8 import concurrent
.futures
9 can_threaded_download
= True
11 can_threaded_download
= False
13 from .common
import FileDownloader
14 from .http
import HttpFD
15 from ..aes
import aes_cbc_decrypt_bytes
, unpad_pkcs7
16 from ..compat
import compat_os_name
, compat_struct_pack
, compat_urllib_error
26 class HttpQuietDownloader(HttpFD
):
27 def to_screen(self
, *args
, **kargs
):
30 def report_retry(self
, err
, count
, retries
):
32 f
'[download] Got server HTTP error: {err}. Retrying (attempt {count} of {self.format_retries(retries)}) ...')
35 class FragmentFD(FileDownloader
):
37 A base file downloader class for fragmented media (e.g. f4m/m3u8 manifests).
41 fragment_retries: Number of times to retry a fragment for HTTP error (DASH
43 skip_unavailable_fragments:
44 Skip unavailable fragments (DASH and hlsnative only)
45 keep_fragments: Keep downloaded fragments on disk after downloading is
47 concurrent_fragment_downloads: The number of threads to use for native hls and dash downloads
48 _no_ytdl_file: Don't use .ytdl file
50 For each incomplete fragment download yt-dlp keeps on disk a special
51 bookkeeping file with download state and metadata (in future such files will
52 be used for any incomplete download handled by yt-dlp). This file is
53 used to properly handle resuming, check download file consistency and detect
54 potential errors. The file has a .ytdl extension and represents a standard
55 JSON file of the following format:
58 Dictionary of extractor related data. TBD.
61 Dictionary of downloader related data. May contain following data:
63 Dictionary with current (being downloaded) fragment data:
64 index: 0-based index of current fragment among all fragments
66 Total count of fragments
68 This feature is experimental and file format may change in future.
71 def report_retry_fragment(self
, err
, frag_index
, count
, retries
):
73 '\r[download] Got server HTTP error: %s. Retrying fragment %d (attempt %d of %s) ...'
74 % (error_to_compat_str(err
), frag_index
, count
, self
.format_retries(retries
)))
76 def report_skip_fragment(self
, frag_index
, err
=None):
77 err
= f
' {err};' if err
else ''
78 self
.to_screen(f
'[download]{err} Skipping fragment {frag_index:d} ...')
80 def _prepare_url(self
, info_dict
, url
):
81 headers
= info_dict
.get('http_headers')
82 return sanitized_Request(url
, None, headers
) if headers
else url
84 def _prepare_and_start_frag_download(self
, ctx
, info_dict
):
85 self
._prepare
_frag
_download
(ctx
)
86 self
._start
_frag
_download
(ctx
, info_dict
)
88 def __do_ytdl_file(self
, ctx
):
89 return ctx
['live'] is not True and ctx
['tmpfilename'] != '-' and not self
.params
.get('_no_ytdl_file')
91 def _read_ytdl_file(self
, ctx
):
92 assert 'ytdl_corrupt' not in ctx
93 stream
, _
= self
.sanitize_open(self
.ytdl_filename(ctx
['filename']), 'r')
95 ytdl_data
= json
.loads(stream
.read())
96 ctx
['fragment_index'] = ytdl_data
['downloader']['current_fragment']['index']
97 if 'extra_state' in ytdl_data
['downloader']:
98 ctx
['extra_state'] = ytdl_data
['downloader']['extra_state']
100 ctx
['ytdl_corrupt'] = True
104 def _write_ytdl_file(self
, ctx
):
105 frag_index_stream
, _
= self
.sanitize_open(self
.ytdl_filename(ctx
['filename']), 'w')
108 'current_fragment': {
109 'index': ctx
['fragment_index'],
112 if 'extra_state' in ctx
:
113 downloader
['extra_state'] = ctx
['extra_state']
114 if ctx
.get('fragment_count') is not None:
115 downloader
['fragment_count'] = ctx
['fragment_count']
116 frag_index_stream
.write(json
.dumps({'downloader': downloader}
))
118 frag_index_stream
.close()
120 def _download_fragment(self
, ctx
, frag_url
, info_dict
, headers
=None, request_data
=None):
121 fragment_filename
= '%s-Frag%d' % (ctx
['tmpfilename'], ctx
['fragment_index'])
122 fragment_info_dict
= {
124 'http_headers': headers
or info_dict
.get('http_headers'),
125 'request_data': request_data
,
126 'ctx_id': ctx
.get('ctx_id'),
128 success
= ctx
['dl'].download(fragment_filename
, fragment_info_dict
)
131 if fragment_info_dict
.get('filetime'):
132 ctx
['fragment_filetime'] = fragment_info_dict
.get('filetime')
133 ctx
['fragment_filename_sanitized'] = fragment_filename
136 def _read_fragment(self
, ctx
):
137 if not ctx
.get('fragment_filename_sanitized'):
140 down
, frag_sanitized
= self
.sanitize_open(ctx
['fragment_filename_sanitized'], 'rb')
141 except FileNotFoundError
:
145 ctx
['fragment_filename_sanitized'] = frag_sanitized
146 frag_content
= down
.read()
150 def _append_fragment(self
, ctx
, frag_content
):
152 ctx
['dest_stream'].write(frag_content
)
153 ctx
['dest_stream'].flush()
155 if self
.__do
_ytdl
_file
(ctx
):
156 self
._write
_ytdl
_file
(ctx
)
157 if not self
.params
.get('keep_fragments', False):
158 self
.try_remove(encodeFilename(ctx
['fragment_filename_sanitized']))
159 del ctx
['fragment_filename_sanitized']
161 def _prepare_frag_download(self
, ctx
):
162 if 'live' not in ctx
:
165 total_frags_str
= '%d' % ctx
['total_frags']
166 ad_frags
= ctx
.get('ad_frags', 0)
168 total_frags_str
+= ' (not including %d ad)' % ad_frags
170 total_frags_str
= 'unknown (live)'
171 self
.to_screen(f
'[{self.FD_NAME}] Total fragments: {total_frags_str}')
172 self
.report_destination(ctx
['filename'])
173 dl
= HttpQuietDownloader(
176 'continuedl': self
.params
.get('continuedl', True),
177 'quiet': self
.params
.get('quiet'),
179 'ratelimit': self
.params
.get('ratelimit'),
180 'retries': self
.params
.get('retries', 0),
181 'nopart': self
.params
.get('nopart', False),
185 tmpfilename
= self
.temp_name(ctx
['filename'])
189 # Establish possible resume length
190 if os
.path
.isfile(encodeFilename(tmpfilename
)):
192 resume_len
= os
.path
.getsize(encodeFilename(tmpfilename
))
194 # Should be initialized before ytdl file check
196 'tmpfilename': tmpfilename
,
200 if self
.__do
_ytdl
_file
(ctx
):
201 if os
.path
.isfile(encodeFilename(self
.ytdl_filename(ctx
['filename']))):
202 self
._read
_ytdl
_file
(ctx
)
203 is_corrupt
= ctx
.get('ytdl_corrupt') is True
204 is_inconsistent
= ctx
['fragment_index'] > 0 and resume_len
== 0
205 if is_corrupt
or is_inconsistent
:
207 '.ytdl file is corrupt' if is_corrupt
else
208 'Inconsistent state of incomplete fragment download')
210 '%s. Restarting from the beginning ...' % message
)
211 ctx
['fragment_index'] = resume_len
= 0
212 if 'ytdl_corrupt' in ctx
:
213 del ctx
['ytdl_corrupt']
214 self
._write
_ytdl
_file
(ctx
)
216 self
._write
_ytdl
_file
(ctx
)
217 assert ctx
['fragment_index'] == 0
219 dest_stream
, tmpfilename
= self
.sanitize_open(tmpfilename
, open_mode
)
223 'dest_stream': dest_stream
,
224 'tmpfilename': tmpfilename
,
225 # Total complete fragments downloaded so far in bytes
226 'complete_frags_downloaded_bytes': resume_len
,
229 def _start_frag_download(self
, ctx
, info_dict
):
230 resume_len
= ctx
['complete_frags_downloaded_bytes']
231 total_frags
= ctx
['total_frags']
232 ctx_id
= ctx
.get('ctx_id')
233 # This dict stores the download progress, it's updated by the progress
236 'status': 'downloading',
237 'downloaded_bytes': resume_len
,
238 'fragment_index': ctx
['fragment_index'],
239 'fragment_count': total_frags
,
240 'filename': ctx
['filename'],
241 'tmpfilename': ctx
['tmpfilename'],
247 'fragment_started': start
,
248 # Amount of fragment's bytes downloaded by the time of the previous
249 # frag progress hook invocation
250 'prev_frag_downloaded_bytes': 0,
253 def frag_progress_hook(s
):
254 if s
['status'] not in ('downloading', 'finished'):
257 if ctx_id
is not None and s
.get('ctx_id') != ctx_id
:
260 state
['max_progress'] = ctx
.get('max_progress')
261 state
['progress_idx'] = ctx
.get('progress_idx')
263 time_now
= time
.time()
264 state
['elapsed'] = time_now
- start
265 frag_total_bytes
= s
.get('total_bytes') or 0
266 s
['fragment_info_dict'] = s
.pop('info_dict', {})
269 (ctx
['complete_frags_downloaded_bytes'] + frag_total_bytes
)
270 / (state
['fragment_index'] + 1) * total_frags
)
271 state
['total_bytes_estimate'] = estimated_size
273 if s
['status'] == 'finished':
274 state
['fragment_index'] += 1
275 ctx
['fragment_index'] = state
['fragment_index']
276 state
['downloaded_bytes'] += frag_total_bytes
- ctx
['prev_frag_downloaded_bytes']
277 ctx
['complete_frags_downloaded_bytes'] = state
['downloaded_bytes']
278 ctx
['speed'] = state
['speed'] = self
.calc_speed(
279 ctx
['fragment_started'], time_now
, frag_total_bytes
)
280 ctx
['fragment_started'] = time
.time()
281 ctx
['prev_frag_downloaded_bytes'] = 0
283 frag_downloaded_bytes
= s
['downloaded_bytes']
284 state
['downloaded_bytes'] += frag_downloaded_bytes
- ctx
['prev_frag_downloaded_bytes']
286 state
['eta'] = self
.calc_eta(
287 start
, time_now
, estimated_size
- resume_len
,
288 state
['downloaded_bytes'] - resume_len
)
289 ctx
['speed'] = state
['speed'] = self
.calc_speed(
290 ctx
['fragment_started'], time_now
, frag_downloaded_bytes
)
291 ctx
['prev_frag_downloaded_bytes'] = frag_downloaded_bytes
292 self
._hook
_progress
(state
, info_dict
)
294 ctx
['dl'].add_progress_hook(frag_progress_hook
)
298 def _finish_frag_download(self
, ctx
, info_dict
):
299 ctx
['dest_stream'].close()
300 if self
.__do
_ytdl
_file
(ctx
):
301 ytdl_filename
= encodeFilename(self
.ytdl_filename(ctx
['filename']))
302 if os
.path
.isfile(ytdl_filename
):
303 self
.try_remove(ytdl_filename
)
304 elapsed
= time
.time() - ctx
['started']
306 if ctx
['tmpfilename'] == '-':
307 downloaded_bytes
= ctx
['complete_frags_downloaded_bytes']
309 self
.try_rename(ctx
['tmpfilename'], ctx
['filename'])
310 if self
.params
.get('updatetime', True):
311 filetime
= ctx
.get('fragment_filetime')
314 os
.utime(ctx
['filename'], (time
.time(), filetime
))
317 downloaded_bytes
= os
.path
.getsize(encodeFilename(ctx
['filename']))
319 self
._hook
_progress
({
320 'downloaded_bytes': downloaded_bytes
,
321 'total_bytes': downloaded_bytes
,
322 'filename': ctx
['filename'],
323 'status': 'finished',
325 'ctx_id': ctx
.get('ctx_id'),
326 'max_progress': ctx
.get('max_progress'),
327 'progress_idx': ctx
.get('progress_idx'),
330 def _prepare_external_frag_download(self
, ctx
):
331 if 'live' not in ctx
:
334 total_frags_str
= '%d' % ctx
['total_frags']
335 ad_frags
= ctx
.get('ad_frags', 0)
337 total_frags_str
+= ' (not including %d ad)' % ad_frags
339 total_frags_str
= 'unknown (live)'
340 self
.to_screen(f
'[{self.FD_NAME}] Total fragments: {total_frags_str}')
342 tmpfilename
= self
.temp_name(ctx
['filename'])
344 # Should be initialized before ytdl file check
346 'tmpfilename': tmpfilename
,
350 def decrypter(self
, info_dict
):
354 if url
not in _key_cache
:
355 _key_cache
[url
] = self
.ydl
.urlopen(self
._prepare
_url
(info_dict
, url
)).read()
356 return _key_cache
[url
]
358 def decrypt_fragment(fragment
, frag_content
):
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 compat_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
, pack_func
=None, finish_func
=None):
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], pack_func
=pack_func
, finish_func
=finish_func
)
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
, pack_func
=pack_func
, finish_func
=finish_func
,
392 tpe
=tpe
, interrupt_trigger
=interrupt_trigger
)
394 class FTPE(concurrent
.futures
.ThreadPoolExecutor
):
395 # has to stop this or it's going to wait on the worker thread itself
396 def __exit__(self
, exc_type
, exc_val
, exc_tb
):
399 if compat_os_name
== 'nt':
400 def future_result(future
):
403 return future
.result(0.1)
404 except KeyboardInterrupt:
406 except concurrent
.futures
.TimeoutError
:
409 def future_result(future
):
410 return future
.result()
412 def interrupt_trigger_iter(fg
):
414 if not interrupt_trigger
[0]:
419 for idx
, (ctx
, fragments
, info_dict
) in enumerate(args
):
420 tpe
= FTPE(math
.ceil(max_workers
/ max_progress
))
421 job
= tpe
.submit(thread_func
, idx
, ctx
, interrupt_trigger_iter(fragments
), info_dict
, tpe
)
422 spins
.append((tpe
, job
))
425 for tpe
, job
in spins
:
427 result
= result
and future_result(job
)
428 except KeyboardInterrupt:
429 interrupt_trigger
[0] = False
431 tpe
.shutdown(wait
=True)
432 if not interrupt_trigger
[0] and not is_live
:
433 raise KeyboardInterrupt()
434 # we expect the user wants to stop and DO WANT the preceding postprocessors to run;
435 # so returning a intermediate result here instead of KeyboardInterrupt on live
438 def download_and_append_fragments(
439 self
, ctx
, fragments
, info_dict
, *, pack_func
=None, finish_func
=None,
440 tpe
=None, interrupt_trigger
=None):
441 if not interrupt_trigger
:
442 interrupt_trigger
= (True, )
444 fragment_retries
= self
.params
.get('fragment_retries', 0)
446 ((lambda _
: False) if info_dict
.get('is_live') else (lambda idx
: idx
== 0))
447 if self
.params
.get('skip_unavailable_fragments', True) else (lambda _
: True))
450 pack_func
= lambda frag_content
, _
: frag_content
452 def download_fragment(fragment
, ctx
):
453 if not interrupt_trigger
[0]:
456 frag_index
= ctx
['fragment_index'] = fragment
['frag_index']
457 ctx
['last_error'] = None
458 headers
= info_dict
.get('http_headers', {}).copy()
459 byte_range
= fragment
.get('byte_range')
461 headers
['Range'] = 'bytes=%d-%d' % (byte_range
['start'], byte_range
['end'] - 1)
463 # Never skip the first fragment
464 fatal
, count
= is_fatal(fragment
.get('index') or (frag_index
- 1)), 0
465 while count
<= fragment_retries
:
467 if self
._download
_fragment
(ctx
, fragment
['url'], info_dict
, headers
):
470 except (compat_urllib_error
.HTTPError
, http
.client
.IncompleteRead
) as err
:
471 # Unavailable (possibly temporary) fragments may be served.
472 # First we try to retry then either skip or abort.
473 # See https://github.com/ytdl-org/youtube-dl/issues/10165,
474 # https://github.com/ytdl-org/youtube-dl/issues/10448).
476 ctx
['last_error'] = err
477 if count
<= fragment_retries
:
478 self
.report_retry_fragment(err
, frag_index
, count
, fragment_retries
)
479 except DownloadError
:
480 # Don't retry fragment if error occurred during HTTP downloading
481 # itself since it has own retry settings
486 if count
> fragment_retries
and fatal
:
487 ctx
['dest_stream'].close()
488 self
.report_error('Giving up after %s fragment retries' % fragment_retries
)
490 def append_fragment(frag_content
, frag_index
, ctx
):
492 self
._append
_fragment
(ctx
, pack_func(frag_content
, frag_index
))
493 elif not is_fatal(frag_index
- 1):
494 self
.report_skip_fragment(frag_index
, 'fragment not found')
496 ctx
['dest_stream'].close()
497 self
.report_error(f
'fragment {frag_index} not found, unable to continue')
501 decrypt_fragment
= self
.decrypter(info_dict
)
503 max_workers
= math
.ceil(
504 self
.params
.get('concurrent_fragment_downloads', 1) / ctx
.get('max_progress', 1))
505 if can_threaded_download
and max_workers
> 1:
507 def _download_fragment(fragment
):
508 ctx_copy
= ctx
.copy()
509 download_fragment(fragment
, ctx_copy
)
510 return fragment
, fragment
['frag_index'], ctx_copy
.get('fragment_filename_sanitized')
512 self
.report_warning('The download speed shown is only of one thread. This is a known issue and patches are welcome')
513 with tpe
or concurrent
.futures
.ThreadPoolExecutor(max_workers
) as pool
:
514 for fragment
, frag_index
, frag_filename
in pool
.map(_download_fragment
, fragments
):
515 ctx
['fragment_filename_sanitized'] = frag_filename
516 ctx
['fragment_index'] = frag_index
517 result
= append_fragment(decrypt_fragment(fragment
, self
._read
_fragment
(ctx
)), frag_index
, ctx
)
521 for fragment
in fragments
:
522 if not interrupt_trigger
[0]:
524 download_fragment(fragment
, ctx
)
525 result
= append_fragment(decrypt_fragment(fragment
, self
._read
_fragment
(ctx
)), fragment
['frag_index'], ctx
)
529 if finish_func
is not None:
530 ctx
['dest_stream'].write(finish_func())
531 ctx
['dest_stream'].flush()
532 self
._finish
_frag
_download
(ctx
, info_dict
)