yt_dlp/downloader/http.py

   1 import os
   2 import random
   3 import time
   4
   5 from .common import FileDownloader
   6 from ..networking import Request
   7 from ..networking.exceptions import (
   8     CertificateVerifyError,
   9     HTTPError,
  10     TransportError,
  11 )
  12 from ..utils import (
  13     ContentTooShortError,
  14     RetryManager,
  15     ThrottledDownload,
  16     XAttrMetadataError,
  17     XAttrUnavailableError,
  18     int_or_none,
  19     parse_http_range,
  20     try_call,
  21     write_xattr,
  22 )
  23 from ..utils.networking import HTTPHeaderDict
  24
  25
  26 class HttpFD(FileDownloader):
  27     def real_download(self, filename, info_dict):
  28         url = info_dict['url']
  29         request_data = info_dict.get('request_data', None)
  30
  31         class DownloadContext(dict):
  32             __getattr__ = dict.get
  33             __setattr__ = dict.__setitem__
  34             __delattr__ = dict.__delitem__
  35
  36         ctx = DownloadContext()
  37         ctx.filename = filename
  38         ctx.tmpfilename = self.temp_name(filename)
  39         ctx.stream = None
  40
  41         # Disable compression
  42         headers = HTTPHeaderDict({'Accept-Encoding': 'identity'}, info_dict.get('http_headers'))
  43
  44         is_test = self.params.get('test', False)
  45         chunk_size = self._TEST_FILE_SIZE if is_test else (
  46             self.params.get('http_chunk_size')
  47             or info_dict.get('downloader_options', {}).get('http_chunk_size')
  48             or 0)
  49
  50         ctx.open_mode = 'wb'
  51         ctx.resume_len = 0
  52         ctx.block_size = self.params.get('buffersize', 1024)
  53         ctx.start_time = time.time()
  54
  55         # parse given Range
  56         req_start, req_end, _ = parse_http_range(headers.get('Range'))
  57
  58         if self.params.get('continuedl', True):
  59             # Establish possible resume length
  60             if os.path.isfile(ctx.tmpfilename):
  61                 ctx.resume_len = os.path.getsize(ctx.tmpfilename)
  62
  63         ctx.is_resume = ctx.resume_len > 0
  64
  65         class SucceedDownload(Exception):
  66             pass
  67
  68         class RetryDownload(Exception):
  69             def __init__(self, source_error):
  70                 self.source_error = source_error
  71
  72         class NextFragment(Exception):
  73             pass
  74
  75         def establish_connection():
  76             ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
  77                               if not is_test and chunk_size else chunk_size)
  78             if ctx.resume_len > 0:
  79                 range_start = ctx.resume_len
  80                 if req_start is not None:
  81                     # offset the beginning of Range to be within request
  82                     range_start += req_start
  83                 if ctx.is_resume:
  84                     self.report_resuming_byte(ctx.resume_len)
  85                 ctx.open_mode = 'ab'
  86             elif req_start is not None:
  87                 range_start = req_start
  88             elif ctx.chunk_size > 0:
  89                 range_start = 0
  90             else:
  91                 range_start = None
  92             ctx.is_resume = False
  93
  94             if ctx.chunk_size:
  95                 chunk_aware_end = range_start + ctx.chunk_size - 1
  96                 # we're not allowed to download outside Range
  97                 range_end = chunk_aware_end if req_end is None else min(chunk_aware_end, req_end)
  98             elif req_end is not None:
  99                 # there's no need for chunked downloads, so download until the end of Range
 100                 range_end = req_end
 101             else:
 102                 range_end = None
 103
 104             if try_call(lambda: range_start > range_end):
 105                 ctx.resume_len = 0
 106                 ctx.open_mode = 'wb'
 107                 raise RetryDownload(Exception(f'Conflicting range. (start={range_start} > end={range_end})'))
 108
 109             if try_call(lambda: range_end >= ctx.content_len):
 110                 range_end = ctx.content_len - 1
 111
 112             request = Request(url, request_data, headers)
 113             has_range = range_start is not None
 114             if has_range:
 115                 request.headers['Range'] = f'bytes={int(range_start)}-{int_or_none(range_end) or ""}'
 116             # Establish connection
 117             try:
 118                 ctx.data = self.ydl.urlopen(request)
 119                 # When trying to resume, Content-Range HTTP header of response has to be checked
 120                 # to match the value of requested Range HTTP header. This is due to a webservers
 121                 # that don't support resuming and serve a whole file with no Content-Range
 122                 # set in response despite of requested Range (see
 123                 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
 124                 if has_range:
 125                     content_range = ctx.data.headers.get('Content-Range')
 126                     content_range_start, content_range_end, content_len = parse_http_range(content_range)
 127                     # Content-Range is present and matches requested Range, resume is possible
 128                     if range_start == content_range_start and (
 129                             # Non-chunked download
 130                             not ctx.chunk_size
 131                             # Chunked download and requested piece or
 132                             # its part is promised to be served
 133                             or content_range_end == range_end
 134                             or content_len < range_end):
 135                         ctx.content_len = content_len
 136                         if content_len or req_end:
 137                             ctx.data_len = min(content_len or req_end, req_end or content_len) - (req_start or 0)
 138                         return
 139                     # Content-Range is either not present or invalid. Assuming remote webserver is
 140                     # trying to send the whole file, resume is not possible, so wiping the local file
 141                     # and performing entire redownload
 142                     elif range_start > 0:
 143                         self.report_unable_to_resume()
 144                     ctx.resume_len = 0
 145                     ctx.open_mode = 'wb'
 146                 ctx.data_len = ctx.content_len = int_or_none(ctx.data.headers.get('Content-length', None))
 147             except HTTPError as err:
 148                 if err.status == 416:
 149                     # Unable to resume (requested range not satisfiable)
 150                     try:
 151                         # Open the connection again without the range header
 152                         ctx.data = self.ydl.urlopen(
 153                             Request(url, request_data, headers))
 154                         content_length = ctx.data.headers['Content-Length']
 155                     except HTTPError as err:
 156                         if err.status < 500 or err.status >= 600:
 157                             raise
 158                     else:
 159                         # Examine the reported length
 160                         if (content_length is not None
 161                                 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
 162                             # The file had already been fully downloaded.
 163                             # Explanation to the above condition: in issue #175 it was revealed that
 164                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 165                             # changing the file size slightly and causing problems for some users. So
 166                             # I decided to implement a suggested change and consider the file
 167                             # completely downloaded if the file size differs less than 100 bytes from
 168                             # the one in the hard drive.
 169                             self.report_file_already_downloaded(ctx.filename)
 170                             self.try_rename(ctx.tmpfilename, ctx.filename)
 171                             self._hook_progress({
 172                                 'filename': ctx.filename,
 173                                 'status': 'finished',
 174                                 'downloaded_bytes': ctx.resume_len,
 175                                 'total_bytes': ctx.resume_len,
 176                             }, info_dict)
 177                             raise SucceedDownload
 178                         else:
 179                             # The length does not match, we start the download over
 180                             self.report_unable_to_resume()
 181                             ctx.resume_len = 0
 182                             ctx.open_mode = 'wb'
 183                             return
 184                 elif err.status < 500 or err.status >= 600:
 185                     # Unexpected HTTP error
 186                     raise
 187                 raise RetryDownload(err)
 188             except CertificateVerifyError:
 189                 raise
 190             except TransportError as err:
 191                 raise RetryDownload(err)
 192
 193         def close_stream():
 194             if ctx.stream is not None:
 195                 if ctx.tmpfilename != '-':
 196                     ctx.stream.close()
 197                 ctx.stream = None
 198
 199         def download():
 200             data_len = ctx.data.headers.get('Content-length')
 201
 202             if ctx.data.headers.get('Content-encoding'):
 203                 # Content-encoding is present, Content-length is not reliable anymore as we are
 204                 # doing auto decompression. (See: https://github.com/yt-dlp/yt-dlp/pull/6176)
 205                 data_len = None
 206
 207             # Range HTTP header may be ignored/unsupported by a webserver
 208             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 209             # However, for a test we still would like to download just a piece of a file.
 210             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 211             # block size when downloading a file.
 212             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 213                 data_len = self._TEST_FILE_SIZE
 214
 215             if data_len is not None:
 216                 data_len = int(data_len) + ctx.resume_len
 217                 min_data_len = self.params.get('min_filesize')
 218                 max_data_len = self.params.get('max_filesize')
 219                 if min_data_len is not None and data_len < min_data_len:
 220                     self.to_screen(
 221                         f'\r[download] File is smaller than min-filesize ({data_len} bytes < {min_data_len} bytes). Aborting.')
 222                     return False
 223                 if max_data_len is not None and data_len > max_data_len:
 224                     self.to_screen(
 225                         f'\r[download] File is larger than max-filesize ({data_len} bytes > {max_data_len} bytes). Aborting.')
 226                     return False
 227
 228             byte_counter = 0 + ctx.resume_len
 229             block_size = ctx.block_size
 230             start = time.time()
 231
 232             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 233             now = None  # needed for slow_down() in the first loop run
 234             before = start  # start measuring
 235
 236             def retry(e):
 237                 close_stream()
 238                 if ctx.tmpfilename == '-':
 239                     ctx.resume_len = byte_counter
 240                 else:
 241                     try:
 242                         ctx.resume_len = os.path.getsize(ctx.tmpfilename)
 243                     except FileNotFoundError:
 244                         ctx.resume_len = 0
 245                 raise RetryDownload(e)
 246
 247             while True:
 248                 try:
 249                     # Download and write
 250                     data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
 251                 except TransportError as err:
 252                     retry(err)
 253
 254                 byte_counter += len(data_block)
 255
 256                 # exit loop when download is finished
 257                 if len(data_block) == 0:
 258                     break
 259
 260                 # Open destination file just in time
 261                 if ctx.stream is None:
 262                     try:
 263                         ctx.stream, ctx.tmpfilename = self.sanitize_open(
 264                             ctx.tmpfilename, ctx.open_mode)
 265                         assert ctx.stream is not None
 266                         ctx.filename = self.undo_temp_name(ctx.tmpfilename)
 267                         self.report_destination(ctx.filename)
 268                     except OSError as err:
 269                         self.report_error(f'unable to open for writing: {err}')
 270                         return False
 271
 272                     if self.params.get('xattr_set_filesize', False) and data_len is not None:
 273                         try:
 274                             write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode())
 275                         except (XAttrUnavailableError, XAttrMetadataError) as err:
 276                             self.report_error(f'unable to set filesize xattr: {err}')
 277
 278                 try:
 279                     ctx.stream.write(data_block)
 280                 except OSError as err:
 281                     self.to_stderr('\n')
 282                     self.report_error(f'unable to write data: {err}')
 283                     return False
 284
 285                 # Apply rate limit
 286                 self.slow_down(start, now, byte_counter - ctx.resume_len)
 287
 288                 # end measuring of one loop run
 289                 now = time.time()
 290                 after = now
 291
 292                 # Adjust block size
 293                 if not self.params.get('noresizebuffer', False):
 294                     block_size = self.best_block_size(after - before, len(data_block))
 295
 296                 before = after
 297
 298                 # Progress message
 299                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
 300                 if ctx.data_len is None:
 301                     eta = None
 302                 else:
 303                     eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
 304
 305                 self._hook_progress({
 306                     'status': 'downloading',
 307                     'downloaded_bytes': byte_counter,
 308                     'total_bytes': ctx.data_len,
 309                     'tmpfilename': ctx.tmpfilename,
 310                     'filename': ctx.filename,
 311                     'eta': eta,
 312                     'speed': speed,
 313                     'elapsed': now - ctx.start_time,
 314                     'ctx_id': info_dict.get('ctx_id'),
 315                 }, info_dict)
 316
 317                 if data_len is not None and byte_counter == data_len:
 318                     break
 319
 320                 if speed and speed < (self.params.get('throttledratelimit') or 0):
 321                     # The speed must stay below the limit for 3 seconds
 322                     # This prevents raising error when the speed temporarily goes down
 323                     if ctx.throttle_start is None:
 324                         ctx.throttle_start = now
 325                     elif now - ctx.throttle_start > 3:
 326                         if ctx.stream is not None and ctx.tmpfilename != '-':
 327                             ctx.stream.close()
 328                         raise ThrottledDownload
 329                 elif speed:
 330                     ctx.throttle_start = None
 331
 332             if ctx.stream is None:
 333                 self.to_stderr('\n')
 334                 self.report_error('Did not get any data blocks')
 335                 return False
 336
 337             if not is_test and ctx.chunk_size and ctx.content_len is not None and byte_counter < ctx.content_len:
 338                 ctx.resume_len = byte_counter
 339                 raise NextFragment
 340
 341             if ctx.tmpfilename != '-':
 342                 ctx.stream.close()
 343
 344             if data_len is not None and byte_counter != data_len:
 345                 err = ContentTooShortError(byte_counter, int(data_len))
 346                 retry(err)
 347
 348             self.try_rename(ctx.tmpfilename, ctx.filename)
 349
 350             # Update file modification time
 351             if self.params.get('updatetime', True):
 352                 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.headers.get('last-modified', None))
 353
 354             self._hook_progress({
 355                 'downloaded_bytes': byte_counter,
 356                 'total_bytes': byte_counter,
 357                 'filename': ctx.filename,
 358                 'status': 'finished',
 359                 'elapsed': time.time() - ctx.start_time,
 360                 'ctx_id': info_dict.get('ctx_id'),
 361             }, info_dict)
 362
 363             return True
 364
 365         for retry in RetryManager(self.params.get('retries'), self.report_retry):
 366             try:
 367                 establish_connection()
 368                 return download()
 369             except RetryDownload as err:
 370                 retry.error = err.source_error
 371                 continue
 372             except NextFragment:
 373                 retry.error = None
 374                 retry.attempt -= 1
 375                 continue
 376             except SucceedDownload:
 377                 return True
 378             except:  # noqa: E722
 379                 close_stream()
 380                 raise
 381         return False