yt_dlp/downloader/http.py

   1 from __future__ import unicode_literals
   2
   3 import errno
   4 import os
   5 import socket
   6 import time
   7 import random
   8 import re
   9
  10 from .common import FileDownloader
  11 from ..compat import (
  12     compat_str,
  13     compat_urllib_error,
  14 )
  15 from ..utils import (
  16     ContentTooShortError,
  17     encodeFilename,
  18     int_or_none,
  19     sanitize_open,
  20     sanitized_Request,
  21     ThrottledDownload,
  22     write_xattr,
  23     XAttrMetadataError,
  24     XAttrUnavailableError,
  25 )
  26
  27
  28 class HttpFD(FileDownloader):
  29     def real_download(self, filename, info_dict):
  30         url = info_dict['url']
  31         request_data = info_dict.get('request_data', None)
  32
  33         class DownloadContext(dict):
  34             __getattr__ = dict.get
  35             __setattr__ = dict.__setitem__
  36             __delattr__ = dict.__delitem__
  37
  38         ctx = DownloadContext()
  39         ctx.filename = filename
  40         ctx.tmpfilename = self.temp_name(filename)
  41         ctx.stream = None
  42
  43         # Do not include the Accept-Encoding header
  44         headers = {'Youtubedl-no-compression': 'True'}
  45         add_headers = info_dict.get('http_headers')
  46         if add_headers:
  47             headers.update(add_headers)
  48
  49         is_test = self.params.get('test', False)
  50         chunk_size = self._TEST_FILE_SIZE if is_test else (
  51             self.params.get('http_chunk_size')
  52             or info_dict.get('downloader_options', {}).get('http_chunk_size')
  53             or 0)
  54
  55         ctx.open_mode = 'wb'
  56         ctx.resume_len = 0
  57         ctx.data_len = None
  58         ctx.block_size = self.params.get('buffersize', 1024)
  59         ctx.start_time = time.time()
  60         ctx.chunk_size = None
  61
  62         if self.params.get('continuedl', True):
  63             # Establish possible resume length
  64             if os.path.isfile(encodeFilename(ctx.tmpfilename)):
  65                 ctx.resume_len = os.path.getsize(
  66                     encodeFilename(ctx.tmpfilename))
  67
  68         ctx.is_resume = ctx.resume_len > 0
  69
  70         count = 0
  71         retries = self.params.get('retries', 0)
  72
  73         class SucceedDownload(Exception):
  74             pass
  75
  76         class RetryDownload(Exception):
  77             def __init__(self, source_error):
  78                 self.source_error = source_error
  79
  80         class NextFragment(Exception):
  81             pass
  82
  83         def set_range(req, start, end):
  84             range_header = 'bytes=%d-' % start
  85             if end:
  86                 range_header += compat_str(end)
  87             req.add_header('Range', range_header)
  88
  89         def establish_connection():
  90             ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
  91                               if not is_test and chunk_size else chunk_size)
  92             if ctx.resume_len > 0:
  93                 range_start = ctx.resume_len
  94                 if ctx.is_resume:
  95                     self.report_resuming_byte(ctx.resume_len)
  96                 ctx.open_mode = 'ab'
  97             elif ctx.chunk_size > 0:
  98                 range_start = 0
  99             else:
 100                 range_start = None
 101             ctx.is_resume = False
 102             range_end = range_start + ctx.chunk_size - 1 if ctx.chunk_size else None
 103             if range_end and ctx.data_len is not None and range_end >= ctx.data_len:
 104                 range_end = ctx.data_len - 1
 105             has_range = range_start is not None
 106             ctx.has_range = has_range
 107             request = sanitized_Request(url, request_data, headers)
 108             if has_range:
 109                 set_range(request, range_start, range_end)
 110             # Establish connection
 111             try:
 112                 try:
 113                     ctx.data = self.ydl.urlopen(request)
 114                 except (compat_urllib_error.URLError, ) as err:
 115                     # reason may not be available, e.g. for urllib2.HTTPError on python 2.6
 116                     reason = getattr(err, 'reason', None)
 117                     if isinstance(reason, socket.timeout):
 118                         raise RetryDownload(err)
 119                     raise err
 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                     if content_range:
 128                         content_range_m = re.search(r'bytes (\d+)-(\d+)?(?:/(\d+))?', content_range)
 129                         # Content-Range is present and matches requested Range, resume is possible
 130                         if content_range_m:
 131                             if range_start == int(content_range_m.group(1)):
 132                                 content_range_end = int_or_none(content_range_m.group(2))
 133                                 content_len = int_or_none(content_range_m.group(3))
 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.data_len = content_len
 143                                     return
 144                     # Content-Range is either not present or invalid. Assuming remote webserver is
 145                     # trying to send the whole file, resume is not possible, so wiping the local file
 146                     # and performing entire redownload
 147                     self.report_unable_to_resume()
 148                     ctx.resume_len = 0
 149                     ctx.open_mode = 'wb'
 150                 ctx.data_len = int_or_none(ctx.data.info().get('Content-length', None))
 151                 return
 152             except (compat_urllib_error.HTTPError, ) as err:
 153                 if err.code == 416:
 154                     # Unable to resume (requested range not satisfiable)
 155                     try:
 156                         # Open the connection again without the range header
 157                         ctx.data = self.ydl.urlopen(
 158                             sanitized_Request(url, request_data, headers))
 159                         content_length = ctx.data.info()['Content-Length']
 160                     except (compat_urllib_error.HTTPError, ) as err:
 161                         if err.code < 500 or err.code >= 600:
 162                             raise
 163                     else:
 164                         # Examine the reported length
 165                         if (content_length is not None
 166                                 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
 167                             # The file had already been fully downloaded.
 168                             # Explanation to the above condition: in issue #175 it was revealed that
 169                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 170                             # changing the file size slightly and causing problems for some users. So
 171                             # I decided to implement a suggested change and consider the file
 172                             # completely downloaded if the file size differs less than 100 bytes from
 173                             # the one in the hard drive.
 174                             self.report_file_already_downloaded(ctx.filename)
 175                             self.try_rename(ctx.tmpfilename, ctx.filename)
 176                             self._hook_progress({
 177                                 'filename': ctx.filename,
 178                                 'status': 'finished',
 179                                 'downloaded_bytes': ctx.resume_len,
 180                                 'total_bytes': ctx.resume_len,
 181                             }, info_dict)
 182                             raise SucceedDownload()
 183                         else:
 184                             # The length does not match, we start the download over
 185                             self.report_unable_to_resume()
 186                             ctx.resume_len = 0
 187                             ctx.open_mode = 'wb'
 188                             return
 189                 elif err.code < 500 or err.code >= 600:
 190                     # Unexpected HTTP error
 191                     raise
 192                 raise RetryDownload(err)
 193             except socket.error as err:
 194                 if err.errno != errno.ECONNRESET:
 195                     # Connection reset is no problem, just retry
 196                     raise
 197                 raise RetryDownload(err)
 198
 199         def download():
 200             data_len = ctx.data.info().get('Content-length', None)
 201
 202             # Range HTTP header may be ignored/unsupported by a webserver
 203             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 204             # However, for a test we still would like to download just a piece of a file.
 205             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 206             # block size when downloading a file.
 207             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 208                 data_len = self._TEST_FILE_SIZE
 209
 210             if data_len is not None:
 211                 data_len = int(data_len) + ctx.resume_len
 212                 min_data_len = self.params.get('min_filesize')
 213                 max_data_len = self.params.get('max_filesize')
 214                 if min_data_len is not None and data_len < min_data_len:
 215                     self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
 216                     return False
 217                 if max_data_len is not None and data_len > max_data_len:
 218                     self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
 219                     return False
 220
 221             byte_counter = 0 + ctx.resume_len
 222             block_size = ctx.block_size
 223             start = time.time()
 224
 225             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 226             now = None  # needed for slow_down() in the first loop run
 227             before = start  # start measuring
 228             throttle_start = None
 229
 230             def retry(e):
 231                 to_stdout = ctx.tmpfilename == '-'
 232                 if ctx.stream is not None:
 233                     if not to_stdout:
 234                         ctx.stream.close()
 235                     ctx.stream = None
 236                 ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
 237                 raise RetryDownload(e)
 238
 239             while True:
 240                 try:
 241                     # Download and write
 242                     data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
 243                 # socket.timeout is a subclass of socket.error but may not have
 244                 # errno set
 245                 except socket.timeout as e:
 246                     retry(e)
 247                 except socket.error as e:
 248                     # SSLError on python 2 (inherits socket.error) may have
 249                     # no errno set but this error message
 250                     if e.errno in (errno.ECONNRESET, errno.ETIMEDOUT) or getattr(e, 'message', None) == 'The read operation timed out':
 251                         retry(e)
 252                     raise
 253
 254                 byte_counter += len(data_block)
 255
 256                 # exit loop when download is finished
 257                 if len(data_block) == 0:
 258                     break
 259
 260                 # Open destination file just in time
 261                 if ctx.stream is None:
 262                     try:
 263                         ctx.stream, ctx.tmpfilename = sanitize_open(
 264                             ctx.tmpfilename, ctx.open_mode)
 265                         assert ctx.stream is not None
 266                         ctx.filename = self.undo_temp_name(ctx.tmpfilename)
 267                         self.report_destination(ctx.filename)
 268                     except (OSError, IOError) as err:
 269                         self.report_error('unable to open for writing: %s' % str(err))
 270                         return False
 271
 272                     if self.params.get('xattr_set_filesize', False) and data_len is not None:
 273                         try:
 274                             write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8'))
 275                         except (XAttrUnavailableError, XAttrMetadataError) as err:
 276                             self.report_error('unable to set filesize xattr: %s' % str(err))
 277
 278                 try:
 279                     ctx.stream.write(data_block)
 280                 except (IOError, OSError) as err:
 281                     self.to_stderr('\n')
 282                     self.report_error('unable to write data: %s' % str(err))
 283                     return False
 284
 285                 # Apply rate limit
 286                 self.slow_down(start, now, byte_counter - ctx.resume_len)
 287
 288                 # end measuring of one loop run
 289                 now = time.time()
 290                 after = now
 291
 292                 # Adjust block size
 293                 if not self.params.get('noresizebuffer', False):
 294                     block_size = self.best_block_size(after - before, len(data_block))
 295
 296                 before = after
 297
 298                 # Progress message
 299                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
 300                 if ctx.data_len is None:
 301                     eta = None
 302                 else:
 303                     eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
 304
 305                 self._hook_progress({
 306                     'status': 'downloading',
 307                     'downloaded_bytes': byte_counter,
 308                     'total_bytes': ctx.data_len,
 309                     'tmpfilename': ctx.tmpfilename,
 310                     'filename': ctx.filename,
 311                     'eta': eta,
 312                     'speed': speed,
 313                     'elapsed': now - ctx.start_time,
 314                     'ctx_id': info_dict.get('ctx_id'),
 315                 }, info_dict)
 316
 317                 if data_len is not None and byte_counter == data_len:
 318                     break
 319
 320                 if speed and speed < (self.params.get('throttledratelimit') or 0):
 321                     # The speed must stay below the limit for 3 seconds
 322                     # This prevents raising error when the speed temporarily goes down
 323                     if throttle_start is None:
 324                         throttle_start = now
 325                     elif now - throttle_start > 3:
 326                         if ctx.stream is not None and ctx.tmpfilename != '-':
 327                             ctx.stream.close()
 328                         raise ThrottledDownload()
 329                 else:
 330                     throttle_start = None
 331
 332             if not is_test and ctx.chunk_size and ctx.data_len is not None and byte_counter < ctx.data_len:
 333                 ctx.resume_len = byte_counter
 334                 # ctx.block_size = block_size
 335                 raise NextFragment()
 336
 337             if ctx.stream is None:
 338                 self.to_stderr('\n')
 339                 self.report_error('Did not get any data blocks')
 340                 return False
 341             if ctx.tmpfilename != '-':
 342                 ctx.stream.close()
 343
 344             if data_len is not None and byte_counter != data_len:
 345                 err = ContentTooShortError(byte_counter, int(data_len))
 346                 if count <= retries:
 347                     retry(err)
 348                 raise 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.info().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         while count <= retries:
 368             try:
 369                 establish_connection()
 370                 return download()
 371             except RetryDownload as e:
 372                 count += 1
 373                 if count <= retries:
 374                     self.report_retry(e.source_error, count, retries)
 375                 continue
 376             except NextFragment:
 377                 continue
 378             except SucceedDownload:
 379                 return True
 380
 381         self.report_error('giving up after %s retries' % retries)
 382         return False