19 class FileDownloader(object):
20 """File Downloader class.
22 File downloader objects are the ones responsible of downloading the
23 actual video file and writing it to disk.
25 File downloaders accept a lot of parameters. In order not to saturate
26 the object constructor with arguments, it receives a dictionary of
31 verbose: Print additional info to stdout.
32 quiet: Do not print messages to stdout.
33 ratelimit: Download speed limit, in bytes/sec.
34 retries: Number of times to retry for HTTP error 5xx
35 buffersize: Size of download buffer in bytes.
36 noresizebuffer: Do not automatically resize the download buffer.
37 continuedl: Try to continue downloads if possible.
38 noprogress: Do not print the progress bar.
39 logtostderr: Log messages to stderr instead of stdout.
40 consoletitle: Display progress in console window's titlebar.
41 nopart: Do not use temporary .part files.
42 updatetime: Use the Last-modified header to set output file timestamps.
43 test: Download only first bytes to test the downloader.
44 min_filesize: Skip files smaller than this size
45 max_filesize: Skip files larger than this size
50 def __init__(self, ydl, params):
51 """Create a FileDownloader object with the given options."""
53 self._progress_hooks = []
57 def format_seconds(seconds):
58 (mins, secs) = divmod(seconds, 60)
59 (hours, mins) = divmod(mins, 60)
63 return '%02d:%02d' % (mins, secs)
65 return '%02d:%02d:%02d' % (hours, mins, secs)
68 def calc_percent(byte_counter, data_len):
71 return float(byte_counter) / float(data_len) * 100.0
74 def format_percent(percent):
77 return '%6s' % ('%3.1f%%' % percent)
80 def calc_eta(start, now, total, current):
84 if current == 0 or dif < 0.001: # One millisecond
86 rate = float(current) / dif
87 return int((float(total) - float(current)) / rate)
93 return FileDownloader.format_seconds(eta)
96 def calc_speed(start, now, bytes):
98 if bytes == 0 or dif < 0.001: # One millisecond
100 return float(bytes) / dif
103 def format_speed(speed):
105 return '%10s' % '---b/s'
106 return '%10s' % ('%s/s' % format_bytes(speed))
109 def best_block_size(elapsed_time, bytes):
110 new_min = max(bytes / 2.0, 1.0)
111 new_max = min(max(bytes * 2.0, 1.0), 4194304) # Do not surpass 4 MB
112 if elapsed_time < 0.001:
114 rate = bytes / elapsed_time
122 def parse_bytes(bytestr):
123 """Parse a string indicating a byte quantity into an integer."""
124 matchobj = re.match(r'(?i)^(\d+(?:\.\d+)?)([kMGTPEZY]?)$', bytestr)
127 number = float(matchobj.group(1))
128 multiplier = 1024.0 ** 'bkmgtpezy'.index(matchobj.group(2).lower())
129 return int(round(number * multiplier))
131 def to_screen(self, *args, **kargs):
132 self.ydl.to_screen(*args, **kargs)
134 def to_stderr(self, message):
135 self.ydl.to_screen(message)
137 def to_console_title(self, message):
138 self.ydl.to_console_title(message)
140 def trouble(self, *args, **kargs):
141 self.ydl.trouble(*args, **kargs)
143 def report_warning(self, *args, **kargs):
144 self.ydl.report_warning(*args, **kargs)
146 def report_error(self, *args, **kargs):
147 self.ydl.report_error(*args, **kargs)
149 def slow_down(self, start_time, byte_counter):
150 """Sleep if the download speed is over the rate limit."""
151 rate_limit = self.params.get('ratelimit', None)
152 if rate_limit is None or byte_counter == 0:
155 elapsed = now - start_time
158 speed = float(byte_counter) / elapsed
159 if speed > rate_limit:
160 time.sleep((byte_counter - rate_limit * (now - start_time)) / rate_limit)
162 def temp_name(self, filename):
163 """Returns a temporary filename for the given filename."""
164 if self.params.get('nopart', False) or filename == u'-' or \
165 (os.path.exists(encodeFilename(filename)) and not os.path.isfile(encodeFilename(filename))):
167 return filename + u'.part'
169 def undo_temp_name(self, filename):
170 if filename.endswith(u'.part'):
171 return filename[:-len(u'.part')]
174 def try_rename(self, old_filename, new_filename):
176 if old_filename == new_filename:
178 os.rename(encodeFilename(old_filename), encodeFilename(new_filename))
179 except (IOError, OSError):
180 self.report_error(u'unable to rename file')
182 def try_utime(self, filename, last_modified_hdr):
183 """Try to set the last-modified time of the given file."""
184 if last_modified_hdr is None:
186 if not os.path.isfile(encodeFilename(filename)):
188 timestr = last_modified_hdr
191 filetime = timeconvert(timestr)
194 # Ignore obviously invalid dates
198 os.utime(filename, (time.time(), filetime))
203 def report_destination(self, filename):
204 """Report destination filename."""
205 self.to_screen(u'[download] Destination: ' + filename)
207 def _report_progress_status(self, msg, is_last_line=False):
208 fullmsg = u'[download] ' + msg
209 if self.params.get('progress_with_newline', False):
210 self.to_screen(fullmsg)
213 prev_len = getattr(self, '_report_progress_prev_line_length',
215 if prev_len > len(fullmsg):
216 fullmsg += u' ' * (prev_len - len(fullmsg))
217 self._report_progress_prev_line_length = len(fullmsg)
220 clear_line = (u'\r\x1b[K' if sys.stderr.isatty() else u'\r')
221 self.to_screen(clear_line + fullmsg, skip_eol=not is_last_line)
222 self.to_console_title(u'youtube-dl ' + msg)
224 def report_progress(self, percent, data_len_str, speed, eta):
225 """Report download progress."""
226 if self.params.get('noprogress', False):
229 eta_str = self.format_eta(eta)
231 eta_str = 'Unknown ETA'
232 if percent is not None:
233 percent_str = self.format_percent(percent)
235 percent_str = 'Unknown %'
236 speed_str = self.format_speed(speed)
238 msg = (u'%s of %s at %s ETA %s' %
239 (percent_str, data_len_str, speed_str, eta_str))
240 self._report_progress_status(msg)
242 def report_finish(self, data_len_str, tot_time):
243 """Report download finished."""
244 if self.params.get('noprogress', False):
245 self.to_screen(u'[download] Download completed')
247 self._report_progress_status(
248 (u'100%% of %s in %s' %
249 (data_len_str, self.format_seconds(tot_time))),
252 def report_resuming_byte(self, resume_len):
253 """Report attempt to resume at given byte."""
254 self.to_screen(u'[download] Resuming download at byte %s' % resume_len)
256 def report_retry(self, count, retries):
257 """Report retry in case of HTTP error 5xx"""
258 self.to_screen(u'[download] Got server HTTP error. Retrying (attempt %d of %d)...' % (count, retries))
260 def report_file_already_downloaded(self, file_name):
261 """Report file has already been fully downloaded."""
263 self.to_screen(u'[download] %s has already been downloaded' % file_name)
264 except UnicodeEncodeError:
265 self.to_screen(u'[download] The file has already been downloaded')
267 def report_unable_to_resume(self):
268 """Report it was impossible to resume download."""
269 self.to_screen(u'[download] Unable to resume')
271 def _download_with_rtmpdump(self, filename, url, player_url, page_url, play_path, tc_url, live):
272 def run_rtmpdump(args):
274 resume_percent = None
275 resume_downloaded_data_len = None
276 proc = subprocess.Popen(args, stderr=subprocess.PIPE)
277 cursor_in_new_line = True
278 proc_stderr_closed = False
279 while not proc_stderr_closed:
280 # read line from stderr
283 char = proc.stderr.read(1)
285 proc_stderr_closed = True
287 if char in [b'\r', b'\n']:
289 line += char.decode('ascii', 'replace')
291 # proc_stderr_closed is True
293 mobj = re.search(r'([0-9]+\.[0-9]{3}) kB / [0-9]+\.[0-9]{2} sec \(([0-9]{1,2}\.[0-9])%\)', line)
295 downloaded_data_len = int(float(mobj.group(1))*1024)
296 percent = float(mobj.group(2))
297 if not resume_percent:
298 resume_percent = percent
299 resume_downloaded_data_len = downloaded_data_len
300 eta = self.calc_eta(start, time.time(), 100-resume_percent, percent-resume_percent)
301 speed = self.calc_speed(start, time.time(), downloaded_data_len-resume_downloaded_data_len)
304 data_len = int(downloaded_data_len * 100 / percent)
305 data_len_str = u'~' + format_bytes(data_len)
306 self.report_progress(percent, data_len_str, speed, eta)
307 cursor_in_new_line = False
308 self._hook_progress({
309 'downloaded_bytes': downloaded_data_len,
310 'total_bytes': data_len,
311 'tmpfilename': tmpfilename,
312 'filename': filename,
313 'status': 'downloading',
317 elif self.params.get('verbose', False):
318 if not cursor_in_new_line:
320 cursor_in_new_line = True
321 self.to_screen(u'[rtmpdump] '+line)
323 if not cursor_in_new_line:
325 return proc.returncode
327 self.report_destination(filename)
328 tmpfilename = self.temp_name(filename)
329 test = self.params.get('test', False)
331 # Check for rtmpdump first
333 subprocess.call(['rtmpdump', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
334 except (OSError, IOError):
335 self.report_error(u'RTMP download detected but "rtmpdump" could not be run')
338 # Download using rtmpdump. rtmpdump returns exit code 2 when
339 # the connection was interrumpted and resuming appears to be
340 # possible. This is part of rtmpdump's normal usage, AFAIK.
341 basic_args = ['rtmpdump', '--verbose', '-r', url, '-o', tmpfilename]
342 if player_url is not None:
343 basic_args += ['--swfVfy', player_url]
344 if page_url is not None:
345 basic_args += ['--pageUrl', page_url]
346 if play_path is not None:
347 basic_args += ['--playpath', play_path]
348 if tc_url is not None:
349 basic_args += ['--tcUrl', url]
351 basic_args += ['--stop', '1']
353 basic_args += ['--live']
354 args = basic_args + [[], ['--resume', '--skip', '1']][self.params.get('continuedl', False)]
356 if sys.platform == 'win32' and sys.version_info < (3, 0):
357 # Windows subprocess module does not actually support Unicode
359 # See http://stackoverflow.com/a/9951851/35070
360 subprocess_encoding = sys.getfilesystemencoding()
361 args = [a.encode(subprocess_encoding, 'ignore') for a in args]
363 subprocess_encoding = None
365 if self.params.get('verbose', False):
366 if subprocess_encoding:
368 a.decode(subprocess_encoding) if isinstance(a, bytes) else a
374 shell_quote = lambda args: ' '.join(map(pipes.quote, str_args))
377 self.to_screen(u'[debug] rtmpdump command line: ' + shell_quote(str_args))
379 retval = run_rtmpdump(args)
381 while (retval == 2 or retval == 1) and not test:
382 prevsize = os.path.getsize(encodeFilename(tmpfilename))
383 self.to_screen(u'[rtmpdump] %s bytes' % prevsize)
384 time.sleep(5.0) # This seems to be needed
385 retval = run_rtmpdump(basic_args + ['-e'] + [[], ['-k', '1']][retval == 1])
386 cursize = os.path.getsize(encodeFilename(tmpfilename))
387 if prevsize == cursize and retval == 1:
389 # Some rtmp streams seem abort after ~ 99.8%. Don't complain for those
390 if prevsize == cursize and retval == 2 and cursize > 1024:
391 self.to_screen(u'[rtmpdump] Could not download the whole video. This can happen for some advertisements.')
394 if retval == 0 or (test and retval == 2):
395 fsize = os.path.getsize(encodeFilename(tmpfilename))
396 self.to_screen(u'[rtmpdump] %s bytes' % fsize)
397 self.try_rename(tmpfilename, filename)
398 self._hook_progress({
399 'downloaded_bytes': fsize,
400 'total_bytes': fsize,
401 'filename': filename,
402 'status': 'finished',
406 self.to_stderr(u"\n")
407 self.report_error(u'rtmpdump exited with code %d' % retval)
410 def _download_with_mplayer(self, filename, url):
411 self.report_destination(filename)
412 tmpfilename = self.temp_name(filename)
414 args = ['mplayer', '-really-quiet', '-vo', 'null', '-vc', 'dummy', '-dumpstream', '-dumpfile', tmpfilename, url]
415 # Check for mplayer first
417 subprocess.call(['mplayer', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
418 except (OSError, IOError):
419 self.report_error(u'MMS or RTSP download detected but "%s" could not be run' % args[0] )
422 # Download using mplayer.
423 retval = subprocess.call(args)
425 fsize = os.path.getsize(encodeFilename(tmpfilename))
426 self.to_screen(u'\r[%s] %s bytes' % (args[0], fsize))
427 self.try_rename(tmpfilename, filename)
428 self._hook_progress({
429 'downloaded_bytes': fsize,
430 'total_bytes': fsize,
431 'filename': filename,
432 'status': 'finished',
436 self.to_stderr(u"\n")
437 self.report_error(u'mplayer exited with code %d' % retval)
440 def _download_m3u8_with_ffmpeg(self, filename, url):
441 self.report_destination(filename)
442 tmpfilename = self.temp_name(filename)
444 args = ['-y', '-i', url, '-f', 'mp4', '-c', 'copy',
445 '-bsf:a', 'aac_adtstoasc', tmpfilename]
447 for program in ['avconv', 'ffmpeg']:
449 subprocess.call([program, '-version'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
451 except (OSError, IOError):
454 self.report_error(u'm3u8 download detected but ffmpeg or avconv could not be found')
455 cmd = [program] + args
457 retval = subprocess.call(cmd)
459 fsize = os.path.getsize(encodeFilename(tmpfilename))
460 self.to_screen(u'\r[%s] %s bytes' % (args[0], fsize))
461 self.try_rename(tmpfilename, filename)
462 self._hook_progress({
463 'downloaded_bytes': fsize,
464 'total_bytes': fsize,
465 'filename': filename,
466 'status': 'finished',
470 self.to_stderr(u"\n")
471 self.report_error(u'ffmpeg exited with code %d' % retval)
475 def _do_download(self, filename, info_dict):
476 url = info_dict['url']
478 # Check file already present
479 if self.params.get('continuedl', False) and os.path.isfile(encodeFilename(filename)) and not self.params.get('nopart', False):
480 self.report_file_already_downloaded(filename)
481 self._hook_progress({
482 'filename': filename,
483 'status': 'finished',
484 'total_bytes': os.path.getsize(encodeFilename(filename)),
488 # Attempt to download using rtmpdump
489 if url.startswith('rtmp'):
490 return self._download_with_rtmpdump(filename, url,
491 info_dict.get('player_url', None),
492 info_dict.get('page_url', None),
493 info_dict.get('play_path', None),
494 info_dict.get('tc_url', None),
495 info_dict.get('rtmp_live', False))
497 # Attempt to download using mplayer
498 if url.startswith('mms') or url.startswith('rtsp'):
499 return self._download_with_mplayer(filename, url)
501 # m3u8 manifest are downloaded with ffmpeg
502 if determine_ext(url) == u'm3u8':
503 return self._download_m3u8_with_ffmpeg(filename, url)
505 tmpfilename = self.temp_name(filename)
508 # Do not include the Accept-Encoding header
509 headers = {'Youtubedl-no-compression': 'True'}
510 if 'user_agent' in info_dict:
511 headers['Youtubedl-user-agent'] = info_dict['user_agent']
512 basic_request = compat_urllib_request.Request(url, None, headers)
513 request = compat_urllib_request.Request(url, None, headers)
515 if self.params.get('test', False):
516 request.add_header('Range','bytes=0-10240')
518 # Establish possible resume length
519 if os.path.isfile(encodeFilename(tmpfilename)):
520 resume_len = os.path.getsize(encodeFilename(tmpfilename))
526 if self.params.get('continuedl', False):
527 self.report_resuming_byte(resume_len)
528 request.add_header('Range','bytes=%d-' % resume_len)
534 retries = self.params.get('retries', 0)
535 while count <= retries:
536 # Establish connection
538 if count == 0 and 'urlhandle' in info_dict:
539 data = info_dict['urlhandle']
540 data = compat_urllib_request.urlopen(request)
542 except (compat_urllib_error.HTTPError, ) as err:
543 if (err.code < 500 or err.code >= 600) and err.code != 416:
544 # Unexpected HTTP error
546 elif err.code == 416:
547 # Unable to resume (requested range not satisfiable)
549 # Open the connection again without the range header
550 data = compat_urllib_request.urlopen(basic_request)
551 content_length = data.info()['Content-Length']
552 except (compat_urllib_error.HTTPError, ) as err:
553 if err.code < 500 or err.code >= 600:
556 # Examine the reported length
557 if (content_length is not None and
558 (resume_len - 100 < int(content_length) < resume_len + 100)):
559 # The file had already been fully downloaded.
560 # Explanation to the above condition: in issue #175 it was revealed that
561 # YouTube sometimes adds or removes a few bytes from the end of the file,
562 # changing the file size slightly and causing problems for some users. So
563 # I decided to implement a suggested change and consider the file
564 # completely downloaded if the file size differs less than 100 bytes from
565 # the one in the hard drive.
566 self.report_file_already_downloaded(filename)
567 self.try_rename(tmpfilename, filename)
568 self._hook_progress({
569 'filename': filename,
570 'status': 'finished',
574 # The length does not match, we start the download over
575 self.report_unable_to_resume()
581 self.report_retry(count, retries)
584 self.report_error(u'giving up after %s retries' % retries)
587 data_len = data.info().get('Content-length', None)
588 if data_len is not None:
589 data_len = int(data_len) + resume_len
590 min_data_len = self.params.get("min_filesize", None)
591 max_data_len = self.params.get("max_filesize", None)
592 if min_data_len is not None and data_len < min_data_len:
593 self.to_screen(u'\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
595 if max_data_len is not None and data_len > max_data_len:
596 self.to_screen(u'\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
599 data_len_str = format_bytes(data_len)
600 byte_counter = 0 + resume_len
601 block_size = self.params.get('buffersize', 1024)
606 data_block = data.read(block_size)
608 if len(data_block) == 0:
610 byte_counter += len(data_block)
612 # Open file just in time
615 (stream, tmpfilename) = sanitize_open(tmpfilename, open_mode)
616 assert stream is not None
617 filename = self.undo_temp_name(tmpfilename)
618 self.report_destination(filename)
619 except (OSError, IOError) as err:
620 self.report_error(u'unable to open for writing: %s' % str(err))
623 stream.write(data_block)
624 except (IOError, OSError) as err:
625 self.to_stderr(u"\n")
626 self.report_error(u'unable to write data: %s' % str(err))
628 if not self.params.get('noresizebuffer', False):
629 block_size = self.best_block_size(after - before, len(data_block))
632 speed = self.calc_speed(start, time.time(), byte_counter - resume_len)
636 percent = self.calc_percent(byte_counter, data_len)
637 eta = self.calc_eta(start, time.time(), data_len - resume_len, byte_counter - resume_len)
638 self.report_progress(percent, data_len_str, speed, eta)
640 self._hook_progress({
641 'downloaded_bytes': byte_counter,
642 'total_bytes': data_len,
643 'tmpfilename': tmpfilename,
644 'filename': filename,
645 'status': 'downloading',
651 self.slow_down(start, byte_counter - resume_len)
654 self.to_stderr(u"\n")
655 self.report_error(u'Did not get any data blocks')
658 self.report_finish(data_len_str, (time.time() - start))
659 if data_len is not None and byte_counter != data_len:
660 raise ContentTooShortError(byte_counter, int(data_len))
661 self.try_rename(tmpfilename, filename)
663 # Update file modification time
664 if self.params.get('updatetime', True):
665 info_dict['filetime'] = self.try_utime(filename, data.info().get('last-modified', None))
667 self._hook_progress({
668 'downloaded_bytes': byte_counter,
669 'total_bytes': byte_counter,
670 'filename': filename,
671 'status': 'finished',
676 def _hook_progress(self, status):
677 for ph in self._progress_hooks:
680 def add_progress_hook(self, ph):
681 """ ph gets called on download progress, with a dictionary with the entries
682 * filename: The final filename
683 * status: One of "downloading" and "finished"
685 It can also have some of the following entries:
687 * downloaded_bytes: Bytes on disks
688 * total_bytes: Total bytes, None if unknown
689 * tmpfilename: The filename we're currently writing to
690 * eta: The estimated time in seconds, None if unknown
691 * speed: The download speed in bytes/second, None if unknown
693 Hooks are guaranteed to be called at least once (with status "finished")
694 if the download is successful.
696 self._progress_hooks.append(ph)