yt_dlp/downloader/fragment.py

   1 import concurrent.futures
   2 import contextlib
   3 import http.client
   4 import json
   5 import math
   6 import os
   7 import struct
   8 import time
   9 import urllib.error
  10
  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
  15 from ..utils import (
  16     DownloadError,
  17     RetryManager,
  18     encodeFilename,
  19     sanitized_Request,
  20     traverse_obj,
  21 )
  22
  23
  24 class HttpQuietDownloader(HttpFD):
  25     def to_screen(self, *args, **kargs):
  26         pass
  27
  28     to_console_title = to_screen
  29
  30
  31 class FragmentFD(FileDownloader):
  32     """
  33     A base file downloader class for fragmented media (e.g. f4m/m3u8 manifests).
  34
  35     Available options:
  36
  37     fragment_retries:   Number of times to retry a fragment for HTTP error (DASH
  38                         and hlsnative only)
  39     skip_unavailable_fragments:
  40                         Skip unavailable fragments (DASH and hlsnative only)
  41     keep_fragments:     Keep downloaded fragments on disk after downloading is
  42                         finished
  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
  45
  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:
  52
  53     extractor:
  54         Dictionary of extractor related data. TBD.
  55
  56     downloader:
  57         Dictionary of downloader related data. May contain following data:
  58             current_fragment:
  59                 Dictionary with current (being downloaded) fragment data:
  60                 index:  0-based index of current fragment among all fragments
  61             fragment_count:
  62                 Total count of fragments
  63
  64     This feature is experimental and file format may change in future.
  65     """
  66
  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)
  71
  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} ...')
  75
  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
  79
  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)
  83
  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')
  86
  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')
  90         try:
  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']
  95         except Exception:
  96             ctx['ytdl_corrupt'] = True
  97         finally:
  98             stream.close()
  99
 100     def _write_ytdl_file(self, ctx):
 101         frag_index_stream, _ = self.sanitize_open(self.ytdl_filename(ctx['filename']), 'w')
 102         try:
 103             downloader = {
 104                 'current_fragment': {
 105                     'index': ctx['fragment_index'],
 106                 },
 107             }
 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}))
 113         finally:
 114             frag_index_stream.close()
 115
 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 = {
 119             'url': frag_url,
 120             'http_headers': headers or info_dict.get('http_headers'),
 121             'request_data': request_data,
 122             'ctx_id': ctx.get('ctx_id'),
 123         }
 124         success, _ = ctx['dl'].download(fragment_filename, fragment_info_dict)
 125         if not success:
 126             return False
 127         if fragment_info_dict.get('filetime'):
 128             ctx['fragment_filetime'] = fragment_info_dict.get('filetime')
 129         ctx['fragment_filename_sanitized'] = fragment_filename
 130         return True
 131
 132     def _read_fragment(self, ctx):
 133         if not ctx.get('fragment_filename_sanitized'):
 134             return None
 135         try:
 136             down, frag_sanitized = self.sanitize_open(ctx['fragment_filename_sanitized'], 'rb')
 137         except FileNotFoundError:
 138             if ctx.get('live'):
 139                 return None
 140             raise
 141         ctx['fragment_filename_sanitized'] = frag_sanitized
 142         frag_content = down.read()
 143         down.close()
 144         return frag_content
 145
 146     def _append_fragment(self, ctx, frag_content):
 147         try:
 148             ctx['dest_stream'].write(frag_content)
 149             ctx['dest_stream'].flush()
 150         finally:
 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']
 156
 157     def _prepare_frag_download(self, ctx):
 158         if 'live' not in ctx:
 159             ctx['live'] = False
 160         if not ctx['live']:
 161             total_frags_str = '%d' % ctx['total_frags']
 162             ad_frags = ctx.get('ad_frags', 0)
 163             if ad_frags:
 164                 total_frags_str += ' (not including %d ad)' % ad_frags
 165         else:
 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, {
 170             **self.params,
 171             'noprogress': True,
 172             'test': False,
 173         })
 174         tmpfilename = self.temp_name(ctx['filename'])
 175         open_mode = 'wb'
 176         resume_len = 0
 177
 178         # Establish possible resume length
 179         if os.path.isfile(encodeFilename(tmpfilename)):
 180             open_mode = 'ab'
 181             resume_len = os.path.getsize(encodeFilename(tmpfilename))
 182
 183         # Should be initialized before ytdl file check
 184         ctx.update({
 185             'tmpfilename': tmpfilename,
 186             'fragment_index': 0,
 187         })
 188
 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:
 195                     message = (
 196                         '.ytdl file is corrupt' if is_corrupt else
 197                         'Inconsistent state of incomplete fragment download')
 198                     self.report_warning(
 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)
 204             else:
 205                 self._write_ytdl_file(ctx)
 206                 assert ctx['fragment_index'] == 0
 207
 208         dest_stream, tmpfilename = self.sanitize_open(tmpfilename, open_mode)
 209
 210         ctx.update({
 211             'dl': dl,
 212             'dest_stream': dest_stream,
 213             'tmpfilename': tmpfilename,
 214             # Total complete fragments downloaded so far in bytes
 215             'complete_frags_downloaded_bytes': resume_len,
 216         })
 217
 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
 223         # hook
 224         state = {
 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'],
 231         }
 232
 233         start = time.time()
 234         ctx.update({
 235             'started': start,
 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,
 240         })
 241
 242         def frag_progress_hook(s):
 243             if s['status'] not in ('downloading', 'finished'):
 244                 return
 245
 246             if not total_frags and ctx.get('fragment_count'):
 247                 state['fragment_count'] = ctx['fragment_count']
 248
 249             if ctx_id is not None and s.get('ctx_id') != ctx_id:
 250                 return
 251
 252             state['max_progress'] = ctx.get('max_progress')
 253             state['progress_idx'] = ctx.get('progress_idx')
 254
 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', {})
 259             if not ctx['live']:
 260                 estimated_size = (
 261                     (ctx['complete_frags_downloaded_bytes'] + frag_total_bytes)
 262                     / (state['fragment_index'] + 1) * total_frags)
 263                 state['total_bytes_estimate'] = estimated_size
 264
 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
 274             else:
 275                 frag_downloaded_bytes = s['downloaded_bytes']
 276                 state['downloaded_bytes'] += frag_downloaded_bytes - ctx['prev_frag_downloaded_bytes']
 277                 if not ctx['live']:
 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)
 285
 286         ctx['dl'].add_progress_hook(frag_progress_hook)
 287
 288         return start
 289
 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']
 297
 298         to_file = ctx['tmpfilename'] != '-'
 299         if to_file:
 300             downloaded_bytes = os.path.getsize(encodeFilename(ctx['tmpfilename']))
 301         else:
 302             downloaded_bytes = ctx['complete_frags_downloaded_bytes']
 303
 304         if not downloaded_bytes:
 305             if to_file:
 306                 self.try_remove(ctx['tmpfilename'])
 307             self.report_error('The downloaded file is empty')
 308             return False
 309         elif to_file:
 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))
 315
 316         self._hook_progress({
 317             'downloaded_bytes': downloaded_bytes,
 318             'total_bytes': downloaded_bytes,
 319             'filename': ctx['filename'],
 320             'status': 'finished',
 321             'elapsed': elapsed,
 322             'ctx_id': ctx.get('ctx_id'),
 323             'max_progress': ctx.get('max_progress'),
 324             'progress_idx': ctx.get('progress_idx'),
 325         }, info_dict)
 326         return True
 327
 328     def _prepare_external_frag_download(self, ctx):
 329         if 'live' not in ctx:
 330             ctx['live'] = False
 331         if not ctx['live']:
 332             total_frags_str = '%d' % ctx['total_frags']
 333             ad_frags = ctx.get('ad_frags', 0)
 334             if ad_frags:
 335                 total_frags_str += ' (not including %d ad)' % ad_frags
 336         else:
 337             total_frags_str = 'unknown (live)'
 338         self.to_screen(f'[{self.FD_NAME}] Total fragments: {total_frags_str}')
 339
 340         tmpfilename = self.temp_name(ctx['filename'])
 341
 342         # Should be initialized before ytdl file check
 343         ctx.update({
 344             'tmpfilename': tmpfilename,
 345             'fragment_index': 0,
 346         })
 347
 348     def decrypter(self, info_dict):
 349         _key_cache = {}
 350
 351         def _get_key(url):
 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]
 355
 356         def decrypt_fragment(fragment, frag_content):
 357             if frag_content is None:
 358                 return
 359             decrypt_info = fragment.get('decrypt_info')
 360             if not decrypt_info or decrypt_info['METHOD'] != 'AES-128':
 361                 return frag_content
 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):
 368                 return frag_content
 369             return unpad_pkcs7(aes_cbc_decrypt_bytes(frag_content, decrypt_info['KEY'], iv))
 370
 371         return decrypt_fragment
 372
 373     def download_and_append_fragments_multiple(self, *args, **kwargs):
 374         '''
 375         @params (ctx1, fragments1, info_dict1), (ctx2, fragments2, info_dict2), ...
 376                 all args must be either tuple or list
 377         '''
 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)
 383         if max_progress > 1:
 384             self._prepare_multiline_status(max_progress)
 385         is_live = any(traverse_obj(args, (..., 2, 'is_live'), default=[]))
 386
 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)
 392
 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):
 396                 pass
 397
 398         if compat_os_name == 'nt':
 399             def future_result(future):
 400                 while True:
 401                     try:
 402                         return future.result(0.1)
 403                     except KeyboardInterrupt:
 404                         raise
 405                     except concurrent.futures.TimeoutError:
 406                         continue
 407         else:
 408             def future_result(future):
 409                 return future.result()
 410
 411         def interrupt_trigger_iter(fg):
 412             for f in fg:
 413                 if not interrupt_trigger[0]:
 414                     break
 415                 yield f
 416
 417         spins = []
 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))
 422
 423         result = True
 424         for tpe, job in spins:
 425             try:
 426                 result = result and future_result(job)
 427             except KeyboardInterrupt:
 428                 interrupt_trigger[0] = False
 429             finally:
 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
 435         return result
 436
 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, )):
 441
 442         if not self.params.get('skip_unavailable_fragments', True):
 443             is_fatal = lambda _: True
 444
 445         def download_fragment(fragment, ctx):
 446             if not interrupt_trigger[0]:
 447                 return
 448
 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')
 453             if byte_range:
 454                 headers['Range'] = 'bytes=%d-%d' % (byte_range['start'], byte_range['end'] - 1)
 455
 456             # Never skip the first fragment
 457             fatal = is_fatal(fragment.get('index') or (frag_index - 1))
 458
 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
 464
 465             for retry in RetryManager(self.params.get('fragment_retries'), error_callback):
 466                 try:
 467                     ctx['fragment_count'] = fragment.get('fragment_count')
 468                     if not self._download_fragment(ctx, fragment['url'], info_dict, headers):
 469                         return
 470                 except (urllib.error.HTTPError, http.client.IncompleteRead) as err:
 471                     retry.error = err
 472                     continue
 473                 except DownloadError:  # has own retry settings
 474                     if fatal:
 475                         raise
 476
 477         def append_fragment(frag_content, frag_index, ctx):
 478             if frag_content:
 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')
 482             else:
 483                 ctx['dest_stream'].close()
 484                 self.report_error(f'fragment {frag_index} not found, unable to continue')
 485                 return False
 486             return True
 487
 488         decrypt_fragment = self.decrypter(info_dict)
 489
 490         max_workers = math.ceil(
 491             self.params.get('concurrent_fragment_downloads', 1) / ctx.get('max_progress', 1))
 492         if max_workers > 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')
 497
 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:
 500                 try:
 501                     for fragment, frag_index, frag_filename in pool.map(_download_fragment, fragments):
 502                         ctx.update({
 503                             'fragment_filename_sanitized': frag_filename,
 504                             'fragment_index': frag_index,
 505                         })
 506                         if not append_fragment(decrypt_fragment(fragment, self._read_fragment(ctx)), frag_index, ctx):
 507                             return False
 508                 except KeyboardInterrupt:
 509                     self._finish_multiline_status()
 510                     self.report_error(
 511                         'Interrupted by user. Waiting for all threads to shutdown...', is_error=False, tb=False)
 512                     pool.shutdown(wait=False)
 513                     raise
 514         else:
 515             for fragment in fragments:
 516                 if not interrupt_trigger[0]:
 517                     break
 518                 try:
 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'):
 524                         break
 525                     raise
 526                 if not result:
 527                     return False
 528
 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)