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