yt_dlp/downloader/fragment.py

   1 import http.client
   2 import json
   3 import math
   4 import os
   5 import time
   6
   7 try:
   8     import concurrent.futures
   9     can_threaded_download = True
  10 except ImportError:
  11     can_threaded_download = False
  12
  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
  17 from ..utils import (
  18     DownloadError,
  19     encodeFilename,
  20     error_to_compat_str,
  21     sanitized_Request,
  22     traverse_obj,
  23 )
  24
  25
  26 class HttpQuietDownloader(HttpFD):
  27     def to_screen(self, *args, **kargs):
  28         pass
  29
  30     def report_retry(self, err, count, retries):
  31         super().to_screen(
  32             f'[download] Got server HTTP error: {err}. Retrying (attempt {count} of {self.format_retries(retries)}) ...')
  33
  34
  35 class FragmentFD(FileDownloader):
  36     """
  37     A base file downloader class for fragmented media (e.g. f4m/m3u8 manifests).
  38
  39     Available options:
  40
  41     fragment_retries:   Number of times to retry a fragment for HTTP error (DASH
  42                         and hlsnative only)
  43     skip_unavailable_fragments:
  44                         Skip unavailable fragments (DASH and hlsnative only)
  45     keep_fragments:     Keep downloaded fragments on disk after downloading is
  46                         finished
  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
  49
  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:
  56
  57     extractor:
  58         Dictionary of extractor related data. TBD.
  59
  60     downloader:
  61         Dictionary of downloader related data. May contain following data:
  62             current_fragment:
  63                 Dictionary with current (being downloaded) fragment data:
  64                 index:  0-based index of current fragment among all fragments
  65             fragment_count:
  66                 Total count of fragments
  67
  68     This feature is experimental and file format may change in future.
  69     """
  70
  71     def report_retry_fragment(self, err, frag_index, count, retries):
  72         self.to_screen(
  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)))
  75
  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} ...')
  79
  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
  83
  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)
  87
  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')
  90
  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')
  94         try:
  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']
  99         except Exception:
 100             ctx['ytdl_corrupt'] = True
 101         finally:
 102             stream.close()
 103
 104     def _write_ytdl_file(self, ctx):
 105         frag_index_stream, _ = self.sanitize_open(self.ytdl_filename(ctx['filename']), 'w')
 106         try:
 107             downloader = {
 108                 'current_fragment': {
 109                     'index': ctx['fragment_index'],
 110                 },
 111             }
 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}))
 117         finally:
 118             frag_index_stream.close()
 119
 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 = {
 123             'url': frag_url,
 124             'http_headers': headers or info_dict.get('http_headers'),
 125             'request_data': request_data,
 126             'ctx_id': ctx.get('ctx_id'),
 127         }
 128         success = ctx['dl'].download(fragment_filename, fragment_info_dict)
 129         if not success:
 130             return False
 131         if fragment_info_dict.get('filetime'):
 132             ctx['fragment_filetime'] = fragment_info_dict.get('filetime')
 133         ctx['fragment_filename_sanitized'] = fragment_filename
 134         return True
 135
 136     def _read_fragment(self, ctx):
 137         if not ctx.get('fragment_filename_sanitized'):
 138             return None
 139         try:
 140             down, frag_sanitized = self.sanitize_open(ctx['fragment_filename_sanitized'], 'rb')
 141         except FileNotFoundError:
 142             if ctx.get('live'):
 143                 return None
 144             raise
 145         ctx['fragment_filename_sanitized'] = frag_sanitized
 146         frag_content = down.read()
 147         down.close()
 148         return frag_content
 149
 150     def _append_fragment(self, ctx, frag_content):
 151         try:
 152             ctx['dest_stream'].write(frag_content)
 153             ctx['dest_stream'].flush()
 154         finally:
 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']
 160
 161     def _prepare_frag_download(self, ctx):
 162         if 'live' not in ctx:
 163             ctx['live'] = False
 164         if not ctx['live']:
 165             total_frags_str = '%d' % ctx['total_frags']
 166             ad_frags = ctx.get('ad_frags', 0)
 167             if ad_frags:
 168                 total_frags_str += ' (not including %d ad)' % ad_frags
 169         else:
 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(
 174             self.ydl,
 175             {
 176                 'continuedl': self.params.get('continuedl', True),
 177                 'quiet': self.params.get('quiet'),
 178                 'noprogress': True,
 179                 'ratelimit': self.params.get('ratelimit'),
 180                 'retries': self.params.get('retries', 0),
 181                 'nopart': self.params.get('nopart', False),
 182                 'test': False,
 183             }
 184         )
 185         tmpfilename = self.temp_name(ctx['filename'])
 186         open_mode = 'wb'
 187         resume_len = 0
 188
 189         # Establish possible resume length
 190         if os.path.isfile(encodeFilename(tmpfilename)):
 191             open_mode = 'ab'
 192             resume_len = os.path.getsize(encodeFilename(tmpfilename))
 193
 194         # Should be initialized before ytdl file check
 195         ctx.update({
 196             'tmpfilename': tmpfilename,
 197             'fragment_index': 0,
 198         })
 199
 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:
 206                     message = (
 207                         '.ytdl file is corrupt' if is_corrupt else
 208                         'Inconsistent state of incomplete fragment download')
 209                     self.report_warning(
 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)
 215             else:
 216                 self._write_ytdl_file(ctx)
 217                 assert ctx['fragment_index'] == 0
 218
 219         dest_stream, tmpfilename = self.sanitize_open(tmpfilename, open_mode)
 220
 221         ctx.update({
 222             'dl': dl,
 223             'dest_stream': dest_stream,
 224             'tmpfilename': tmpfilename,
 225             # Total complete fragments downloaded so far in bytes
 226             'complete_frags_downloaded_bytes': resume_len,
 227         })
 228
 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
 234         # hook
 235         state = {
 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'],
 242         }
 243
 244         start = time.time()
 245         ctx.update({
 246             'started': start,
 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,
 251         })
 252
 253         def frag_progress_hook(s):
 254             if s['status'] not in ('downloading', 'finished'):
 255                 return
 256
 257             if ctx_id is not None and s.get('ctx_id') != ctx_id:
 258                 return
 259
 260             state['max_progress'] = ctx.get('max_progress')
 261             state['progress_idx'] = ctx.get('progress_idx')
 262
 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', {})
 267             if not ctx['live']:
 268                 estimated_size = (
 269                     (ctx['complete_frags_downloaded_bytes'] + frag_total_bytes)
 270                     / (state['fragment_index'] + 1) * total_frags)
 271                 state['total_bytes_estimate'] = estimated_size
 272
 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
 282             else:
 283                 frag_downloaded_bytes = s['downloaded_bytes']
 284                 state['downloaded_bytes'] += frag_downloaded_bytes - ctx['prev_frag_downloaded_bytes']
 285                 if not ctx['live']:
 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)
 293
 294         ctx['dl'].add_progress_hook(frag_progress_hook)
 295
 296         return start
 297
 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']
 305
 306         if ctx['tmpfilename'] == '-':
 307             downloaded_bytes = ctx['complete_frags_downloaded_bytes']
 308         else:
 309             self.try_rename(ctx['tmpfilename'], ctx['filename'])
 310             if self.params.get('updatetime', True):
 311                 filetime = ctx.get('fragment_filetime')
 312                 if filetime:
 313                     try:
 314                         os.utime(ctx['filename'], (time.time(), filetime))
 315                     except Exception:
 316                         pass
 317             downloaded_bytes = os.path.getsize(encodeFilename(ctx['filename']))
 318
 319         self._hook_progress({
 320             'downloaded_bytes': downloaded_bytes,
 321             'total_bytes': downloaded_bytes,
 322             'filename': ctx['filename'],
 323             'status': 'finished',
 324             'elapsed': elapsed,
 325             'ctx_id': ctx.get('ctx_id'),
 326             'max_progress': ctx.get('max_progress'),
 327             'progress_idx': ctx.get('progress_idx'),
 328         }, info_dict)
 329
 330     def _prepare_external_frag_download(self, ctx):
 331         if 'live' not in ctx:
 332             ctx['live'] = False
 333         if not ctx['live']:
 334             total_frags_str = '%d' % ctx['total_frags']
 335             ad_frags = ctx.get('ad_frags', 0)
 336             if ad_frags:
 337                 total_frags_str += ' (not including %d ad)' % ad_frags
 338         else:
 339             total_frags_str = 'unknown (live)'
 340         self.to_screen(f'[{self.FD_NAME}] Total fragments: {total_frags_str}')
 341
 342         tmpfilename = self.temp_name(ctx['filename'])
 343
 344         # Should be initialized before ytdl file check
 345         ctx.update({
 346             'tmpfilename': tmpfilename,
 347             'fragment_index': 0,
 348         })
 349
 350     def decrypter(self, info_dict):
 351         _key_cache = {}
 352
 353         def _get_key(url):
 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]
 357
 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':
 361                 return frag_content
 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):
 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, pack_func=None, finish_func=None):
 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], pack_func=pack_func, finish_func=finish_func)
 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, pack_func=pack_func, finish_func=finish_func,
 392                 tpe=tpe, interrupt_trigger=interrupt_trigger)
 393
 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):
 397                 pass
 398
 399         if compat_os_name == 'nt':
 400             def future_result(future):
 401                 while True:
 402                     try:
 403                         return future.result(0.1)
 404                     except KeyboardInterrupt:
 405                         raise
 406                     except concurrent.futures.TimeoutError:
 407                         continue
 408         else:
 409             def future_result(future):
 410                 return future.result()
 411
 412         def interrupt_trigger_iter(fg):
 413             for f in fg:
 414                 if not interrupt_trigger[0]:
 415                     break
 416                 yield f
 417
 418         spins = []
 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))
 423
 424         result = True
 425         for tpe, job in spins:
 426             try:
 427                 result = result and future_result(job)
 428             except KeyboardInterrupt:
 429                 interrupt_trigger[0] = False
 430             finally:
 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
 436         return result
 437
 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, )
 443
 444         fragment_retries = self.params.get('fragment_retries', 0)
 445         is_fatal = (
 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))
 448
 449         if not pack_func:
 450             pack_func = lambda frag_content, _: frag_content
 451
 452         def download_fragment(fragment, ctx):
 453             if not interrupt_trigger[0]:
 454                 return
 455
 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')
 460             if byte_range:
 461                 headers['Range'] = 'bytes=%d-%d' % (byte_range['start'], byte_range['end'] - 1)
 462
 463             # Never skip the first fragment
 464             fatal, count = is_fatal(fragment.get('index') or (frag_index - 1)), 0
 465             while count <= fragment_retries:
 466                 try:
 467                     if self._download_fragment(ctx, fragment['url'], info_dict, headers):
 468                         break
 469                     return
 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).
 475                     count += 1
 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
 482                     if not fatal:
 483                         break
 484                     raise
 485
 486             if count > fragment_retries and fatal:
 487                 ctx['dest_stream'].close()
 488                 self.report_error('Giving up after %s fragment retries' % fragment_retries)
 489
 490         def append_fragment(frag_content, frag_index, ctx):
 491             if frag_content:
 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')
 495             else:
 496                 ctx['dest_stream'].close()
 497                 self.report_error(f'fragment {frag_index} not found, unable to continue')
 498                 return False
 499             return True
 500
 501         decrypt_fragment = self.decrypter(info_dict)
 502
 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:
 506
 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')
 511
 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)
 518                     if not result:
 519                         return False
 520         else:
 521             for fragment in fragments:
 522                 if not interrupt_trigger[0]:
 523                     break
 524                 download_fragment(fragment, ctx)
 525                 result = append_fragment(decrypt_fragment(fragment, self._read_fragment(ctx)), fragment['frag_index'], ctx)
 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         self._finish_frag_download(ctx, info_dict)
 533         return True