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 (DASH
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'),
124 success
, _
= ctx
['dl'].download(fragment_filename
, fragment_info_dict
)
127 if fragment_info_dict
.get('filetime'):
128 ctx
['fragment_filetime'] = fragment_info_dict
.get('filetime')
129 ctx
['fragment_filename_sanitized'] = fragment_filename
132 def _read_fragment(self
, ctx
):
133 if not ctx
.get('fragment_filename_sanitized'):
136 down
, frag_sanitized
= self
.sanitize_open(ctx
['fragment_filename_sanitized'], 'rb')
137 except FileNotFoundError
:
141 ctx
['fragment_filename_sanitized'] = frag_sanitized
142 frag_content
= down
.read()
146 def _append_fragment(self
, ctx
, frag_content
):
148 ctx
['dest_stream'].write(frag_content
)
149 ctx
['dest_stream'].flush()
151 if self
.__do
_ytdl
_file
(ctx
):
152 self
._write
_ytdl
_file
(ctx
)
153 if not self
.params
.get('keep_fragments', False):
154 self
.try_remove(encodeFilename(ctx
['fragment_filename_sanitized']))
155 del ctx
['fragment_filename_sanitized']
157 def _prepare_frag_download(self
, ctx
):
158 if 'live' not in ctx
:
161 total_frags_str
= '%d' % ctx
['total_frags']
162 ad_frags
= ctx
.get('ad_frags', 0)
164 total_frags_str
+= ' (not including %d ad)' % ad_frags
166 total_frags_str
= 'unknown (live)'
167 self
.to_screen(f
'[{self.FD_NAME}] Total fragments: {total_frags_str}')
168 self
.report_destination(ctx
['filename'])
169 dl
= HttpQuietDownloader(self
.ydl
, {
174 tmpfilename
= self
.temp_name(ctx
['filename'])
178 # Establish possible resume length
179 if os
.path
.isfile(encodeFilename(tmpfilename
)):
181 resume_len
= os
.path
.getsize(encodeFilename(tmpfilename
))
183 # Should be initialized before ytdl file check
185 'tmpfilename': tmpfilename
,
189 if self
.__do
_ytdl
_file
(ctx
):
190 if os
.path
.isfile(encodeFilename(self
.ytdl_filename(ctx
['filename']))):
191 self
._read
_ytdl
_file
(ctx
)
192 is_corrupt
= ctx
.get('ytdl_corrupt') is True
193 is_inconsistent
= ctx
['fragment_index'] > 0 and resume_len
== 0
194 if is_corrupt
or is_inconsistent
:
196 '.ytdl file is corrupt' if is_corrupt
else
197 'Inconsistent state of incomplete fragment download')
199 '%s. Restarting from the beginning ...' % message
)
200 ctx
['fragment_index'] = resume_len
= 0
201 if 'ytdl_corrupt' in ctx
:
202 del ctx
['ytdl_corrupt']
203 self
._write
_ytdl
_file
(ctx
)
205 self
._write
_ytdl
_file
(ctx
)
206 assert ctx
['fragment_index'] == 0
208 dest_stream
, tmpfilename
= self
.sanitize_open(tmpfilename
, open_mode
)
212 'dest_stream': dest_stream
,
213 'tmpfilename': tmpfilename
,
214 # Total complete fragments downloaded so far in bytes
215 'complete_frags_downloaded_bytes': resume_len
,
218 def _start_frag_download(self
, ctx
, info_dict
):
219 resume_len
= ctx
['complete_frags_downloaded_bytes']
220 total_frags
= ctx
['total_frags']
221 ctx_id
= ctx
.get('ctx_id')
222 # This dict stores the download progress, it's updated by the progress
225 'status': 'downloading',
226 'downloaded_bytes': resume_len
,
227 'fragment_index': ctx
['fragment_index'],
228 'fragment_count': total_frags
,
229 'filename': ctx
['filename'],
230 'tmpfilename': ctx
['tmpfilename'],
236 'fragment_started': start
,
237 # Amount of fragment's bytes downloaded by the time of the previous
238 # frag progress hook invocation
239 'prev_frag_downloaded_bytes': 0,
242 def frag_progress_hook(s
):
243 if s
['status'] not in ('downloading', 'finished'):
246 if not total_frags
and ctx
.get('fragment_count'):
247 state
['fragment_count'] = ctx
['fragment_count']
249 if ctx_id
is not None and s
.get('ctx_id') != ctx_id
:
252 state
['max_progress'] = ctx
.get('max_progress')
253 state
['progress_idx'] = ctx
.get('progress_idx')
255 time_now
= time
.time()
256 state
['elapsed'] = time_now
- start
257 frag_total_bytes
= s
.get('total_bytes') or 0
258 s
['fragment_info_dict'] = s
.pop('info_dict', {})
261 (ctx
['complete_frags_downloaded_bytes'] + frag_total_bytes
)
262 / (state
['fragment_index'] + 1) * total_frags
)
263 state
['total_bytes_estimate'] = estimated_size
265 if s
['status'] == 'finished':
266 state
['fragment_index'] += 1
267 ctx
['fragment_index'] = state
['fragment_index']
268 state
['downloaded_bytes'] += frag_total_bytes
- ctx
['prev_frag_downloaded_bytes']
269 ctx
['complete_frags_downloaded_bytes'] = state
['downloaded_bytes']
270 ctx
['speed'] = state
['speed'] = self
.calc_speed(
271 ctx
['fragment_started'], time_now
, frag_total_bytes
)
272 ctx
['fragment_started'] = time
.time()
273 ctx
['prev_frag_downloaded_bytes'] = 0
275 frag_downloaded_bytes
= s
['downloaded_bytes']
276 state
['downloaded_bytes'] += frag_downloaded_bytes
- ctx
['prev_frag_downloaded_bytes']
278 state
['eta'] = self
.calc_eta(
279 start
, time_now
, estimated_size
- resume_len
,
280 state
['downloaded_bytes'] - resume_len
)
281 ctx
['speed'] = state
['speed'] = self
.calc_speed(
282 ctx
['fragment_started'], time_now
, frag_downloaded_bytes
)
283 ctx
['prev_frag_downloaded_bytes'] = frag_downloaded_bytes
284 self
._hook
_progress
(state
, info_dict
)
286 ctx
['dl'].add_progress_hook(frag_progress_hook
)
290 def _finish_frag_download(self
, ctx
, info_dict
):
291 ctx
['dest_stream'].close()
292 if self
.__do
_ytdl
_file
(ctx
):
293 ytdl_filename
= encodeFilename(self
.ytdl_filename(ctx
['filename']))
294 if os
.path
.isfile(ytdl_filename
):
295 self
.try_remove(ytdl_filename
)
296 elapsed
= time
.time() - ctx
['started']
298 to_file
= ctx
['tmpfilename'] != '-'
300 downloaded_bytes
= os
.path
.getsize(encodeFilename(ctx
['tmpfilename']))
302 downloaded_bytes
= ctx
['complete_frags_downloaded_bytes']
304 if not downloaded_bytes
:
306 self
.try_remove(ctx
['tmpfilename'])
307 self
.report_error('The downloaded file is empty')
310 self
.try_rename(ctx
['tmpfilename'], ctx
['filename'])
311 filetime
= ctx
.get('fragment_filetime')
312 if self
.params
.get('updatetime', True) and filetime
:
313 with contextlib
.suppress(Exception):
314 os
.utime(ctx
['filename'], (time
.time(), filetime
))
316 self
._hook
_progress
({
317 'downloaded_bytes': downloaded_bytes
,
318 'total_bytes': downloaded_bytes
,
319 'filename': ctx
['filename'],
320 'status': 'finished',
322 'ctx_id': ctx
.get('ctx_id'),
323 'max_progress': ctx
.get('max_progress'),
324 'progress_idx': ctx
.get('progress_idx'),
328 def _prepare_external_frag_download(self
, ctx
):
329 if 'live' not in ctx
:
332 total_frags_str
= '%d' % ctx
['total_frags']
333 ad_frags
= ctx
.get('ad_frags', 0)
335 total_frags_str
+= ' (not including %d ad)' % ad_frags
337 total_frags_str
= 'unknown (live)'
338 self
.to_screen(f
'[{self.FD_NAME}] Total fragments: {total_frags_str}')
340 tmpfilename
= self
.temp_name(ctx
['filename'])
342 # Should be initialized before ytdl file check
344 'tmpfilename': tmpfilename
,
348 def decrypter(self
, info_dict
):
352 if url
not in _key_cache
:
353 _key_cache
[url
] = self
.ydl
.urlopen(self
._prepare
_url
(info_dict
, url
)).read()
354 return _key_cache
[url
]
356 def decrypt_fragment(fragment
, frag_content
):
357 if frag_content
is None:
359 decrypt_info
= fragment
.get('decrypt_info')
360 if not decrypt_info
or decrypt_info
['METHOD'] != 'AES-128':
362 iv
= decrypt_info
.get('IV') or struct
.pack('>8xq', fragment
['media_sequence'])
363 decrypt_info
['KEY'] = decrypt_info
.get('KEY') or _get_key(info_dict
.get('_decryption_key_url') or decrypt_info
['URI'])
364 # Don't decrypt the content in tests since the data is explicitly truncated and it's not to a valid block
365 # size (see https://github.com/ytdl-org/youtube-dl/pull/27660). Tests only care that the correct data downloaded,
366 # not what it decrypts to.
367 if self
.params
.get('test', False):
369 return unpad_pkcs7(aes_cbc_decrypt_bytes(frag_content
, decrypt_info
['KEY'], iv
))
371 return decrypt_fragment
373 def download_and_append_fragments_multiple(self
, *args
, **kwargs
):
375 @params (ctx1, fragments1, info_dict1), (ctx2, fragments2, info_dict2), ...
376 all args must be either tuple or list
378 interrupt_trigger
= [True]
379 max_progress
= len(args
)
380 if max_progress
== 1:
381 return self
.download_and_append_fragments(*args
[0], **kwargs
)
382 max_workers
= self
.params
.get('concurrent_fragment_downloads', 1)
384 self
._prepare
_multiline
_status
(max_progress
)
385 is_live
= any(traverse_obj(args
, (..., 2, 'is_live'), default
=[]))
387 def thread_func(idx
, ctx
, fragments
, info_dict
, tpe
):
388 ctx
['max_progress'] = max_progress
389 ctx
['progress_idx'] = idx
390 return self
.download_and_append_fragments(
391 ctx
, fragments
, info_dict
, **kwargs
, tpe
=tpe
, interrupt_trigger
=interrupt_trigger
)
393 class FTPE(concurrent
.futures
.ThreadPoolExecutor
):
394 # has to stop this or it's going to wait on the worker thread itself
395 def __exit__(self
, exc_type
, exc_val
, exc_tb
):
398 if compat_os_name
== 'nt':
399 def future_result(future
):
402 return future
.result(0.1)
403 except KeyboardInterrupt:
405 except concurrent
.futures
.TimeoutError
:
408 def future_result(future
):
409 return future
.result()
411 def interrupt_trigger_iter(fg
):
413 if not interrupt_trigger
[0]:
418 for idx
, (ctx
, fragments
, info_dict
) in enumerate(args
):
419 tpe
= FTPE(math
.ceil(max_workers
/ max_progress
))
420 job
= tpe
.submit(thread_func
, idx
, ctx
, interrupt_trigger_iter(fragments
), info_dict
, tpe
)
421 spins
.append((tpe
, job
))
424 for tpe
, job
in spins
:
426 result
= result
and future_result(job
)
427 except KeyboardInterrupt:
428 interrupt_trigger
[0] = False
430 tpe
.shutdown(wait
=True)
431 if not interrupt_trigger
[0] and not is_live
:
432 raise KeyboardInterrupt()
433 # we expect the user wants to stop and DO WANT the preceding postprocessors to run;
434 # so returning a intermediate result here instead of KeyboardInterrupt on live
437 def download_and_append_fragments(
438 self
, ctx
, fragments
, info_dict
, *, is_fatal
=(lambda idx
: False),
439 pack_func
=(lambda content
, idx
: content
), finish_func
=None,
440 tpe
=None, interrupt_trigger
=(True, )):
442 if not self
.params
.get('skip_unavailable_fragments', True):
443 is_fatal
= lambda _
: True
445 def download_fragment(fragment
, ctx
):
446 if not interrupt_trigger
[0]:
449 frag_index
= ctx
['fragment_index'] = fragment
['frag_index']
450 ctx
['last_error'] = None
451 headers
= info_dict
.get('http_headers', {}).copy()
452 byte_range
= fragment
.get('byte_range')
454 headers
['Range'] = 'bytes=%d-%d' % (byte_range
['start'], byte_range
['end'] - 1)
456 # Never skip the first fragment
457 fatal
= is_fatal(fragment
.get('index') or (frag_index
- 1))
459 def error_callback(err
, count
, retries
):
460 if fatal
and count
> retries
:
461 ctx
['dest_stream'].close()
462 self
.report_retry(err
, count
, retries
, frag_index
, fatal
)
463 ctx
['last_error'] = err
465 for retry
in RetryManager(self
.params
.get('fragment_retries'), error_callback
):
467 ctx
['fragment_count'] = fragment
.get('fragment_count')
468 if not self
._download
_fragment
(ctx
, fragment
['url'], info_dict
, headers
):
470 except (urllib
.error
.HTTPError
, http
.client
.IncompleteRead
) as err
:
473 except DownloadError
: # has own retry settings
477 def append_fragment(frag_content
, frag_index
, ctx
):
479 self
._append
_fragment
(ctx
, pack_func(frag_content
, frag_index
))
480 elif not is_fatal(frag_index
- 1):
481 self
.report_skip_fragment(frag_index
, 'fragment not found')
483 ctx
['dest_stream'].close()
484 self
.report_error(f
'fragment {frag_index} not found, unable to continue')
488 decrypt_fragment
= self
.decrypter(info_dict
)
490 max_workers
= math
.ceil(
491 self
.params
.get('concurrent_fragment_downloads', 1) / ctx
.get('max_progress', 1))
493 def _download_fragment(fragment
):
494 ctx_copy
= ctx
.copy()
495 download_fragment(fragment
, ctx_copy
)
496 return fragment
, fragment
['frag_index'], ctx_copy
.get('fragment_filename_sanitized')
498 self
.report_warning('The download speed shown is only of one thread. This is a known issue and patches are welcome')
499 with tpe
or concurrent
.futures
.ThreadPoolExecutor(max_workers
) as pool
:
501 for fragment
, frag_index
, frag_filename
in pool
.map(_download_fragment
, fragments
):
503 'fragment_filename_sanitized': frag_filename
,
504 'fragment_index': frag_index
,
506 if not append_fragment(decrypt_fragment(fragment
, self
._read
_fragment
(ctx
)), frag_index
, ctx
):
508 except KeyboardInterrupt:
509 self
._finish
_multiline
_status
()
511 'Interrupted by user. Waiting for all threads to shutdown...', is_error
=False, tb
=False)
512 pool
.shutdown(wait
=False)
515 for fragment
in fragments
:
516 if not interrupt_trigger
[0]:
519 download_fragment(fragment
, ctx
)
520 result
= append_fragment(
521 decrypt_fragment(fragment
, self
._read
_fragment
(ctx
)), fragment
['frag_index'], ctx
)
522 except KeyboardInterrupt:
523 if info_dict
.get('is_live'):
529 if finish_func
is not None:
530 ctx
['dest_stream'].write(finish_func())
531 ctx
['dest_stream'].flush()
532 return self
._finish
_frag
_download
(ctx
, info_dict
)