yt_dlp/downloader/http.py

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