youtube_dl/FileDownloader.py

   1 import math
   2 import os
   3 import re
   4 import subprocess
   5 import sys
   6 import time
   7
   8 if os.name == 'nt':
   9     import ctypes
  10
  11 from .utils import (
  12     compat_urllib_error,
  13     compat_urllib_request,
  14     ContentTooShortError,
  15     determine_ext,
  16     encodeFilename,
  17     sanitize_open,
  18     timeconvert,
  19 )
  20
  21
  22 class FileDownloader(object):
  23     """File Downloader class.
  24
  25     File downloader objects are the ones responsible of downloading the
  26     actual video file and writing it to disk.
  27
  28     File downloaders accept a lot of parameters. In order not to saturate
  29     the object constructor with arguments, it receives a dictionary of
  30     options instead.
  31
  32     Available options:
  33
  34     verbose:           Print additional info to stdout.
  35     quiet:             Do not print messages to stdout.
  36     ratelimit:         Download speed limit, in bytes/sec.
  37     retries:           Number of times to retry for HTTP error 5xx
  38     buffersize:        Size of download buffer in bytes.
  39     noresizebuffer:    Do not automatically resize the download buffer.
  40     continuedl:        Try to continue downloads if possible.
  41     noprogress:        Do not print the progress bar.
  42     logtostderr:       Log messages to stderr instead of stdout.
  43     consoletitle:      Display progress in console window's titlebar.
  44     nopart:            Do not use temporary .part files.
  45     updatetime:        Use the Last-modified header to set output file timestamps.
  46     test:              Download only first bytes to test the downloader.
  47     min_filesize:      Skip files smaller than this size
  48     max_filesize:      Skip files larger than this size
  49     """
  50
  51     params = None
  52
  53     def __init__(self, ydl, params):
  54         """Create a FileDownloader object with the given options."""
  55         self.ydl = ydl
  56         self._progress_hooks = []
  57         self.params = params
  58
  59     @staticmethod
  60     def format_bytes(bytes):
  61         if bytes is None:
  62             return 'N/A'
  63         if type(bytes) is str:
  64             bytes = float(bytes)
  65         if bytes == 0.0:
  66             exponent = 0
  67         else:
  68             exponent = int(math.log(bytes, 1024.0))
  69         suffix = ['B','KiB','MiB','GiB','TiB','PiB','EiB','ZiB','YiB'][exponent]
  70         converted = float(bytes) / float(1024 ** exponent)
  71         return '%.2f%s' % (converted, suffix)
  72
  73     @staticmethod
  74     def format_seconds(seconds):
  75         (mins, secs) = divmod(seconds, 60)
  76         (hours, mins) = divmod(mins, 60)
  77         if hours > 99:
  78             return '--:--:--'
  79         if hours == 0:
  80             return '%02d:%02d' % (mins, secs)
  81         else:
  82             return '%02d:%02d:%02d' % (hours, mins, secs)
  83
  84     @staticmethod
  85     def calc_percent(byte_counter, data_len):
  86         if data_len is None:
  87             return None
  88         return float(byte_counter) / float(data_len) * 100.0
  89
  90     @staticmethod
  91     def format_percent(percent):
  92         if percent is None:
  93             return '---.-%'
  94         return '%6s' % ('%3.1f%%' % percent)
  95
  96     @staticmethod
  97     def calc_eta(start, now, total, current):
  98         if total is None:
  99             return None
 100         dif = now - start
 101         if current == 0 or dif < 0.001: # One millisecond
 102             return None
 103         rate = float(current) / dif
 104         return int((float(total) - float(current)) / rate)
 105
 106     @staticmethod
 107     def format_eta(eta):
 108         if eta is None:
 109             return '--:--'
 110         return FileDownloader.format_seconds(eta)
 111
 112     @staticmethod
 113     def calc_speed(start, now, bytes):
 114         dif = now - start
 115         if bytes == 0 or dif < 0.001: # One millisecond
 116             return None
 117         return float(bytes) / dif
 118
 119     @staticmethod
 120     def format_speed(speed):
 121         if speed is None:
 122             return '%10s' % '---b/s'
 123         return '%10s' % ('%s/s' % FileDownloader.format_bytes(speed))
 124
 125     @staticmethod
 126     def best_block_size(elapsed_time, bytes):
 127         new_min = max(bytes / 2.0, 1.0)
 128         new_max = min(max(bytes * 2.0, 1.0), 4194304) # Do not surpass 4 MB
 129         if elapsed_time < 0.001:
 130             return int(new_max)
 131         rate = bytes / elapsed_time
 132         if rate > new_max:
 133             return int(new_max)
 134         if rate < new_min:
 135             return int(new_min)
 136         return int(rate)
 137
 138     @staticmethod
 139     def parse_bytes(bytestr):
 140         """Parse a string indicating a byte quantity into an integer."""
 141         matchobj = re.match(r'(?i)^(\d+(?:\.\d+)?)([kMGTPEZY]?)$', bytestr)
 142         if matchobj is None:
 143             return None
 144         number = float(matchobj.group(1))
 145         multiplier = 1024.0 ** 'bkmgtpezy'.index(matchobj.group(2).lower())
 146         return int(round(number * multiplier))
 147
 148     def to_screen(self, *args, **kargs):
 149         self.ydl.to_screen(*args, **kargs)
 150
 151     def to_stderr(self, message):
 152         self.ydl.to_screen(message)
 153
 154     def to_cons_title(self, message):
 155         """Set console/terminal window title to message."""
 156         if not self.params.get('consoletitle', False):
 157             return
 158         if os.name == 'nt' and ctypes.windll.kernel32.GetConsoleWindow():
 159             # c_wchar_p() might not be necessary if `message` is
 160             # already of type unicode()
 161             ctypes.windll.kernel32.SetConsoleTitleW(ctypes.c_wchar_p(message))
 162         elif 'TERM' in os.environ:
 163             self.to_screen('\033]0;%s\007' % message, skip_eol=True)
 164
 165     def trouble(self, *args, **kargs):
 166         self.ydl.trouble(*args, **kargs)
 167
 168     def report_warning(self, *args, **kargs):
 169         self.ydl.report_warning(*args, **kargs)
 170
 171     def report_error(self, *args, **kargs):
 172         self.ydl.report_error(*args, **kargs)
 173
 174     def slow_down(self, start_time, byte_counter):
 175         """Sleep if the download speed is over the rate limit."""
 176         rate_limit = self.params.get('ratelimit', None)
 177         if rate_limit is None or byte_counter == 0:
 178             return
 179         now = time.time()
 180         elapsed = now - start_time
 181         if elapsed <= 0.0:
 182             return
 183         speed = float(byte_counter) / elapsed
 184         if speed > rate_limit:
 185             time.sleep((byte_counter - rate_limit * (now - start_time)) / rate_limit)
 186
 187     def temp_name(self, filename):
 188         """Returns a temporary filename for the given filename."""
 189         if self.params.get('nopart', False) or filename == u'-' or \
 190                 (os.path.exists(encodeFilename(filename)) and not os.path.isfile(encodeFilename(filename))):
 191             return filename
 192         return filename + u'.part'
 193
 194     def undo_temp_name(self, filename):
 195         if filename.endswith(u'.part'):
 196             return filename[:-len(u'.part')]
 197         return filename
 198
 199     def try_rename(self, old_filename, new_filename):
 200         try:
 201             if old_filename == new_filename:
 202                 return
 203             os.rename(encodeFilename(old_filename), encodeFilename(new_filename))
 204         except (IOError, OSError):
 205             self.report_error(u'unable to rename file')
 206
 207     def try_utime(self, filename, last_modified_hdr):
 208         """Try to set the last-modified time of the given file."""
 209         if last_modified_hdr is None:
 210             return
 211         if not os.path.isfile(encodeFilename(filename)):
 212             return
 213         timestr = last_modified_hdr
 214         if timestr is None:
 215             return
 216         filetime = timeconvert(timestr)
 217         if filetime is None:
 218             return filetime
 219         # Ignore obviously invalid dates
 220         if filetime == 0:
 221             return
 222         try:
 223             os.utime(filename, (time.time(), filetime))
 224         except:
 225             pass
 226         return filetime
 227
 228     def report_destination(self, filename):
 229         """Report destination filename."""
 230         self.to_screen(u'[download] Destination: ' + filename)
 231
 232     def report_progress(self, percent, data_len_str, speed, eta):
 233         """Report download progress."""
 234         if self.params.get('noprogress', False):
 235             return
 236         clear_line = (u'\x1b[K' if sys.stderr.isatty() and os.name != 'nt' else u'')
 237         eta_str = self.format_eta(eta)
 238         percent_str = self.format_percent(percent)
 239         speed_str = self.format_speed(speed)
 240         if self.params.get('progress_with_newline', False):
 241             self.to_screen(u'[download] %s of %s at %s ETA %s' %
 242                 (percent_str, data_len_str, speed_str, eta_str))
 243         else:
 244             self.to_screen(u'\r%s[download] %s of %s at %s ETA %s' %
 245                 (clear_line, percent_str, data_len_str, speed_str, eta_str), skip_eol=True)
 246         self.to_cons_title(u'youtube-dl - %s of %s at %s ETA %s' %
 247                 (percent_str.strip(), data_len_str.strip(), speed_str.strip(), eta_str.strip()))
 248
 249     def report_resuming_byte(self, resume_len):
 250         """Report attempt to resume at given byte."""
 251         self.to_screen(u'[download] Resuming download at byte %s' % resume_len)
 252
 253     def report_retry(self, count, retries):
 254         """Report retry in case of HTTP error 5xx"""
 255         self.to_screen(u'[download] Got server HTTP error. Retrying (attempt %d of %d)...' % (count, retries))
 256
 257     def report_file_already_downloaded(self, file_name):
 258         """Report file has already been fully downloaded."""
 259         try:
 260             self.to_screen(u'[download] %s has already been downloaded' % file_name)
 261         except UnicodeEncodeError:
 262             self.to_screen(u'[download] The file has already been downloaded')
 263
 264     def report_unable_to_resume(self):
 265         """Report it was impossible to resume download."""
 266         self.to_screen(u'[download] Unable to resume')
 267
 268     def report_finish(self, data_len_str, tot_time):
 269         """Report download finished."""
 270         if self.params.get('noprogress', False):
 271             self.to_screen(u'[download] Download completed')
 272         else:
 273             clear_line = (u'\x1b[K' if sys.stderr.isatty() and os.name != 'nt' else u'')
 274             self.to_screen(u'\r%s[download] 100%% of %s in %s' %
 275                 (clear_line, data_len_str, self.format_seconds(tot_time)))
 276
 277     def _download_with_rtmpdump(self, filename, url, player_url, page_url, play_path, tc_url):
 278         self.report_destination(filename)
 279         tmpfilename = self.temp_name(filename)
 280         test = self.params.get('test', False)
 281
 282         # Check for rtmpdump first
 283         try:
 284             subprocess.call(['rtmpdump', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
 285         except (OSError, IOError):
 286             self.report_error(u'RTMP download detected but "rtmpdump" could not be run')
 287             return False
 288         verbosity_option = '--verbose' if self.params.get('verbose', False) else '--quiet'
 289
 290         # Download using rtmpdump. rtmpdump returns exit code 2 when
 291         # the connection was interrumpted and resuming appears to be
 292         # possible. This is part of rtmpdump's normal usage, AFAIK.
 293         basic_args = ['rtmpdump', verbosity_option, '-r', url, '-o', tmpfilename]
 294         if player_url is not None:
 295             basic_args += ['--swfVfy', player_url]
 296         if page_url is not None:
 297             basic_args += ['--pageUrl', page_url]
 298         if play_path is not None:
 299             basic_args += ['--playpath', play_path]
 300         if tc_url is not None:
 301             basic_args += ['--tcUrl', url]
 302         if test:
 303             basic_args += ['--stop', '1']
 304         args = basic_args + [[], ['--resume', '--skip', '1']][self.params.get('continuedl', False)]
 305         if self.params.get('verbose', False):
 306             try:
 307                 import pipes
 308                 shell_quote = lambda args: ' '.join(map(pipes.quote, args))
 309             except ImportError:
 310                 shell_quote = repr
 311             self.to_screen(u'[debug] rtmpdump command line: ' + shell_quote(args))
 312         retval = subprocess.call(args)
 313         while (retval == 2 or retval == 1) and not test:
 314             prevsize = os.path.getsize(encodeFilename(tmpfilename))
 315             self.to_screen(u'\r[rtmpdump] %s bytes' % prevsize, skip_eol=True)
 316             time.sleep(5.0) # This seems to be needed
 317             retval = subprocess.call(basic_args + ['-e'] + [[], ['-k', '1']][retval == 1])
 318             cursize = os.path.getsize(encodeFilename(tmpfilename))
 319             if prevsize == cursize and retval == 1:
 320                 break
 321              # Some rtmp streams seem abort after ~ 99.8%. Don't complain for those
 322             if prevsize == cursize and retval == 2 and cursize > 1024:
 323                 self.to_screen(u'\r[rtmpdump] Could not download the whole video. This can happen for some advertisements.')
 324                 retval = 0
 325                 break
 326         if retval == 0 or (test and retval == 2):
 327             fsize = os.path.getsize(encodeFilename(tmpfilename))
 328             self.to_screen(u'\r[rtmpdump] %s bytes' % fsize)
 329             self.try_rename(tmpfilename, filename)
 330             self._hook_progress({
 331                 'downloaded_bytes': fsize,
 332                 'total_bytes': fsize,
 333                 'filename': filename,
 334                 'status': 'finished',
 335             })
 336             return True
 337         else:
 338             self.to_stderr(u"\n")
 339             self.report_error(u'rtmpdump exited with code %d' % retval)
 340             return False
 341
 342     def _download_with_mplayer(self, filename, url):
 343         self.report_destination(filename)
 344         tmpfilename = self.temp_name(filename)
 345
 346         args = ['mplayer', '-really-quiet', '-vo', 'null', '-vc', 'dummy', '-dumpstream', '-dumpfile', tmpfilename, url]
 347         # Check for mplayer first
 348         try:
 349             subprocess.call(['mplayer', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
 350         except (OSError, IOError):
 351             self.report_error(u'MMS or RTSP download detected but "%s" could not be run' % args[0] )
 352             return False
 353
 354         # Download using mplayer.
 355         retval = subprocess.call(args)
 356         if retval == 0:
 357             fsize = os.path.getsize(encodeFilename(tmpfilename))
 358             self.to_screen(u'\r[%s] %s bytes' % (args[0], fsize))
 359             self.try_rename(tmpfilename, filename)
 360             self._hook_progress({
 361                 'downloaded_bytes': fsize,
 362                 'total_bytes': fsize,
 363                 'filename': filename,
 364                 'status': 'finished',
 365             })
 366             return True
 367         else:
 368             self.to_stderr(u"\n")
 369             self.report_error(u'mplayer exited with code %d' % retval)
 370             return False
 371
 372     def _download_m3u8_with_ffmpeg(self, filename, url):
 373         self.report_destination(filename)
 374         tmpfilename = self.temp_name(filename)
 375
 376         args = ['ffmpeg', '-y', '-i', url, '-f', 'mp4', tmpfilename]
 377         # Check for ffmpeg first
 378         try:
 379             subprocess.call(['ffmpeg', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
 380         except (OSError, IOError):
 381             self.report_error(u'm3u8 download detected but "%s" could not be run' % args[0] )
 382             return False
 383
 384         retval = subprocess.call(args)
 385         if retval == 0:
 386             fsize = os.path.getsize(encodeFilename(tmpfilename))
 387             self.to_screen(u'\r[%s] %s bytes' % (args[0], fsize))
 388             self.try_rename(tmpfilename, filename)
 389             self._hook_progress({
 390                 'downloaded_bytes': fsize,
 391                 'total_bytes': fsize,
 392                 'filename': filename,
 393                 'status': 'finished',
 394             })
 395             return True
 396         else:
 397             self.to_stderr(u"\n")
 398             self.report_error(u'ffmpeg exited with code %d' % retval)
 399             return False
 400
 401
 402     def _do_download(self, filename, info_dict):
 403         url = info_dict['url']
 404
 405         # Check file already present
 406         if self.params.get('continuedl', False) and os.path.isfile(encodeFilename(filename)) and not self.params.get('nopart', False):
 407             self.report_file_already_downloaded(filename)
 408             self._hook_progress({
 409                 'filename': filename,
 410                 'status': 'finished',
 411                 'total_bytes': os.path.getsize(encodeFilename(filename)),
 412             })
 413             return True
 414
 415         # Attempt to download using rtmpdump
 416         if url.startswith('rtmp'):
 417             return self._download_with_rtmpdump(filename, url,
 418                                                 info_dict.get('player_url', None),
 419                                                 info_dict.get('page_url', None),
 420                                                 info_dict.get('play_path', None),
 421                                                 info_dict.get('tc_url', None))
 422
 423         # Attempt to download using mplayer
 424         if url.startswith('mms') or url.startswith('rtsp'):
 425             return self._download_with_mplayer(filename, url)
 426
 427         # m3u8 manifest are downloaded with ffmpeg
 428         if determine_ext(url) == u'm3u8':
 429             return self._download_m3u8_with_ffmpeg(filename, url)
 430
 431         tmpfilename = self.temp_name(filename)
 432         stream = None
 433
 434         # Do not include the Accept-Encoding header
 435         headers = {'Youtubedl-no-compression': 'True'}
 436         if 'user_agent' in info_dict:
 437             headers['Youtubedl-user-agent'] = info_dict['user_agent']
 438         basic_request = compat_urllib_request.Request(url, None, headers)
 439         request = compat_urllib_request.Request(url, None, headers)
 440
 441         if self.params.get('test', False):
 442             request.add_header('Range','bytes=0-10240')
 443
 444         # Establish possible resume length
 445         if os.path.isfile(encodeFilename(tmpfilename)):
 446             resume_len = os.path.getsize(encodeFilename(tmpfilename))
 447         else:
 448             resume_len = 0
 449
 450         open_mode = 'wb'
 451         if resume_len != 0:
 452             if self.params.get('continuedl', False):
 453                 self.report_resuming_byte(resume_len)
 454                 request.add_header('Range','bytes=%d-' % resume_len)
 455                 open_mode = 'ab'
 456             else:
 457                 resume_len = 0
 458
 459         count = 0
 460         retries = self.params.get('retries', 0)
 461         while count <= retries:
 462             # Establish connection
 463             try:
 464                 if count == 0 and 'urlhandle' in info_dict:
 465                     data = info_dict['urlhandle']
 466                 data = compat_urllib_request.urlopen(request)
 467                 break
 468             except (compat_urllib_error.HTTPError, ) as err:
 469                 if (err.code < 500 or err.code >= 600) and err.code != 416:
 470                     # Unexpected HTTP error
 471                     raise
 472                 elif err.code == 416:
 473                     # Unable to resume (requested range not satisfiable)
 474                     try:
 475                         # Open the connection again without the range header
 476                         data = compat_urllib_request.urlopen(basic_request)
 477                         content_length = data.info()['Content-Length']
 478                     except (compat_urllib_error.HTTPError, ) as err:
 479                         if err.code < 500 or err.code >= 600:
 480                             raise
 481                     else:
 482                         # Examine the reported length
 483                         if (content_length is not None and
 484                                 (resume_len - 100 < int(content_length) < resume_len + 100)):
 485                             # The file had already been fully downloaded.
 486                             # Explanation to the above condition: in issue #175 it was revealed that
 487                             # YouTube sometimes adds or removes a few bytes from the end of the file,
 488                             # changing the file size slightly and causing problems for some users. So
 489                             # I decided to implement a suggested change and consider the file
 490                             # completely downloaded if the file size differs less than 100 bytes from
 491                             # the one in the hard drive.
 492                             self.report_file_already_downloaded(filename)
 493                             self.try_rename(tmpfilename, filename)
 494                             self._hook_progress({
 495                                 'filename': filename,
 496                                 'status': 'finished',
 497                             })
 498                             return True
 499                         else:
 500                             # The length does not match, we start the download over
 501                             self.report_unable_to_resume()
 502                             open_mode = 'wb'
 503                             break
 504             # Retry
 505             count += 1
 506             if count <= retries:
 507                 self.report_retry(count, retries)
 508
 509         if count > retries:
 510             self.report_error(u'giving up after %s retries' % retries)
 511             return False
 512
 513         data_len = data.info().get('Content-length', None)
 514         if data_len is not None:
 515             data_len = int(data_len) + resume_len
 516             min_data_len = self.params.get("min_filesize", None)
 517             max_data_len =  self.params.get("max_filesize", None)
 518             if min_data_len is not None and data_len < min_data_len:
 519                 self.to_screen(u'\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
 520                 return False
 521             if max_data_len is not None and data_len > max_data_len:
 522                 self.to_screen(u'\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
 523                 return False
 524
 525         data_len_str = self.format_bytes(data_len)
 526         byte_counter = 0 + resume_len
 527         block_size = self.params.get('buffersize', 1024)
 528         start = time.time()
 529         while True:
 530             # Download and write
 531             before = time.time()
 532             data_block = data.read(block_size)
 533             after = time.time()
 534             if len(data_block) == 0:
 535                 break
 536             byte_counter += len(data_block)
 537
 538             # Open file just in time
 539             if stream is None:
 540                 try:
 541                     (stream, tmpfilename) = sanitize_open(tmpfilename, open_mode)
 542                     assert stream is not None
 543                     filename = self.undo_temp_name(tmpfilename)
 544                     self.report_destination(filename)
 545                 except (OSError, IOError) as err:
 546                     self.report_error(u'unable to open for writing: %s' % str(err))
 547                     return False
 548             try:
 549                 stream.write(data_block)
 550             except (IOError, OSError) as err:
 551                 self.to_stderr(u"\n")
 552                 self.report_error(u'unable to write data: %s' % str(err))
 553                 return False
 554             if not self.params.get('noresizebuffer', False):
 555                 block_size = self.best_block_size(after - before, len(data_block))
 556
 557             # Progress message
 558             speed = self.calc_speed(start, time.time(), byte_counter - resume_len)
 559             if data_len is None:
 560                 self.report_progress('Unknown %', data_len_str, speed, 'Unknown ETA')
 561                 eta = None
 562             else:
 563                 percent = self.calc_percent(byte_counter, data_len)
 564                 eta = self.calc_eta(start, time.time(), data_len - resume_len, byte_counter - resume_len)
 565                 self.report_progress(percent, data_len_str, speed, eta)
 566
 567             self._hook_progress({
 568                 'downloaded_bytes': byte_counter,
 569                 'total_bytes': data_len,
 570                 'tmpfilename': tmpfilename,
 571                 'filename': filename,
 572                 'status': 'downloading',
 573                 'eta': eta,
 574                 'speed': speed,
 575             })
 576
 577             # Apply rate limit
 578             self.slow_down(start, byte_counter - resume_len)
 579
 580         if stream is None:
 581             self.to_stderr(u"\n")
 582             self.report_error(u'Did not get any data blocks')
 583             return False
 584         stream.close()
 585         self.report_finish(data_len_str, (time.time() - start))
 586         if data_len is not None and byte_counter != data_len:
 587             raise ContentTooShortError(byte_counter, int(data_len))
 588         self.try_rename(tmpfilename, filename)
 589
 590         # Update file modification time
 591         if self.params.get('updatetime', True):
 592             info_dict['filetime'] = self.try_utime(filename, data.info().get('last-modified', None))
 593
 594         self._hook_progress({
 595             'downloaded_bytes': byte_counter,
 596             'total_bytes': byte_counter,
 597             'filename': filename,
 598             'status': 'finished',
 599         })
 600
 601         return True
 602
 603     def _hook_progress(self, status):
 604         for ph in self._progress_hooks:
 605             ph(status)
 606
 607     def add_progress_hook(self, ph):
 608         """ ph gets called on download progress, with a dictionary with the entries
 609         * filename: The final filename
 610         * status: One of "downloading" and "finished"
 611
 612         It can also have some of the following entries:
 613
 614         * downloaded_bytes: Bytes on disks
 615         * total_bytes: Total bytes, None if unknown
 616         * tmpfilename: The filename we're currently writing to
 617         * eta: The estimated time in seconds, None if unknown
 618         * speed: The download speed in bytes/second, None if unknown
 619
 620         Hooks are guaranteed to be called at least once (with status "finished")
 621         if the download is successful.
 622         """
 623         self._progress_hooks.append(ph)