yt_dlp/downloader/fragment.py

   1 from __future__ import division, unicode_literals
   2
   3 import http.client
   4 import json
   5 import math
   6 import os
   7 import time
   8
   9 try:
  10     import concurrent.futures
  11     can_threaded_download = True
  12 except ImportError:
  13     can_threaded_download = False
  14
  15 from .common import FileDownloader
  16 from .http import HttpFD
  17 from ..aes import aes_cbc_decrypt_bytes
  18 from ..compat import (
  19     compat_os_name,
  20     compat_urllib_error,
  21     compat_struct_pack,
  22 )
  23 from ..utils import (
  24     DownloadError,
  25     error_to_compat_str,
  26     encodeFilename,
  27     sanitize_open,
  28     sanitized_Request,
  29 )
  30
  31
  32 class HttpQuietDownloader(HttpFD):
  33     def to_screen(self, *args, **kargs):
  34         pass
  35
  36     def report_retry(self, err, count, retries):
  37         super().to_screen(
  38             f'[download] Got server HTTP error: {err}. Retrying (attempt {count} of {self.format_retries(retries)}) ...')
  39
  40
  41 class FragmentFD(FileDownloader):
  42     """
  43     A base file downloader class for fragmented media (e.g. f4m/m3u8 manifests).
  44
  45     Available options:
  46
  47     fragment_retries:   Number of times to retry a fragment for HTTP error (DASH
  48                         and hlsnative only)
  49     skip_unavailable_fragments:
  50                         Skip unavailable fragments (DASH and hlsnative only)
  51     keep_fragments:     Keep downloaded fragments on disk after downloading is
  52                         finished
  53     concurrent_fragment_downloads:  The number of threads to use for native hls and dash downloads
  54     _no_ytdl_file:      Don't use .ytdl file
  55
  56     For each incomplete fragment download yt-dlp keeps on disk a special
  57     bookkeeping file with download state and metadata (in future such files will
  58     be used for any incomplete download handled by yt-dlp). This file is
  59     used to properly handle resuming, check download file consistency and detect
  60     potential errors. The file has a .ytdl extension and represents a standard
  61     JSON file of the following format:
  62
  63     extractor:
  64         Dictionary of extractor related data. TBD.
  65
  66     downloader:
  67         Dictionary of downloader related data. May contain following data:
  68             current_fragment:
  69                 Dictionary with current (being downloaded) fragment data:
  70                 index:  0-based index of current fragment among all fragments
  71             fragment_count:
  72                 Total count of fragments
  73
  74     This feature is experimental and file format may change in future.
  75     """
  76
  77     def report_retry_fragment(self, err, frag_index, count, retries):
  78         self.to_screen(
  79             '\r[download] Got server HTTP error: %s. Retrying fragment %d (attempt %d of %s) ...'
  80             % (error_to_compat_str(err), frag_index, count, self.format_retries(retries)))
  81
  82     def report_skip_fragment(self, frag_index, err=None):
  83         err = f' {err};' if err else ''
  84         self.to_screen(f'[download]{err} Skipping fragment {frag_index:d} ...')
  85
  86     def _prepare_url(self, info_dict, url):
  87         headers = info_dict.get('http_headers')
  88         return sanitized_Request(url, None, headers) if headers else url
  89
  90     def _prepare_and_start_frag_download(self, ctx, info_dict):
  91         self._prepare_frag_download(ctx)
  92         self._start_frag_download(ctx, info_dict)
  93
  94     def __do_ytdl_file(self, ctx):
  95         return ctx['live'] is not True and ctx['tmpfilename'] != '-' and not self.params.get('_no_ytdl_file')
  96
  97     def _read_ytdl_file(self, ctx):
  98         assert 'ytdl_corrupt' not in ctx
  99         stream, _ = sanitize_open(self.ytdl_filename(ctx['filename']), 'r')
 100         try:
 101             ytdl_data = json.loads(stream.read())
 102             ctx['fragment_index'] = ytdl_data['downloader']['current_fragment']['index']
 103             if 'extra_state' in ytdl_data['downloader']:
 104                 ctx['extra_state'] = ytdl_data['downloader']['extra_state']
 105         except Exception:
 106             ctx['ytdl_corrupt'] = True
 107         finally:
 108             stream.close()
 109
 110     def _write_ytdl_file(self, ctx):
 111         frag_index_stream, _ = sanitize_open(self.ytdl_filename(ctx['filename']), 'w')
 112         try:
 113             downloader = {
 114                 'current_fragment': {
 115                     'index': ctx['fragment_index'],
 116                 },
 117             }
 118             if 'extra_state' in ctx:
 119                 downloader['extra_state'] = ctx['extra_state']
 120             if ctx.get('fragment_count') is not None:
 121                 downloader['fragment_count'] = ctx['fragment_count']
 122             frag_index_stream.write(json.dumps({'downloader': downloader}))
 123         finally:
 124             frag_index_stream.close()
 125
 126     def _download_fragment(self, ctx, frag_url, info_dict, headers=None, request_data=None):
 127         fragment_filename = '%s-Frag%d' % (ctx['tmpfilename'], ctx['fragment_index'])
 128         fragment_info_dict = {
 129             'url': frag_url,
 130             'http_headers': headers or info_dict.get('http_headers'),
 131             'request_data': request_data,
 132             'ctx_id': ctx.get('ctx_id'),
 133         }
 134         success = ctx['dl'].download(fragment_filename, fragment_info_dict)
 135         if not success:
 136             return False, None
 137         if fragment_info_dict.get('filetime'):
 138             ctx['fragment_filetime'] = fragment_info_dict.get('filetime')
 139         ctx['fragment_filename_sanitized'] = fragment_filename
 140         return True, self._read_fragment(ctx)
 141
 142     def _read_fragment(self, ctx):
 143         down, frag_sanitized = sanitize_open(ctx['fragment_filename_sanitized'], 'rb')
 144         ctx['fragment_filename_sanitized'] = frag_sanitized
 145         frag_content = down.read()
 146         down.close()
 147         return frag_content
 148
 149     def _append_fragment(self, ctx, frag_content):
 150         try:
 151             ctx['dest_stream'].write(frag_content)
 152             ctx['dest_stream'].flush()
 153         finally:
 154             if self.__do_ytdl_file(ctx):
 155                 self._write_ytdl_file(ctx)
 156             if not self.params.get('keep_fragments', False):
 157                 os.remove(encodeFilename(ctx['fragment_filename_sanitized']))
 158             del ctx['fragment_filename_sanitized']
 159
 160     def _prepare_frag_download(self, ctx):
 161         if 'live' not in ctx:
 162             ctx['live'] = False
 163         if not ctx['live']:
 164             total_frags_str = '%d' % ctx['total_frags']
 165             ad_frags = ctx.get('ad_frags', 0)
 166             if ad_frags:
 167                 total_frags_str += ' (not including %d ad)' % ad_frags
 168         else:
 169             total_frags_str = 'unknown (live)'
 170         self.to_screen(
 171             '[%s] Total fragments: %s' % (self.FD_NAME, total_frags_str))
 172         self.report_destination(ctx['filename'])
 173         dl = HttpQuietDownloader(
 174             self.ydl,
 175             {
 176                 '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': self.params.get('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 = 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                 os.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(
 341             '[%s] Total fragments: %s' % (self.FD_NAME, total_frags_str))
 342
 343         tmpfilename = self.temp_name(ctx['filename'])
 344
 345         # Should be initialized before ytdl file check
 346         ctx.update({
 347             'tmpfilename': tmpfilename,
 348             'fragment_index': 0,
 349         })
 350
 351     def decrypter(self, info_dict):
 352         _key_cache = {}
 353
 354         def _get_key(url):
 355             if url not in _key_cache:
 356                 _key_cache[url] = self.ydl.urlopen(self._prepare_url(info_dict, url)).read()
 357             return _key_cache[url]
 358
 359         def decrypt_fragment(fragment, frag_content):
 360             decrypt_info = fragment.get('decrypt_info')
 361             if not decrypt_info or decrypt_info['METHOD'] != 'AES-128':
 362                 return frag_content
 363             iv = decrypt_info.get('IV') or compat_struct_pack('>8xq', fragment['media_sequence'])
 364             decrypt_info['KEY'] = decrypt_info.get('KEY') or _get_key(info_dict.get('_decryption_key_url') or decrypt_info['URI'])
 365             # Don't decrypt the content in tests since the data is explicitly truncated and it's not to a valid block
 366             # size (see https://github.com/ytdl-org/youtube-dl/pull/27660). Tests only care that the correct data downloaded,
 367             # not what it decrypts to.
 368             if self.params.get('test', False):
 369                 return frag_content
 370             decrypted_data = aes_cbc_decrypt_bytes(frag_content, decrypt_info['KEY'], iv)
 371             return decrypted_data[:-decrypted_data[-1]]
 372
 373         return decrypt_fragment
 374
 375     def download_and_append_fragments_multiple(self, *args, pack_func=None, finish_func=None):
 376         '''
 377         @params (ctx1, fragments1, info_dict1), (ctx2, fragments2, info_dict2), ...
 378                 all args must be either tuple or list
 379         '''
 380         interrupt_trigger = [True]
 381         max_progress = len(args)
 382         if max_progress == 1:
 383             return self.download_and_append_fragments(*args[0], pack_func=pack_func, finish_func=finish_func)
 384         max_workers = self.params.get('concurrent_fragment_downloads', 1)
 385         if max_progress > 1:
 386             self._prepare_multiline_status(max_progress)
 387
 388         def thread_func(idx, ctx, fragments, info_dict, tpe):
 389             ctx['max_progress'] = max_progress
 390             ctx['progress_idx'] = idx
 391             return self.download_and_append_fragments(
 392                 ctx, fragments, info_dict, pack_func=pack_func, finish_func=finish_func,
 393                 tpe=tpe, interrupt_trigger=interrupt_trigger)
 394
 395         class FTPE(concurrent.futures.ThreadPoolExecutor):
 396             # has to stop this or it's going to wait on the worker thread itself
 397             def __exit__(self, exc_type, exc_val, exc_tb):
 398                 pass
 399
 400         spins = []
 401         if compat_os_name == 'nt':
 402             self.report_warning('Ctrl+C does not work on Windows when used with parallel threads. '
 403                                 'This is a known issue and patches are welcome')
 404         for idx, (ctx, fragments, info_dict) in enumerate(args):
 405             tpe = FTPE(math.ceil(max_workers / max_progress))
 406             job = tpe.submit(thread_func, idx, ctx, fragments, info_dict, tpe)
 407             spins.append((tpe, job))
 408
 409         result = True
 410         for tpe, job in spins:
 411             try:
 412                 result = result and job.result()
 413             except KeyboardInterrupt:
 414                 interrupt_trigger[0] = False
 415             finally:
 416                 tpe.shutdown(wait=True)
 417         if not interrupt_trigger[0]:
 418             raise KeyboardInterrupt()
 419         return result
 420
 421     def download_and_append_fragments(
 422             self, ctx, fragments, info_dict, *, pack_func=None, finish_func=None,
 423             tpe=None, interrupt_trigger=None):
 424         if not interrupt_trigger:
 425             interrupt_trigger = (True, )
 426
 427         fragment_retries = self.params.get('fragment_retries', 0)
 428         is_fatal = (
 429             ((lambda _: False) if info_dict.get('is_live') else (lambda idx: idx == 0))
 430             if self.params.get('skip_unavailable_fragments', True) else (lambda _: True))
 431
 432         if not pack_func:
 433             pack_func = lambda frag_content, _: frag_content
 434
 435         def download_fragment(fragment, ctx):
 436             frag_index = ctx['fragment_index'] = fragment['frag_index']
 437             if not interrupt_trigger[0]:
 438                 return False, frag_index
 439             headers = info_dict.get('http_headers', {}).copy()
 440             byte_range = fragment.get('byte_range')
 441             if byte_range:
 442                 headers['Range'] = 'bytes=%d-%d' % (byte_range['start'], byte_range['end'] - 1)
 443
 444             # Never skip the first fragment
 445             fatal = is_fatal(fragment.get('index') or (frag_index - 1))
 446             count, frag_content = 0, None
 447             while count <= fragment_retries:
 448                 try:
 449                     success, frag_content = self._download_fragment(ctx, fragment['url'], info_dict, headers)
 450                     if not success:
 451                         return False, frag_index
 452                     break
 453                 except (compat_urllib_error.HTTPError, http.client.IncompleteRead) as err:
 454                     # Unavailable (possibly temporary) fragments may be served.
 455                     # First we try to retry then either skip or abort.
 456                     # See https://github.com/ytdl-org/youtube-dl/issues/10165,
 457                     # https://github.com/ytdl-org/youtube-dl/issues/10448).
 458                     count += 1
 459                     if count <= fragment_retries:
 460                         self.report_retry_fragment(err, frag_index, count, fragment_retries)
 461                 except DownloadError:
 462                     # Don't retry fragment if error occurred during HTTP downloading
 463                     # itself since it has own retry settings
 464                     if not fatal:
 465                         break
 466                     raise
 467
 468             if count > fragment_retries:
 469                 if not fatal:
 470                     return False, frag_index
 471                 ctx['dest_stream'].close()
 472                 self.report_error('Giving up after %s fragment retries' % fragment_retries)
 473                 return False, frag_index
 474             return frag_content, frag_index
 475
 476         def append_fragment(frag_content, frag_index, ctx):
 477             if not frag_content:
 478                 if not is_fatal(frag_index - 1):
 479                     self.report_skip_fragment(frag_index, 'fragment not found')
 480                     return True
 481                 else:
 482                     ctx['dest_stream'].close()
 483                     self.report_error(
 484                         'fragment %s not found, unable to continue' % frag_index)
 485                     return False
 486             self._append_fragment(ctx, pack_func(frag_content, frag_index))
 487             return True
 488
 489         decrypt_fragment = self.decrypter(info_dict)
 490
 491         max_workers = math.ceil(
 492             self.params.get('concurrent_fragment_downloads', 1) / ctx.get('max_progress', 1))
 493         if can_threaded_download and max_workers > 1:
 494
 495             def _download_fragment(fragment):
 496                 ctx_copy = ctx.copy()
 497                 frag_content, frag_index = download_fragment(fragment, ctx_copy)
 498                 return fragment, frag_content, frag_index, ctx_copy.get('fragment_filename_sanitized')
 499
 500             self.report_warning('The download speed shown is only of one thread. This is a known issue and patches are welcome')
 501             with tpe or concurrent.futures.ThreadPoolExecutor(max_workers) as pool:
 502                 for fragment, frag_content, frag_index, frag_filename in pool.map(_download_fragment, fragments):
 503                     if not interrupt_trigger[0]:
 504                         break
 505                     ctx['fragment_filename_sanitized'] = frag_filename
 506                     ctx['fragment_index'] = frag_index
 507                     result = append_fragment(decrypt_fragment(fragment, frag_content), frag_index, ctx)
 508                     if not result:
 509                         return False
 510         else:
 511             for fragment in fragments:
 512                 if not interrupt_trigger[0]:
 513                     break
 514                 frag_content, frag_index = download_fragment(fragment, ctx)
 515                 result = append_fragment(decrypt_fragment(fragment, frag_content), frag_index, ctx)
 516                 if not result:
 517                     return False
 518
 519         if finish_func is not None:
 520             ctx['dest_stream'].write(finish_func())
 521             ctx['dest_stream'].flush()
 522         self._finish_frag_download(ctx, info_dict)
 523         return True