yt_dlp/downloader/http.py

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