IncompleteRead
) as err
:
482 except DownloadError
: # has own retry settings
486 def append_fragment(frag_content
, frag_index
, ctx
):
488 self
._append
_fragment
(ctx
, pack_func(frag_content
, frag_index
))
489 elif not is_fatal(frag_index
- 1):
490 self
.report_skip_fragment(frag_index
, 'fragment not found')
492 ctx
['dest_stream'].close()
493 self
.report_error(f
'fragment {frag_index} not found, unable to continue')
497 decrypt_fragment
= self
.decrypter(info_dict
)
499 max_workers
= math
.ceil(
500 self
.params
.get('concurrent_fragment_downloads', 1) / ctx
.get('max_progress', 1))
502 def _download_fragment(fragment
):
503 ctx_copy
= ctx
.copy()
504 download_fragment(fragment
, ctx_copy
)
505 return fragment
, fragment
['frag_index'], ctx_copy
.get('fragment_filename_sanitized')
507 self
.report_warning('The download speed shown is only of one thread. This is a known issue')
508 with tpe
or concurrent
.futures
.ThreadPoolExecutor(max_workers
) as pool
:
510 for fragment
, frag_index
, frag_filename
in pool
.map(_download_fragment
, fragments
):
512 'fragment_filename_sanitized': frag_filename
,
513 'fragment_index': frag_index
,
515 if not append_fragment(decrypt_fragment(fragment
, self
._read
_fragment
(ctx
)), frag_index
, ctx
):
517 except KeyboardInterrupt:
518 self
._finish
_multiline
_status
()
520 'Interrupted by user. Waiting for all threads to shutdown...', is_error
=False, tb
=False)
521 pool
.shutdown(wait
=False)
524 for fragment
in fragments
:
525 if not interrupt_trigger
[0]:
528 download_fragment(fragment
, ctx
)
529 result
= append_fragment(
530 decrypt_fragment(fragment
, self
._read
_fragment
(ctx
)), fragment
['frag_index'], ctx
)
531 except KeyboardInterrupt:
532 if info_dict
.get('is_live'):
538 if finish_func
is not None:
539 ctx
['dest_stream'].write(finish_func())
540 ctx
['dest_stream'].flush()
541 return self
._finish
_frag
_download
(ctx
, info_dict
)