Raphaël G. Git Repositories - youtubedl/blob - youtube_dl/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     write_xattr,
  22     XAttrMetadataError,
  23     XAttrUnavailableError,
  24 )
  25
  26
  27 class HttpFD(FileDownloader):
  28     def real_download(self, filename, info_dict):
  29         url = info_dict['url']
  30
  31         class DownloadContext(dict):
  32             __getattr__ = dict.get
  33             __setattr__ = dict.__setitem__
  34             __delattr__ = dict.__delitem__
  35
  36         ctx = DownloadContext()
  37         ctx.filename = filename
  38         ctx.tmpfilename = self.temp_name(filename)
  39         ctx.stream = None
  40
  41         # Do not include the Accept-Encoding header
  42         headers = {'Youtubedl-no-compression': 'True'}
  43         add_headers = info_dict.get('http_headers')
  44         if add_headers:
  45             headers.update(add_headers)
  46
  47         is_test = self.params.get('test', False)
  48         chunk_size = self._TEST_FILE_SIZE if is_test else (
  49             info_dict.get('downloader_options', {}).get('http_chunk_size') or
  50             self.params.get('http_chunk_size') or 0)
  51
  52         ctx.open_mode = 'wb'
  53         ctx.resume_len = 0
  54         ctx.data_len = None
  55         ctx.block_size = self.params.get('buffersize', 1024)
  56         ctx.start_time = time.time()
  57         ctx.chunk_size = None
  58
  59         if self.params.get('continuedl', True):
  60             # Establish possible resume length
  61             if os.path.isfile(encodeFilename(ctx.tmpfilename)):
  62                 ctx.resume_len = os.path.getsize(
  63                     encodeFilename(ctx.tmpfilename))
  64
  65         ctx.is_resume = ctx.resume_len > 0
  66
  67         count = 0
  68         retries = self.params.get('retries', 0)
  69
  70         class SucceedDownload(Exception):
  71             pass
  72
  73         class RetryDownload(Exception):
  74             def __init__(self, source_error):
  75                 self.source_error = source_error
  76
  77         class NextFragment(Exception):
  78             pass
  79
  80         def set_range(req, start, end):
  81             range_header = 'bytes=%d-' % start
  82             if end:
  83                 range_header += compat_str(end)
  84             req.add_header('Range', range_header)
  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 ctx.is_resume:
  92                     self.report_resuming_byte(ctx.resume_len)
  93                 ctx.open_mode = 'ab'
  94             elif ctx.chunk_size > 0:
  95                 range_start = 0
  96             else:
  97                 range_start = None
  98             ctx.is_resume = False
  99             range_end = range_start + ctx.chunk_size - 1 if ctx.chunk_size else None
 100             if range_end and ctx.data_len is not None and range_end >= ctx.data_len:
 101                 range_end = ctx.data_len - 1
 102             has_range = range_start is not None
 103             ctx.has_range = has_range
 104             request = sanitized_Request(url, None, headers)
 105             if has_range:
 106                 set_range(request, range_start, range_end)
 107             # Establish connection
 108             try:
 109                 ctx.data = self.ydl.urlopen(request)
 110                 # When trying to resume, Content-Range HTTP header of response has to be checked
 111                 # to match the value of requested Range HTTP header. This is due to a webservers
 112                 # that don't support resuming and serve a whole file with no Content-Range
 113                 # set in response despite of requested Range (see
 114                 # https://github.com/rg3/youtube-dl/issues/6057#issuecomment-126129799)
 115                 if has_range:
 116                     content_range = ctx.data.headers.get('Content-Range')
 117                     if content_range:
 118                         content_range_m = re.search(r'bytes (\d+)-(\d+)?(?:/(\d+))?', content_range)
 119                         # Content-Range is present and matches requested Range, resume is possible
 120                         if content_range_m:
 121                             if range_start == int(content_range_m.group(1)):
 122                                 content_range_end = int_or_none(content_range_m.group(2))
 123                                 content_len = int_or_none(content_range_m.group(3))
 124                                 accept_content_len = (
 125                                     # Non-chunked download
 126                                     not ctx.chunk_size or
 127                                     # Chunked download and requested piece or
 128                                     # its part is promised to be served
 129                                     content_range_end == range_end or
 130                                     content_len < range_end)
 131                                 if accept_content_len:
 132                                     ctx.data_len = content_len
 133                                     return
 134                     # Content-Range is either not present or invalid. Assuming remote webserver is
 135                     # trying to send the whole file, resume is not possible, so wiping the local file
 136                     # and performing entire redownload
 137                     self.report_unable_to_resume()
 138                     ctx.resume_len = 0
 139                     ctx.open_mode = 'wb'
 140                 ctx.data_len = int_or_none(ctx.data.info().get('Content-length', None))
 141                 return
 142             except (compat_urllib_error.HTTPError, ) as err:
 143                 if err.code == 416:
 144                     # Unable to resume (requested range not satisfiable)
 145                     try:
 146                         # Open the connection again without the range header
 147                         ctx.data = self.ydl.urlopen(
 148                             sanitized_Request(url, None, headers))
 149                         content_length = ctx.data.info()['Content-Length']
 150                     except (compat_urllib_error.HTTPError, ) as err:
 151                         if err.code < 500 or err.code >= 600:
 152                             raise
 153                     else:
 154                         # Examine the reported length
 155                         if (content_length is not None and
 156                                 (ctx.resume_len - 100 < int(content_length) < ctx.resume_len + 100)):
 157                             # The file had already been fully downloaded.
 158                             # Explanation to the above condition: in issue #175 it was revealed that
 159                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 160                             # changing the file size slightly and causing problems for some users. So
 161                             # I decided to implement a suggested change and consider the file
 162                             # completely downloaded if the file size differs less than 100 bytes from
 163                             # the one in the hard drive.
 164                             self.report_file_already_downloaded(ctx.filename)
 165                             self.try_rename(ctx.tmpfilename, ctx.filename)
 166                             self._hook_progress({
 167                                 'filename': ctx.filename,
 168                                 'status': 'finished',
 169                                 'downloaded_bytes': ctx.resume_len,
 170                                 'total_bytes': ctx.resume_len,
 171                             })
 172                             raise SucceedDownload()
 173                         else:
 174                             # The length does not match, we start the download over
 175                             self.report_unable_to_resume()
 176                             ctx.resume_len = 0
 177                             ctx.open_mode = 'wb'
 178                             return
 179                 elif err.code < 500 or err.code >= 600:
 180                     # Unexpected HTTP error
 181                     raise
 182                 raise RetryDownload(err)
 183             except socket.error as err:
 184                 if err.errno != errno.ECONNRESET:
 185                     # Connection reset is no problem, just retry
 186                     raise
 187                 raise RetryDownload(err)
 188
 189         def download():
 190             data_len = ctx.data.info().get('Content-length', None)
 191
 192             # Range HTTP header may be ignored/unsupported by a webserver
 193             # (e.g. extractor/scivee.py, extractor/bambuser.py).
 194             # However, for a test we still would like to download just a piece of a file.
 195             # To achieve this we limit data_len to _TEST_FILE_SIZE and manually control
 196             # block size when downloading a file.
 197             if is_test and (data_len is None or int(data_len) > self._TEST_FILE_SIZE):
 198                 data_len = self._TEST_FILE_SIZE
 199
 200             if data_len is not None:
 201                 data_len = int(data_len) + ctx.resume_len
 202                 min_data_len = self.params.get('min_filesize')
 203                 max_data_len = self.params.get('max_filesize')
 204                 if min_data_len is not None and data_len < min_data_len:
 205                     self.to_screen('\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
 206                     return False
 207                 if max_data_len is not None and data_len > max_data_len:
 208                     self.to_screen('\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
 209                     return False
 210
 211             byte_counter = 0 + ctx.resume_len
 212             block_size = ctx.block_size
 213             start = time.time()
 214
 215             # measure time over whole while-loop, so slow_down() and best_block_size() work together properly
 216             now = None  # needed for slow_down() in the first loop run
 217             before = start  # start measuring
 218
 219             def retry(e):
 220                 if ctx.tmpfilename != '-':
 221                     ctx.stream.close()
 222                 ctx.stream = None
 223                 ctx.resume_len = os.path.getsize(encodeFilename(ctx.tmpfilename))
 224                 raise RetryDownload(e)
 225
 226             while True:
 227                 try:
 228                     # Download and write
 229                     data_block = ctx.data.read(block_size if not is_test else min(block_size, data_len - byte_counter))
 230                 # socket.timeout is a subclass of socket.error but may not have
 231                 # errno set
 232                 except socket.timeout as e:
 233                     retry(e)
 234                 except socket.error as e:
 235                     if e.errno not in (errno.ECONNRESET, errno.ETIMEDOUT):
 236                         raise
 237                     retry(e)
 238
 239                 byte_counter += len(data_block)
 240
 241                 # exit loop when download is finished
 242                 if len(data_block) == 0:
 243                     break
 244
 245                 # Open destination file just in time
 246                 if ctx.stream is None:
 247                     try:
 248                         ctx.stream, ctx.tmpfilename = sanitize_open(
 249                             ctx.tmpfilename, ctx.open_mode)
 250                         assert ctx.stream is not None
 251                         ctx.filename = self.undo_temp_name(ctx.tmpfilename)
 252                         self.report_destination(ctx.filename)
 253                     except (OSError, IOError) as err:
 254                         self.report_error('unable to open for writing: %s' % str(err))
 255                         return False
 256
 257                     if self.params.get('xattr_set_filesize', False) and data_len is not None:
 258                         try:
 259                             write_xattr(ctx.tmpfilename, 'user.ytdl.filesize', str(data_len).encode('utf-8'))
 260                         except (XAttrUnavailableError, XAttrMetadataError) as err:
 261                             self.report_error('unable to set filesize xattr: %s' % str(err))
 262
 263                 try:
 264                     ctx.stream.write(data_block)
 265                 except (IOError, OSError) as err:
 266                     self.to_stderr('\n')
 267                     self.report_error('unable to write data: %s' % str(err))
 268                     return False
 269
 270                 # Apply rate limit
 271                 self.slow_down(start, now, byte_counter - ctx.resume_len)
 272
 273                 # end measuring of one loop run
 274                 now = time.time()
 275                 after = now
 276
 277                 # Adjust block size
 278                 if not self.params.get('noresizebuffer', False):
 279                     block_size = self.best_block_size(after - before, len(data_block))
 280
 281                 before = after
 282
 283                 # Progress message
 284                 speed = self.calc_speed(start, now, byte_counter - ctx.resume_len)
 285                 if ctx.data_len is None:
 286                     eta = None
 287                 else:
 288                     eta = self.calc_eta(start, time.time(), ctx.data_len - ctx.resume_len, byte_counter - ctx.resume_len)
 289
 290                 self._hook_progress({
 291                     'status': 'downloading',
 292                     'downloaded_bytes': byte_counter,
 293                     'total_bytes': ctx.data_len,
 294                     'tmpfilename': ctx.tmpfilename,
 295                     'filename': ctx.filename,
 296                     'eta': eta,
 297                     'speed': speed,
 298                     'elapsed': now - ctx.start_time,
 299                 })
 300
 301                 if is_test and byte_counter == data_len:
 302                     break
 303
 304             if not is_test and ctx.chunk_size and ctx.data_len is not None and byte_counter < ctx.data_len:
 305                 ctx.resume_len = byte_counter
 306                 # ctx.block_size = block_size
 307                 raise NextFragment()
 308
 309             if ctx.stream is None:
 310                 self.to_stderr('\n')
 311                 self.report_error('Did not get any data blocks')
 312                 return False
 313             if ctx.tmpfilename != '-':
 314                 ctx.stream.close()
 315
 316             if data_len is not None and byte_counter != data_len:
 317                 err = ContentTooShortError(byte_counter, int(data_len))
 318                 if count <= retries:
 319                     retry(err)
 320                 raise err
 321
 322             self.try_rename(ctx.tmpfilename, ctx.filename)
 323
 324             # Update file modification time
 325             if self.params.get('updatetime', True):
 326                 info_dict['filetime'] = self.try_utime(ctx.filename, ctx.data.info().get('last-modified', None))
 327
 328             self._hook_progress({
 329                 'downloaded_bytes': byte_counter,
 330                 'total_bytes': byte_counter,
 331                 'filename': ctx.filename,
 332                 'status': 'finished',
 333                 'elapsed': time.time() - ctx.start_time,
 334             })
 335
 336             return True
 337
 338         while count <= retries:
 339             try:
 340                 establish_connection()
 341                 return download()
 342             except RetryDownload as e:
 343                 count += 1
 344                 if count <= retries:
 345                     self.report_retry(e.source_error, count, retries)
 346                 continue
 347             except NextFragment:
 348                 continue
 349             except SucceedDownload:
 350                 return True
 351
 352         self.report_error('giving up after %s retries' % retries)
 353         return False