yt_dlp/downloader/http.py

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