yt_dlp/downloader/http.py

   1 import os
   2 import random
   3 import socket
   4 import ssl
   5 import time
   6
   7 from .common import FileDownloader
   8 from ..compat import compat_http_client, compat_urllib_error
   9 from ..utils import (
  10     ContentTooShortError,
  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 RESPONSE_READ_EXCEPTIONS = (
  23     TimeoutError,
  24     socket.timeout,  # compat: py < 3.10
  25     ConnectionError,
  26     ssl.SSLError,
  27     compat_http_client.HTTPException
  28 )
  29
  30
  31 class HttpFD(FileDownloader):
  32     def real_download(self, filename, info_dict):
  33         url = info_dict['url']
  34         request_data = info_dict.get('request_data', None)
  35
  36         class DownloadContext(dict):
  37             __getattr__ = dict.get
  38             __setattr__ = dict.__setitem__
  39             __delattr__ = dict.__delitem__
  40
  41         ctx = DownloadContext()
  42         ctx.filename = filename
  43         ctx.tmpfilename = self.temp_name(filename)
  44         ctx.stream = None
  45
  46         # Do not include the Accept-Encoding header
  47         headers = {'Youtubedl-no-compression': 'True'}
  48         add_headers = info_dict.get('http_headers')
  49         if add_headers:
  50             headers.update(add_headers)
  51
  52         is_test = self.params.get('test', False)
  53         chunk_size = self._TEST_FILE_SIZE if is_test else (
  54             self.params.get('http_chunk_size')
  55             or info_dict.get('downloader_options', {}).get('http_chunk_size')
  56             or 0)
  57
  58         ctx.open_mode = 'wb'
  59         ctx.resume_len = 0
  60         ctx.block_size = self.params.get('buffersize', 1024)
  61         ctx.start_time = time.time()
  62
  63         # parse given Range
  64         req_start, req_end, _ = parse_http_range(headers.get('Range'))
  65
  66         if self.params.get('continuedl', True):
  67             # Establish possible resume length
  68             if os.path.isfile(encodeFilename(ctx.tmpfilename)):
  69                 ctx.resume_len = os.path.getsize(
  70                     encodeFilename(ctx.tmpfilename))
  71
  72         ctx.is_resume = ctx.resume_len > 0
  73
  74         count = 0
  75         retries = self.params.get('retries', 0)
  76
  77         class SucceedDownload(Exception):
  78             pass
  79
  80         class RetryDownload(Exception):
  81             def __init__(self, source_error):
  82                 self.source_error = source_error
  83
  84         class NextFragment(Exception):
  85             pass
  86
  87         def establish_connection():
  88             ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
  89                               if not is_test and chunk_size else chunk_size)
  90             if ctx.resume_len > 0:
  91                 range_start = ctx.resume_len
  92                 if req_start is not None:
  93                     # offset the beginning of Range to be within request
  94                     range_start += req_start
  95                 if ctx.is_resume:
  96                     self.report_resuming_byte(ctx.resume_len)
  97                 ctx.open_mode = 'ab'
  98             elif req_start is not None:
  99                 range_start = req_start
 100             elif ctx.chunk_size > 0:
 101                 range_start = 0
 102             else:
 103                 range_start = None
 104             ctx.is_resume = False
 105
 106             if ctx.chunk_size:
 107                 chunk_aware_end = range_start + ctx.chunk_size - 1
 108                 # we're not allowed to download outside Range
 109                 range_end = chunk_aware_end if req_end is None else min(chunk_aware_end, req_end)
 110             elif req_end is not None:
 111                 # there's no need for chunked downloads, so download until the end of Range
 112                 range_end = req_end
 113             else:
 114                 range_end = None
 115
 116             if try_call(lambda: range_start > range_end):
 117                 ctx.resume_len = 0
 118                 ctx.open_mode = 'wb'
 119                 raise RetryDownload(Exception(f'Conflicting range. (start={range_start} > end={range_end})'))
 120
 121             if try_call(lambda: range_end >= ctx.content_len):
 122                 range_end = ctx.content_len - 1
 123
 124             request = sanitized_Request(url, request_data, headers)
 125             has_range = range_start is not None
 126             if has_range:
 127                 request.add_header('Range', f'bytes={int(range_start)}-{int_or_none(range_end) or ""}')
 128             # Establish connection
 129             try:
 130                 ctx.data = self.ydl.urlopen(request)
 131                 # When trying to resume, Content-Range HTTP header of response has to be checked
 132                 # to match the value of requested Range HTTP header. This is due to a webservers
 133                 # that don't support resuming and serve a whole file with no Content-Range
 134                 # set in response despite of requested Range (see
 135                 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
 136                 if has_range:
 137                     content_range = ctx.data.headers.get('Content-Range')
 138                     content_range_start, content_range_end, content_len = parse_http_range(content_range)
 139                     # Content-Range is present and matches requested Range, resume is possible
 140                     if range_start == content_range_start and (
 141                             # Non-chunked download
 142                             not ctx.chunk_size
 143                             # Chunked download and requested piece or
 144                             # its part is promised to be served
 145                             or content_range_end == range_end
 146                             or content_len < range_end):
 147                         ctx.content_len = content_len
 148                         if content_len or req_end:
 149                             ctx.data_len = min(content_len or req_end, req_end or content_len) - (req_start or 0)
 150                         return
 151                     # Content-Range is either not present or invalid. Assuming remote webserver is
 152                     # trying to send the whole file, resume is not possible, so wiping the local file
 153                     # and performing entire redownload
 154                     self.report_unable_to_resume()
 155                     ctx.resume_len = 0
 156                     ctx.open_mode = 'wb'
 157                 ctx.data_len = ctx.content_len = int_or_none(ctx.data.info().get('Content-length', None))
 158             except compat_urllib_error.HTTPError as err:
 159                 if err.code == 416:
 160                     # Unable to resume (requested range not satisfiable)
 161                     try:
 162                         # Open the connection again without the range header
 163                         ctx.data = self.ydl.urlopen(
 164                             sanitized_Request(url, request_data, headers))
 165                         content_length = ctx.data.info()['Content-Length']
 166                     except compat_urllib_error.HTTPError as err:
 167                         if err.code < 500 or err.code >= 600:
 168                             raise
 169                     else:
 170                         # Examine the reported length
 171                         if (content_length is not None
 172                                 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
 173                             # The file had already been fully downloaded.
 174                             # Explanation to the above condition: in issue #175 it was revealed that
 175                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 176                             # changing the file size slightly and causing problems for some users. So
 177                             # I decided to implement a suggested change and consider the file
 178                             # completely downloaded if the file size differs less than 100 bytes from
 179                             # the one in the hard drive.
 180                             self.report_file_already_downloaded(ctx.filename)
 181                             self.try_rename(ctx.tmpfilename, ctx.filename)
 182                             self._hook_progress({
 183                                 'filename': ctx.filename,
 184                                 'status': 'finished',
 185                                 'downloaded_bytes': ctx.resume_len,
 186                                 'total_bytes': ctx.resume_len,
 187                             }, info_dict)
 188                             raise SucceedDownload()
 189                         else:
 190                             # The length does not match, we start the download over
 191                             self.report_unable_to_resume()
 192                             ctx.resume_len = 0
 193                             ctx.open_mode = 'wb'
 194                             return
 195                 elif err.code < 500 or err.code >= 600:
 196                     # Unexpected HTTP error
 197                     raise
 198                 raise RetryDownload(err)
 199             except compat_urllib_error.URLError as err:
 200                 if isinstance(err.reason, ssl.CertificateError):
 201                     raise
 202                 raise RetryDownload(err)
 203             # In urllib.request.AbstractHTTPHandler, the response is partially read on request.
 204             # Any errors that occur during this will not be wrapped by URLError
 205             except RESPONSE_READ_EXCEPTIONS as err:
 206                 raise RetryDownload(err)
 207
 208         def download():
 209             data_len = ctx.data.info().get('Content-length', None)
 210
 211             # Range HTTP header may be ignored/unsupported by a webserver
 212             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 213             # However, for a test we still would like to download just a piece of a file.
 214             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 215             # block size when downloading a file.
 216             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 217                 data_len = self._TEST_FILE_SIZE
 218
 219             if data_len is not None:
 220                 data_len = int(data_len) + ctx.resume_len
 221                 min_data_len = self.params.get('min_filesize')
 222                 max_data_len = self.params.get('max_filesize')
 223                 if min_data_len is not None and data_len < min_data_len:
 224                     self.to_screen(
 225                         f'\r[download] File is smaller than min-filesize ({data_len} bytes < {min_data_len} bytes). Aborting.')
 226                     return False
 227                 if max_data_len is not None and data_len > max_data_len:
 228                     self.to_screen(
 229                         f'\r[download] File is larger than max-filesize ({data_len} bytes > {max_data_len} bytes). Aborting.')
 230                     return False
 231
 232             byte_counter = 0 + ctx.resume_len
 233             block_size = ctx.block_size
 234             start = time.time()
 235
 236             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 237             now = None  # needed for slow_down() in the first loop run
 238             before = start  # start measuring
 239
 240             def retry(e):
 241                 to_stdout = ctx.tmpfilename == '-'
 242                 if ctx.stream is not None:
 243                     if not to_stdout:
 244                         ctx.stream.close()
 245                     ctx.stream = None
 246                 ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
 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 RESPONSE_READ_EXCEPTIONS 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('unable to open for writing: %s' % str(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('unable to set filesize xattr: %s' % str(err))
 279
 280                 try:
 281                     ctx.stream.write(data_block)
 282                 except OSError as err:
 283                     self.to_stderr('\n')
 284                     self.report_error('unable to write data: %s' % str(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 not is_test and ctx.chunk_size and ctx.content_len is not None and byte_counter < ctx.content_len:
 335                 ctx.resume_len = byte_counter
 336                 # ctx.block_size = block_size
 337                 raise NextFragment()
 338
 339             if ctx.stream is None:
 340                 self.to_stderr('\n')
 341                 self.report_error('Did not get any data blocks')
 342                 return False
 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                 if count <= retries:
 349                     retry(err)
 350                 raise err
 351
 352             self.try_rename(ctx.tmpfilename, ctx.filename)
 353
 354             # Update file modification time
 355             if self.params.get('updatetime', True):
 356                 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
 357
 358             self._hook_progress({
 359                 'downloaded_bytes': byte_counter,
 360                 'total_bytes': byte_counter,
 361                 'filename': ctx.filename,
 362                 'status': 'finished',
 363                 'elapsed': time.time() - ctx.start_time,
 364                 'ctx_id': info_dict.get('ctx_id'),
 365             }, info_dict)
 366
 367             return True
 368
 369         while count <= retries:
 370             try:
 371                 establish_connection()
 372                 return download()
 373             except RetryDownload as e:
 374                 count += 1
 375                 if count <= retries:
 376                     self.report_retry(e.source_error, count, retries)
 377                 else:
 378                     self.to_screen(f'[download] Got server HTTP error: {e.source_error}')
 379                 continue
 380             except NextFragment:
 381                 continue
 382             except SucceedDownload:
 383                 return True
 384
 385         self.report_error('giving up after %s retries' % retries)
 386         return False