yt_dlp/downloader/http.py

   1 import os
   2 import random
   3 import socket
   4 import ssl
   5 import time
   6 import urllib.error
   7 import http.client
   8
   9 from .common import FileDownloader
  10 from ..utils import (
  11     ContentTooShortError,
  12     ThrottledDownload,
  13     XAttrMetadataError,
  14     XAttrUnavailableError,
  15     encodeFilename,
  16     int_or_none,
  17     parse_http_range,
  18     sanitized_Request,
  19     try_call,
  20     write_xattr,
  21 )
  22
  23 RESPONSE_READ_EXCEPTIONS = (
  24     TimeoutError,
  25     socket.timeout,  # compat: py < 3.10
  26     ConnectionError,
  27     ssl.SSLError,
  28     http.client.HTTPException
  29 )
  30
  31
  32 class HttpFD(FileDownloader):
  33     def real_download(self, filename, info_dict):
  34         url = info_dict['url']
  35         request_data = info_dict.get('request_data', None)
  36
  37         class DownloadContext(dict):
  38             __getattr__ = dict.get
  39             __setattr__ = dict.__setitem__
  40             __delattr__ = dict.__delitem__
  41
  42         ctx = DownloadContext()
  43         ctx.filename = filename
  44         ctx.tmpfilename = self.temp_name(filename)
  45         ctx.stream = None
  46
  47         # Do not include the Accept-Encoding header
  48         headers = {'Youtubedl-no-compression': 'True'}
  49         add_headers = info_dict.get('http_headers')
  50         if add_headers:
  51             headers.update(add_headers)
  52
  53         is_test = self.params.get('test', False)
  54         chunk_size = self._TEST_FILE_SIZE if is_test else (
  55             self.params.get('http_chunk_size')
  56             or info_dict.get('downloader_options', {}).get('http_chunk_size')
  57             or 0)
  58
  59         ctx.open_mode = 'wb'
  60         ctx.resume_len = 0
  61         ctx.block_size = self.params.get('buffersize', 1024)
  62         ctx.start_time = time.time()
  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 establish_connection():
  89             ctx.chunk_size = (random.randint(int(chunk_size * 0.95), chunk_size)
  90                               if not is_test and chunk_size else chunk_size)
  91             if ctx.resume_len > 0:
  92                 range_start = ctx.resume_len
  93                 if req_start is not None:
  94                     # offset the beginning of Range to be within request
  95                     range_start += req_start
  96                 if ctx.is_resume:
  97                     self.report_resuming_byte(ctx.resume_len)
  98                 ctx.open_mode = 'ab'
  99             elif req_start is not None:
 100                 range_start = req_start
 101             elif ctx.chunk_size > 0:
 102                 range_start = 0
 103             else:
 104                 range_start = None
 105             ctx.is_resume = False
 106
 107             if ctx.chunk_size:
 108                 chunk_aware_end = range_start + ctx.chunk_size - 1
 109                 # we're not allowed to download outside Range
 110                 range_end = chunk_aware_end if req_end is None else min(chunk_aware_end, req_end)
 111             elif req_end is not None:
 112                 # there's no need for chunked downloads, so download until the end of Range
 113                 range_end = req_end
 114             else:
 115                 range_end = None
 116
 117             if try_call(lambda: range_start > range_end):
 118                 ctx.resume_len = 0
 119                 ctx.open_mode = 'wb'
 120                 raise RetryDownload(Exception(f'Conflicting range. (start={range_start} > end={range_end})'))
 121
 122             if try_call(lambda: range_end >= ctx.content_len):
 123                 range_end = ctx.content_len - 1
 124
 125             request = sanitized_Request(url, request_data, headers)
 126             has_range = range_start is not None
 127             if has_range:
 128                 request.add_header('Range', f'bytes={int(range_start)}-{int_or_none(range_end) or ""}')
 129             # Establish connection
 130             try:
 131                 ctx.data = self.ydl.urlopen(request)
 132                 # When trying to resume, Content-Range HTTP header of response has to be checked
 133                 # to match the value of requested Range HTTP header. This is due to a webservers
 134                 # that don't support resuming and serve a whole file with no Content-Range
 135                 # set in response despite of requested Range (see
 136                 # https://github.com/ytdl-org/youtube-dl/issues/6057#issuecomment-126129799)
 137                 if has_range:
 138                     content_range = ctx.data.headers.get('Content-Range')
 139                     content_range_start, content_range_end, content_len = parse_http_range(content_range)
 140                     # Content-Range is present and matches requested Range, resume is possible
 141                     if range_start == content_range_start and (
 142                             # Non-chunked download
 143                             not ctx.chunk_size
 144                             # Chunked download and requested piece or
 145                             # its part is promised to be served
 146                             or content_range_end == range_end
 147                             or content_len < range_end):
 148                         ctx.content_len = content_len
 149                         if content_len or req_end:
 150                             ctx.data_len = min(content_len or req_end, req_end or content_len) - (req_start or 0)
 151                         return
 152                     # Content-Range is either not present or invalid. Assuming remote webserver is
 153                     # trying to send the whole file, resume is not possible, so wiping the local file
 154                     # and performing entire redownload
 155                     self.report_unable_to_resume()
 156                     ctx.resume_len = 0
 157                     ctx.open_mode = 'wb'
 158                 ctx.data_len = ctx.content_len = int_or_none(ctx.data.info().get('Content-length', None))
 159             except urllib.error.HTTPError as err:
 160                 if err.code == 416:
 161                     # Unable to resume (requested range not satisfiable)
 162                     try:
 163                         # Open the connection again without the range header
 164                         ctx.data = self.ydl.urlopen(
 165                             sanitized_Request(url, request_data, headers))
 166                         content_length = ctx.data.info()['Content-Length']
 167                     except urllib.error.HTTPError as err:
 168                         if err.code < 500 or err.code >= 600:
 169                             raise
 170                     else:
 171                         # Examine the reported length
 172                         if (content_length is not None
 173                                 and (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
 174                             # The file had already been fully downloaded.
 175                             # Explanation to the above condition: in issue #175 it was revealed that
 176                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 177                             # changing the file size slightly and causing problems for some users. So
 178                             # I decided to implement a suggested change and consider the file
 179                             # completely downloaded if the file size differs less than 100 bytes from
 180                             # the one in the hard drive.
 181                             self.report_file_already_downloaded(ctx.filename)
 182                             self.try_rename(ctx.tmpfilename, ctx.filename)
 183                             self._hook_progress({
 184                                 'filename': ctx.filename,
 185                                 'status': 'finished',
 186                                 'downloaded_bytes': ctx.resume_len,
 187                                 'total_bytes': ctx.resume_len,
 188                             }, info_dict)
 189                             raise SucceedDownload()
 190                         else:
 191                             # The length does not match, we start the download over
 192                             self.report_unable_to_resume()
 193                             ctx.resume_len = 0
 194                             ctx.open_mode = 'wb'
 195                             return
 196                 elif err.code < 500 or err.code >= 600:
 197                     # Unexpected HTTP error
 198                     raise
 199                 raise RetryDownload(err)
 200             except urllib.error.URLError as err:
 201                 if isinstance(err.reason, ssl.CertificateError):
 202                     raise
 203                 raise RetryDownload(err)
 204             # In urllib.request.AbstractHTTPHandler, the response is partially read on request.
 205             # Any errors that occur during this will not be wrapped by URLError
 206             except RESPONSE_READ_EXCEPTIONS as err:
 207                 raise RetryDownload(err)
 208
 209         def download():
 210             data_len = ctx.data.info().get('Content-length', None)
 211
 212             # Range HTTP header may be ignored/unsupported by a webserver
 213             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 214             # However, for a test we still would like to download just a piece of a file.
 215             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 216             # block size when downloading a file.
 217             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 218                 data_len = self._TEST_FILE_SIZE
 219
 220             if data_len is not None:
 221                 data_len = int(data_len) + ctx.resume_len
 222                 min_data_len = self.params.get('min_filesize')
 223                 max_data_len = self.params.get('max_filesize')
 224                 if min_data_len is not None and data_len < min_data_len:
 225                     self.to_screen(
 226                         f'\r[download] File is smaller than min-filesize ({data_len} bytes < {min_data_len} bytes). Aborting.')
 227                     return False
 228                 if max_data_len is not None and data_len > max_data_len:
 229                     self.to_screen(
 230                         f'\r[download] File is larger than max-filesize ({data_len} bytes > {max_data_len} bytes). Aborting.')
 231                     return False
 232
 233             byte_counter = 0 + ctx.resume_len
 234             block_size = ctx.block_size
 235             start = time.time()
 236
 237             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 238             now = None  # needed for slow_down() in the first loop run
 239             before = start  # start measuring
 240
 241             def retry(e):
 242                 to_stdout = ctx.tmpfilename == '-'
 243                 if ctx.stream is not None:
 244                     if not to_stdout:
 245                         ctx.stream.close()
 246                     ctx.stream = None
 247                 ctx.resume_len = byte_counter if to_stdout else os.path.getsize(encodeFilename(ctx.tmpfilename))
 248                 raise RetryDownload(e)
 249
 250             while True:
 251                 try:
 252                     # Download and write
 253                     data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
 254                 except RESPONSE_READ_EXCEPTIONS as err:
 255                     retry(err)
 256
 257                 byte_counter += len(data_block)
 258
 259                 # exit loop when download is finished
 260                 if len(data_block) == 0:
 261                     break
 262
 263                 # Open destination file just in time
 264                 if ctx.stream is None:
 265                     try:
 266                         ctx.stream, ctx.tmpfilename = self.sanitize_open(
 267                             ctx.tmpfilename, ctx.open_mode)
 268                         assert ctx.stream is not None
 269                         ctx.filename = self.undo_temp_name(ctx.tmpfilename)
 270                         self.report_destination(ctx.filename)
 271                     except OSError as err:
 272                         self.report_error('unable to open for writing: %s' % str(err))
 273                         return False
 274
 275                     if self.params.get('xattr_set_filesize', False) and data_len is not None:
 276                         try:
 277                             write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode())
 278                         except (XAttrUnavailableError, XAttrMetadataError) as err:
 279                             self.report_error('unable to set filesize xattr: %s' % str(err))
 280
 281                 try:
 282                     ctx.stream.write(data_block)
 283                 except OSError as err:
 284                     self.to_stderr('\n')
 285                     self.report_error('unable to write data: %s' % str(err))
 286                     return False
 287
 288                 # Apply rate limit
 289                 self.slow_down(start, now, byte_counter - ctx.resume_len)
 290
 291                 # end measuring of one loop run
 292                 now = time.time()
 293                 after = now
 294
 295                 # Adjust block size
 296                 if not self.params.get('noresizebuffer', False):
 297                     block_size = self.best_block_size(after - before, len(data_block))
 298
 299                 before = after
 300
 301                 # Progress message
 302                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
 303                 if ctx.data_len is None:
 304                     eta = None
 305                 else:
 306                     eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
 307
 308                 self._hook_progress({
 309                     'status': 'downloading',
 310                     'downloaded_bytes': byte_counter,
 311                     'total_bytes': ctx.data_len,
 312                     'tmpfilename': ctx.tmpfilename,
 313                     'filename': ctx.filename,
 314                     'eta': eta,
 315                     'speed': speed,
 316                     'elapsed': now - ctx.start_time,
 317                     'ctx_id': info_dict.get('ctx_id'),
 318                 }, info_dict)
 319
 320                 if data_len is not None and byte_counter == data_len:
 321                     break
 322
 323                 if speed and speed < (self.params.get('throttledratelimit') or 0):
 324                     # The speed must stay below the limit for 3 seconds
 325                     # This prevents raising error when the speed temporarily goes down
 326                     if ctx.throttle_start is None:
 327                         ctx.throttle_start = now
 328                     elif now - ctx.throttle_start > 3:
 329                         if ctx.stream is not None and ctx.tmpfilename != '-':
 330                             ctx.stream.close()
 331                         raise ThrottledDownload()
 332                 elif speed:
 333                     ctx.throttle_start = None
 334
 335             if not is_test and ctx.chunk_size and ctx.content_len is not None and byte_counter < ctx.content_len:
 336                 ctx.resume_len = byte_counter
 337                 # ctx.block_size = block_size
 338                 raise NextFragment()
 339
 340             if ctx.stream is None:
 341                 self.to_stderr('\n')
 342                 self.report_error('Did not get any data blocks')
 343                 return False
 344             if ctx.tmpfilename != '-':
 345                 ctx.stream.close()
 346
 347             if data_len is not None and byte_counter != data_len:
 348                 err = ContentTooShortError(byte_counter, int(data_len))
 349                 if count <= retries:
 350                     retry(err)
 351                 raise err
 352
 353             self.try_rename(ctx.tmpfilename, ctx.filename)
 354
 355             # Update file modification time
 356             if self.params.get('updatetime', True):
 357                 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
 358
 359             self._hook_progress({
 360                 'downloaded_bytes': byte_counter,
 361                 'total_bytes': byte_counter,
 362                 'filename': ctx.filename,
 363                 'status': 'finished',
 364                 'elapsed': time.time() - ctx.start_time,
 365                 'ctx_id': info_dict.get('ctx_id'),
 366             }, info_dict)
 367
 368             return True
 369
 370         while count <= retries:
 371             try:
 372                 establish_connection()
 373                 return download()
 374             except RetryDownload as e:
 375                 count += 1
 376                 if count <= retries:
 377                     self.report_retry(e.source_error, count, retries)
 378                 else:
 379                     self.to_screen(f'[download] Got server HTTP error: {e.source_error}')
 380                 continue
 381             except NextFragment:
 382                 continue
 383             except SucceedDownload:
 384                 return True
 385
 386         self.report_error('giving up after %s retries' % retries)
 387         return False