youtube_dl/FileDownloader.py

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