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_str,
  11     compat_urllib_error,
  12     compat_http_client
  13 )
  14 from ..utils import (
  15     ContentTooShortError,
  16     encodeFilename,
  17     int_or_none,
  18     parse_http_range,
  19     sanitized_Request,
  20     ThrottledDownload,
  21     try_call,
  22     write_xattr,
  23     XAttrMetadataError,
  24     XAttrUnavailableError,
  25 )
  26
  27 RESPONSE_READ_EXCEPTIONS = (TimeoutError, ConnectionError, ssl.SSLError, compat_http_client.HTTPException)
  28
  29
  30 class HttpFD(FileDownloader):
  31     def real_download(self, filename, info_dict):
  32         url = info_dict['url']
  33         request_data = info_dict.get('request_data', None)
  34
  35         class DownloadContext(dict):
  36             __getattr__ = dict.get
  37             __setattr__ = dict.__setitem__
  38             __delattr__ = dict.__delitem__
  39
  40         ctx = DownloadContext()
  41         ctx.filename = filename
  42         ctx.tmpfilename = self.temp_name(filename)
  43         ctx.stream = None
  44
  45         # Do not include the Accept-Encoding header
  46         headers = {'Youtubedl-no-compression': 'True'}
  47         add_headers = info_dict.get('http_headers')
  48         if add_headers:
  49             headers.update(add_headers)
  50
  51         is_test = self.params.get('test', False)
  52         chunk_size = self._TEST_FILE_SIZE if is_test else (
  53             self.params.get('http_chunk_size')
  54             or info_dict.get('downloader_options', {}).get('http_chunk_size')
  55             or 0)
  56
  57         ctx.open_mode = 'wb'
  58         ctx.resume_len = 0
  59         ctx.block_size = self.params.get('buffersize', 1024)
  60         ctx.start_time = time.time()
  61         ctx.chunk_size = None
  62         throttle_start = None
  63
  64         # parse given Range
  65         req_start, req_end, _ = parse_http_range(headers.get('Range'))
  66
  67         if self.params.get('continuedl', True):
  68             # Establish possible resume length
  69             if os.path.isfile(encodeFilename(ctx.tmpfilename)):
  70                 ctx.resume_len = os.path.getsize(
  71                     encodeFilename(ctx.tmpfilename))
  72
  73         ctx.is_resume = ctx.resume_len > 0
  74
  75         count = 0
  76         retries = self.params.get('retries', 0)
  77
  78         class SucceedDownload(Exception):
  79             pass
  80
  81         class RetryDownload(Exception):
  82             def __init__(self, source_error):
  83                 self.source_error = source_error
  84
  85         class NextFragment(Exception):
  86             pass
  87
  88         def set_range(req, start, end):
  89             range_header = 'bytes=%d-' % start
  90             if end:
  91                 range_header += compat_str(end)
  92             req.add_header('Range', range_header)
  93
  94         def establish_connection():
  95             ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
  96                               if not is_test and chunk_size else chunk_size)
  97             if ctx.resume_len > 0:
  98                 range_start = ctx.resume_len
  99                 if req_start is not None:
 100                     # offset the beginning of Range to be within request
 101                     range_start += req_start
 102                 if ctx.is_resume:
 103                     self.report_resuming_byte(ctx.resume_len)
 104                 ctx.open_mode = 'ab'
 105             elif req_start is not None:
 106                 range_start = req_start
 107             elif ctx.chunk_size > 0:
 108                 range_start = 0
 109             else:
 110                 range_start = None
 111             ctx.is_resume = False
 112
 113             if ctx.chunk_size:
 114                 chunk_aware_end = range_start + ctx.chunk_size - 1
 115                 # we're not allowed to download outside Range
 116                 range_end = chunk_aware_end if req_end is None else min(chunk_aware_end, req_end)
 117             elif req_end is not None:
 118                 # there's no need for chunked downloads, so download until the end of Range
 119                 range_end = req_end
 120             else:
 121                 range_end = None
 122
 123             if try_call(lambda: range_start > range_end):
 124                 ctx.resume_len = 0
 125                 ctx.open_mode = 'wb'
 126                 raise RetryDownload(Exception(f'Conflicting range. (start={range_start} > end={range_end})'))
 127
 128             if try_call(lambda: range_end >= ctx.content_len):
 129                 range_end = ctx.content_len - 1
 130
 131             request = sanitized_Request(url, request_data, headers)
 132             has_range = range_start is not None
 133             if has_range:
 134                 set_range(request, range_start, range_end)
 135             # Establish connection
 136             try:
 137                 ctx.data = self.ydl.urlopen(request)
 138                 # When trying to resume, Content-Range HTTP header of response has to be checked
 139                 # to match the value of requested Range HTTP header. This is due to a webservers
 140                 # that don't support resuming and serve a whole file with no Content-Range
 141                 # set in response despite of requested Range (see
 142                 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
 143                 if has_range:
 144                     content_range = ctx.data.headers.get('Content-Range')
 145                     content_range_start, content_range_end, content_len = parse_http_range(content_range)
 146                     if content_range_start is not None and range_start == content_range_start:
 147                         # Content-Range is present and matches requested Range, resume is possible
 148                         accept_content_len = (
 149                             # Non-chunked download
 150                             not ctx.chunk_size
 151                             # Chunked download and requested piece or
 152                             # its part is promised to be served
 153                             or content_range_end == range_end
 154                             or content_len < range_end)
 155                         if accept_content_len:
 156                             ctx.content_len = content_len
 157                             ctx.data_len = min(content_len, req_end or content_len) - (req_start or 0)
 158                             return
 159                     # Content-Range is either not present or invalid. Assuming remote webserver is
 160                     # trying to send the whole file, resume is not possible, so wiping the local file
 161                     # and performing entire redownload
 162                     self.report_unable_to_resume()
 163                     ctx.resume_len = 0
 164                     ctx.open_mode = 'wb'
 165                 ctx.data_len = ctx.content_len = int_or_none(ctx.data.info().get('Content-length', None))
 166             except (compat_urllib_error.HTTPError, ) as err:
 167                 if err.code == 416:
 168                     # Unable to resume (requested range not satisfiable)
 169                     try:
 170                         # Open the connection again without the range header
 171                         ctx.data = self.ydl.urlopen(
 172                             sanitized_Request(url, request_data, headers))
 173                         content_length = ctx.data.info()['Content-Length']
 174                     except (compat_urllib_error.HTTPError, ) as err:
 175                         if err.code < 500 or err.code >= 600:
 176                             raise
 177                     else:
 178                         # Examine the reported length
 179                         if (content_length is not None
 180                                 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
 181                             # The file had already been fully downloaded.
 182                             # Explanation to the above condition: in issue #175 it was revealed that
 183                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 184                             # changing the file size slightly and causing problems for some users. So
 185                             # I decided to implement a suggested change and consider the file
 186                             # completely downloaded if the file size differs less than 100 bytes from
 187                             # the one in the hard drive.
 188                             self.report_file_already_downloaded(ctx.filename)
 189                             self.try_rename(ctx.tmpfilename, ctx.filename)
 190                             self._hook_progress({
 191                                 'filename': ctx.filename,
 192                                 'status': 'finished',
 193                                 'downloaded_bytes': ctx.resume_len,
 194                                 'total_bytes': ctx.resume_len,
 195                             }, info_dict)
 196                             raise SucceedDownload()
 197                         else:
 198                             # The length does not match, we start the download over
 199                             self.report_unable_to_resume()
 200                             ctx.resume_len = 0
 201                             ctx.open_mode = 'wb'
 202                             return
 203                 elif err.code < 500 or err.code >= 600:
 204                     # Unexpected HTTP error
 205                     raise
 206                 raise RetryDownload(err)
 207             except compat_urllib_error.URLError as err:
 208                 if isinstance(err.reason, ssl.CertificateError):
 209                     raise
 210                 raise RetryDownload(err)
 211             # In urllib.request.AbstractHTTPHandler, the response is partially read on request.
 212             # Any errors that occur during this will not be wrapped by URLError
 213             except RESPONSE_READ_EXCEPTIONS as err:
 214                 raise RetryDownload(err)
 215
 216         def download():
 217             nonlocal throttle_start
 218             data_len = ctx.data.info().get('Content-length', None)
 219
 220             # Range HTTP header may be ignored/unsupported by a webserver
 221             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 222             # However, for a test we still would like to download just a piece of a file.
 223             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 224             # block size when downloading a file.
 225             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 226                 data_len = self._TEST_FILE_SIZE
 227
 228             if data_len is not None:
 229                 data_len = int(data_len) + ctx.resume_len
 230                 min_data_len = self.params.get('min_filesize')
 231                 max_data_len = self.params.get('max_filesize')
 232                 if min_data_len is not None and data_len < min_data_len:
 233                     self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
 234                     return False
 235                 if max_data_len is not None and data_len > max_data_len:
 236                     self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
 237                     return False
 238
 239             byte_counter = 0 + ctx.resume_len
 240             block_size = ctx.block_size
 241             start = time.time()
 242
 243             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 244             now = None  # needed for slow_down() in the first loop run
 245             before = start  # start measuring
 246
 247             def retry(e):
 248                 to_stdout = ctx.tmpfilename == '-'
 249                 if ctx.stream is not None:
 250                     if not to_stdout:
 251                         ctx.stream.close()
 252                     ctx.stream = None
 253                 ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
 254                 raise RetryDownload(e)
 255
 256             while True:
 257                 try:
 258                     # Download and write
 259                     data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
 260                 except RESPONSE_READ_EXCEPTIONS as err:
 261                     retry(err)
 262
 263                 byte_counter += len(data_block)
 264
 265                 # exit loop when download is finished
 266                 if len(data_block) == 0:
 267                     break
 268
 269                 # Open destination file just in time
 270                 if ctx.stream is None:
 271                     try:
 272                         ctx.stream, ctx.tmpfilename = self.sanitize_open(
 273                             ctx.tmpfilename, ctx.open_mode)
 274                         assert ctx.stream is not None
 275                         ctx.filename = self.undo_temp_name(ctx.tmpfilename)
 276                         self.report_destination(ctx.filename)
 277                     except (OSError, IOError) as err:
 278                         self.report_error('unable to open for writing: %s' % str(err))
 279                         return False
 280
 281                     if self.params.get('xattr_set_filesize', False) and data_len is not None:
 282                         try:
 283                             write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8'))
 284                         except (XAttrUnavailableError, XAttrMetadataError) as err:
 285                             self.report_error('unable to set filesize xattr: %s' % str(err))
 286
 287                 try:
 288                     ctx.stream.write(data_block)
 289                 except (IOError, OSError) as err:
 290                     self.to_stderr('\n')
 291                     self.report_error('unable to write data: %s' % str(err))
 292                     return False
 293
 294                 # Apply rate limit
 295                 self.slow_down(start, now, byte_counter - ctx.resume_len)
 296
 297                 # end measuring of one loop run
 298                 now = time.time()
 299                 after = now
 300
 301                 # Adjust block size
 302                 if not self.params.get('noresizebuffer', False):
 303                     block_size = self.best_block_size(after - before, len(data_block))
 304
 305                 before = after
 306
 307                 # Progress message
 308                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
 309                 if ctx.data_len is None:
 310                     eta = None
 311                 else:
 312                     eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
 313
 314                 self._hook_progress({
 315                     'status': 'downloading',
 316                     'downloaded_bytes': byte_counter,
 317                     'total_bytes': ctx.data_len,
 318                     'tmpfilename': ctx.tmpfilename,
 319                     'filename': ctx.filename,
 320                     'eta': eta,
 321                     'speed': speed,
 322                     'elapsed': now - ctx.start_time,
 323                     'ctx_id': info_dict.get('ctx_id'),
 324                 }, info_dict)
 325
 326                 if data_len is not None and byte_counter == data_len:
 327                     break
 328
 329                 if speed and speed < (self.params.get('throttledratelimit') or 0):
 330                     # The speed must stay below the limit for 3 seconds
 331                     # This prevents raising error when the speed temporarily goes down
 332                     if throttle_start is None:
 333                         throttle_start = now
 334                     elif now - throttle_start > 3:
 335                         if ctx.stream is not None and ctx.tmpfilename != '-':
 336                             ctx.stream.close()
 337                         raise ThrottledDownload()
 338                 elif speed:
 339                     throttle_start = None
 340
 341             if not is_test and ctx.chunk_size and ctx.content_len is not None and byte_counter < ctx.content_len:
 342                 ctx.resume_len = byte_counter
 343                 # ctx.block_size = block_size
 344                 raise NextFragment()
 345
 346             if ctx.stream is None:
 347                 self.to_stderr('\n')
 348                 self.report_error('Did not get any data blocks')
 349                 return False
 350             if ctx.tmpfilename != '-':
 351                 ctx.stream.close()
 352
 353             if data_len is not None and byte_counter != data_len:
 354                 err = ContentTooShortError(byte_counter, int(data_len))
 355                 if count <= retries:
 356                     retry(err)
 357                 raise err
 358
 359             self.try_rename(ctx.tmpfilename, ctx.filename)
 360
 361             # Update file modification time
 362             if self.params.get('updatetime', True):
 363                 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
 364
 365             self._hook_progress({
 366                 'downloaded_bytes': byte_counter,
 367                 'total_bytes': byte_counter,
 368                 'filename': ctx.filename,
 369                 'status': 'finished',
 370                 'elapsed': time.time() - ctx.start_time,
 371                 'ctx_id': info_dict.get('ctx_id'),
 372             }, info_dict)
 373
 374             return True
 375
 376         while count <= retries:
 377             try:
 378                 establish_connection()
 379                 return download()
 380             except RetryDownload as e:
 381                 count += 1
 382                 if count <= retries:
 383                     self.report_retry(e.source_error, count, retries)
 384                 else:
 385                     self.to_screen(f'[download] Got server HTTP error: {e.source_error}')
 386                 continue
 387             except NextFragment:
 388                 continue
 389             except SucceedDownload:
 390                 return True
 391
 392         self.report_error('giving up after %s retries' % retries)
 393         return False