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