yt_dlp/downloader/http.py

   1 import http.client
   2 import os
   3 import random
   4 import socket
   5 import ssl
   6 import time
   7 import urllib.error
   8
   9 from .common import FileDownloader
  10 from ..utils import (
  11     ContentTooShortError,
  12     RetryManager,
  13     ThrottledDownload,
  14     XAttrMetadataError,
  15     XAttrUnavailableError,
  16     encodeFilename,
  17     int_or_none,
  18     parse_http_range,
  19     sanitized_Request,
  20     try_call,
  21     write_xattr,
  22 )
  23
  24 RESPONSE_READ_EXCEPTIONS = (
  25     TimeoutError,
  26     socket.timeout,  # compat: py < 3.10
  27     ConnectionError,
  28     ssl.SSLError,
  29     http.client.HTTPException
  30 )
  31
  32
  33 class HttpFD(FileDownloader):
  34     def real_download(self, filename, info_dict):
  35         url = info_dict['url']
  36         request_data = info_dict.get('request_data', None)
  37
  38         class DownloadContext(dict):
  39             __getattr__ = dict.get
  40             __setattr__ = dict.__setitem__
  41             __delattr__ = dict.__delitem__
  42
  43         ctx = DownloadContext()
  44         ctx.filename = filename
  45         ctx.tmpfilename = self.temp_name(filename)
  46         ctx.stream = None
  47
  48         # Disable compression
  49         headers = {'Accept-Encoding': 'identity'}
  50         add_headers = info_dict.get('http_headers')
  51         if add_headers:
  52             headers.update(add_headers)
  53
  54         is_test = self.params.get('test', False)
  55         chunk_size = self._TEST_FILE_SIZE if is_test else (
  56             self.params.get('http_chunk_size')
  57             or info_dict.get('downloader_options', {}).get('http_chunk_size')
  58             or 0)
  59
  60         ctx.open_mode = 'wb'
  61         ctx.resume_len = 0
  62         ctx.block_size = self.params.get('buffersize', 1024)
  63         ctx.start_time = time.time()
  64
  65         # parse given Range
  66         req_start, req_end, _ = parse_http_range(headers.get('Range'))
  67
  68         if self.params.get('continuedl', True):
  69             # Establish possible resume length
  70             if os.path.isfile(encodeFilename(ctx.tmpfilename)):
  71                 ctx.resume_len = os.path.getsize(
  72                     encodeFilename(ctx.tmpfilename))
  73
  74         ctx.is_resume = ctx.resume_len > 0
  75
  76         class SucceedDownload(Exception):
  77             pass
  78
  79         class RetryDownload(Exception):
  80             def __init__(self, source_error):
  81                 self.source_error = source_error
  82
  83         class NextFragment(Exception):
  84             pass
  85
  86         def establish_connection():
  87             ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
  88                               if not is_test and chunk_size else chunk_size)
  89             if ctx.resume_len > 0:
  90                 range_start = ctx.resume_len
  91                 if req_start is not None:
  92                     # offset the beginning of Range to be within request
  93                     range_start += req_start
  94                 if ctx.is_resume:
  95                     self.report_resuming_byte(ctx.resume_len)
  96                 ctx.open_mode = 'ab'
  97             elif req_start is not None:
  98                 range_start = req_start
  99             elif ctx.chunk_size > 0:
 100                 range_start = 0
 101             else:
 102                 range_start = None
 103             ctx.is_resume = False
 104
 105             if ctx.chunk_size:
 106                 chunk_aware_end = range_start + ctx.chunk_size - 1
 107                 # we're not allowed to download outside Range
 108                 range_end = chunk_aware_end if req_end is None else min(chunk_aware_end, req_end)
 109             elif req_end is not None:
 110                 # there's no need for chunked downloads, so download until the end of Range
 111                 range_end = req_end
 112             else:
 113                 range_end = None
 114
 115             if try_call(lambda: range_start > range_end):
 116                 ctx.resume_len = 0
 117                 ctx.open_mode = 'wb'
 118                 raise RetryDownload(Exception(f'Conflicting range. (start={range_start} > end={range_end})'))
 119
 120             if try_call(lambda: range_end >= ctx.content_len):
 121                 range_end = ctx.content_len - 1
 122
 123             request = sanitized_Request(url, request_data, headers)
 124             has_range = range_start is not None
 125             if has_range:
 126                 request.add_header('Range', f'bytes={int(range_start)}-{int_or_none(range_end) or ""}')
 127             # Establish connection
 128             try:
 129                 ctx.data = self.ydl.urlopen(request)
 130                 # When trying to resume, Content-Range HTTP header of response has to be checked
 131                 # to match the value of requested Range HTTP header. This is due to a webservers
 132                 # that don't support resuming and serve a whole file with no Content-Range
 133                 # set in response despite of requested Range (see
 134                 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
 135                 if has_range:
 136                     content_range = ctx.data.headers.get('Content-Range')
 137                     content_range_start, content_range_end, content_len = parse_http_range(content_range)
 138                     # Content-Range is present and matches requested Range, resume is possible
 139                     if range_start == content_range_start and (
 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                         ctx.content_len = content_len
 147                         if content_len or req_end:
 148                             ctx.data_len = min(content_len or req_end, 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                     elif range_start > 0:
 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 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 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 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 close_stream():
 209             if ctx.stream is not None:
 210                 if not ctx.tmpfilename == '-':
 211                     ctx.stream.close()
 212                 ctx.stream = None
 213
 214         def download():
 215             data_len = ctx.data.info().get('Content-length')
 216
 217             if ctx.data.info().get('Content-encoding'):
 218                 # Content-encoding is present, Content-length is not reliable anymore as we are
 219                 # doing auto decompression. (See: https://github.com/yt-dlp/yt-dlp/pull/6176)
 220                 data_len = None
 221
 222             # Range HTTP header may be ignored/unsupported by a webserver
 223             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 224             # However, for a test we still would like to download just a piece of a file.
 225             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 226             # block size when downloading a file.
 227             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 228                 data_len = self._TEST_FILE_SIZE
 229
 230             if data_len is not None:
 231                 data_len = int(data_len) + ctx.resume_len
 232                 min_data_len = self.params.get('min_filesize')
 233                 max_data_len = self.params.get('max_filesize')
 234                 if min_data_len is not None and data_len < min_data_len:
 235                     self.to_screen(
 236                         f'\r[download] File is smaller than min-filesize ({data_len} bytes < {min_data_len} bytes). Aborting.')
 237                     return False
 238                 if max_data_len is not None and data_len > max_data_len:
 239                     self.to_screen(
 240                         f'\r[download] File is larger than max-filesize ({data_len} bytes > {max_data_len} bytes). Aborting.')
 241                     return False
 242
 243             byte_counter = 0 + ctx.resume_len
 244             block_size = ctx.block_size
 245             start = time.time()
 246
 247             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 248             now = None  # needed for slow_down() in the first loop run
 249             before = start  # start measuring
 250
 251             def retry(e):
 252                 close_stream()
 253                 ctx.resume_len = (byte_counter if ctx.tmpfilename == '-'
 254                                   else os.path.getsize(encodeFilename(ctx.tmpfilename)))
 255                 raise RetryDownload(e)
 256
 257             while True:
 258                 try:
 259                     # Download and write
 260                     data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
 261                 except RESPONSE_READ_EXCEPTIONS as err:
 262                     retry(err)
 263
 264                 byte_counter += len(data_block)
 265
 266                 # exit loop when download is finished
 267                 if len(data_block) == 0:
 268                     break
 269
 270                 # Open destination file just in time
 271                 if ctx.stream is None:
 272                     try:
 273                         ctx.stream, ctx.tmpfilename = self.sanitize_open(
 274                             ctx.tmpfilename, ctx.open_mode)
 275                         assert ctx.stream is not None
 276                         ctx.filename = self.undo_temp_name(ctx.tmpfilename)
 277                         self.report_destination(ctx.filename)
 278                     except OSError as err:
 279                         self.report_error('unable to open for writing: %s' % str(err))
 280                         return False
 281
 282                     if self.params.get('xattr_set_filesize', False) and data_len is not None:
 283                         try:
 284                             write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode())
 285                         except (XAttrUnavailableError, XAttrMetadataError) as err:
 286                             self.report_error('unable to set filesize xattr: %s' % str(err))
 287
 288                 try:
 289                     ctx.stream.write(data_block)
 290                 except OSError as err:
 291                     self.to_stderr('\n')
 292                     self.report_error('unable to write data: %s' % str(err))
 293                     return False
 294
 295                 # Apply rate limit
 296                 self.slow_down(start, now, byte_counter - ctx.resume_len)
 297
 298                 # end measuring of one loop run
 299                 now = time.time()
 300                 after = now
 301
 302                 # Adjust block size
 303                 if not self.params.get('noresizebuffer', False):
 304                     block_size = self.best_block_size(after - before, len(data_block))
 305
 306                 before = after
 307
 308                 # Progress message
 309                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
 310                 if ctx.data_len is None:
 311                     eta = None
 312                 else:
 313                     eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
 314
 315                 self._hook_progress({
 316                     'status': 'downloading',
 317                     'downloaded_bytes': byte_counter,
 318                     'total_bytes': ctx.data_len,
 319                     'tmpfilename': ctx.tmpfilename,
 320                     'filename': ctx.filename,
 321                     'eta': eta,
 322                     'speed': speed,
 323                     'elapsed': now - ctx.start_time,
 324                     'ctx_id': info_dict.get('ctx_id'),
 325                 }, info_dict)
 326
 327                 if data_len is not None and byte_counter == data_len:
 328                     break
 329
 330                 if speed and speed < (self.params.get('throttledratelimit') or 0):
 331                     # The speed must stay below the limit for 3 seconds
 332                     # This prevents raising error when the speed temporarily goes down
 333                     if ctx.throttle_start is None:
 334                         ctx.throttle_start = now
 335                     elif now - ctx.throttle_start > 3:
 336                         if ctx.stream is not None and ctx.tmpfilename != '-':
 337                             ctx.stream.close()
 338                         raise ThrottledDownload()
 339                 elif speed:
 340                     ctx.throttle_start = None
 341
 342             if not is_test and ctx.chunk_size and ctx.content_len is not None and byte_counter < ctx.content_len:
 343                 ctx.resume_len = byte_counter
 344                 # ctx.block_size = block_size
 345                 raise NextFragment()
 346
 347             if ctx.stream is None:
 348                 self.to_stderr('\n')
 349                 self.report_error('Did not get any data blocks')
 350                 return False
 351             if ctx.tmpfilename != '-':
 352                 ctx.stream.close()
 353
 354             if data_len is not None and byte_counter != data_len:
 355                 err = ContentTooShortError(byte_counter, int(data_len))
 356                 retry(err)
 357
 358             self.try_rename(ctx.tmpfilename, ctx.filename)
 359
 360             # Update file modification time
 361             if self.params.get('updatetime', True):
 362                 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
 363
 364             self._hook_progress({
 365                 'downloaded_bytes': byte_counter,
 366                 'total_bytes': byte_counter,
 367                 'filename': ctx.filename,
 368                 'status': 'finished',
 369                 'elapsed': time.time() - ctx.start_time,
 370                 'ctx_id': info_dict.get('ctx_id'),
 371             }, info_dict)
 372
 373             return True
 374
 375         for retry in RetryManager(self.params.get('retries'), self.report_retry):
 376             try:
 377                 establish_connection()
 378                 return download()
 379             except RetryDownload as err:
 380                 retry.error = err.source_error
 381                 continue
 382             except NextFragment:
 383                 retry.error = None
 384                 retry.attempt -= 1
 385                 continue
 386             except SucceedDownload:
 387                 return True
 388             except:  # noqa: E722
 389                 close_stream()
 390                 raise
 391         return False