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 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] ' + msg)
212 self.to_screen(u'\r%s[download] %s' % (clear_line, msg),
213 skip_eol=not is_last_line)
214 self.to_console_title(u'youtube-dl ' + msg)
216 def report_progress(self, percent, data_len_str, speed, eta):
217 """Report download progress."""
218 if self.params.get('noprogress', False):
221 eta_str = self.format_eta(eta)
223 eta_str = 'Unknown ETA'
224 if percent is not None:
225 percent_str = self.format_percent(percent)
227 percent_str = 'Unknown %'
228 speed_str = self.format_speed(speed)
230 msg = (u'%s of %s at %s ETA %s' %
231 (percent_str, data_len_str, speed_str, eta_str))
232 self._report_progress_status(msg)
234 def report_finish(self, data_len_str, tot_time):
235 """Report download finished."""
236 if self.params.get('noprogress', False):
237 self.to_screen(u'[download] Download completed')
239 self._report_progress_status(
240 (u'100%% of %s in %s' %
241 (data_len_str, self.format_seconds(tot_time))),
244 def report_resuming_byte(self, resume_len):
245 """Report attempt to resume at given byte."""
246 self.to_screen(u'[download] Resuming download at byte %s' % resume_len)
248 def report_retry(self, count, retries):
249 """Report retry in case of HTTP error 5xx"""
250 self.to_screen(u'[download] Got server HTTP error. Retrying (attempt %d of %d)...' % (count, retries))
252 def report_file_already_downloaded(self, file_name):
253 """Report file has already been fully downloaded."""
255 self.to_screen(u'[download] %s has already been downloaded' % file_name)
256 except UnicodeEncodeError:
257 self.to_screen(u'[download] The file has already been downloaded')
259 def report_unable_to_resume(self):
260 """Report it was impossible to resume download."""
261 self.to_screen(u'[download] Unable to resume')
263 def _download_with_rtmpdump(self, filename, url, player_url, page_url, play_path, tc_url, live):
264 def run_rtmpdump(args):
266 resume_percent = None
267 resume_downloaded_data_len = None
268 proc = subprocess.Popen(args, stderr=subprocess.PIPE)
269 cursor_in_new_line = True
270 proc_stderr_closed = False
271 while not proc_stderr_closed:
272 # read line from stderr
275 char = proc.stderr.read(1)
277 proc_stderr_closed = True
279 if char in [b'\r', b'\n']:
281 line += char.decode('ascii', 'replace')
283 # proc_stderr_closed is True
285 mobj = re.search(r'([0-9]+\.[0-9]{3}) kB / [0-9]+\.[0-9]{2} sec \(([0-9]{1,2}\.[0-9])%\)', line)
287 downloaded_data_len = int(float(mobj.group(1))*1024)
288 percent = float(mobj.group(2))
289 if not resume_percent:
290 resume_percent = percent
291 resume_downloaded_data_len = downloaded_data_len
292 eta = self.calc_eta(start, time.time(), 100-resume_percent, percent-resume_percent)
293 speed = self.calc_speed(start, time.time(), downloaded_data_len-resume_downloaded_data_len)
296 data_len = int(downloaded_data_len * 100 / percent)
297 data_len_str = u'~' + format_bytes(data_len)
298 self.report_progress(percent, data_len_str, speed, eta)
299 cursor_in_new_line = False
300 self._hook_progress({
301 'downloaded_bytes': downloaded_data_len,
302 'total_bytes': data_len,
303 'tmpfilename': tmpfilename,
304 'filename': filename,
305 'status': 'downloading',
309 elif self.params.get('verbose', False):
310 if not cursor_in_new_line:
312 cursor_in_new_line = True
313 self.to_screen(u'[rtmpdump] '+line)
315 if not cursor_in_new_line:
317 return proc.returncode
319 self.report_destination(filename)
320 tmpfilename = self.temp_name(filename)
321 test = self.params.get('test', False)
323 # Check for rtmpdump first
325 subprocess.call(['rtmpdump', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
326 except (OSError, IOError):
327 self.report_error(u'RTMP download detected but "rtmpdump" could not be run')
330 # Download using rtmpdump. rtmpdump returns exit code 2 when
331 # the connection was interrumpted and resuming appears to be
332 # possible. This is part of rtmpdump's normal usage, AFAIK.
333 basic_args = ['rtmpdump', '--verbose', '-r', url, '-o', tmpfilename]
334 if player_url is not None:
335 basic_args += ['--swfVfy', player_url]
336 if page_url is not None:
337 basic_args += ['--pageUrl', page_url]
338 if play_path is not None:
339 basic_args += ['--playpath', play_path]
340 if tc_url is not None:
341 basic_args += ['--tcUrl', url]
343 basic_args += ['--stop', '1']
345 basic_args += ['--live']
346 args = basic_args + [[], ['--resume', '--skip', '1']][self.params.get('continuedl', False)]
348 if sys.platform == 'win32' and sys.version_info < (3, 0):
349 # Windows subprocess module does not actually support Unicode
351 # See http://stackoverflow.com/a/9951851/35070
352 subprocess_encoding = sys.getfilesystemencoding()
353 args = [a.encode(subprocess_encoding, 'ignore') for a in args]
355 subprocess_encoding = None
357 if self.params.get('verbose', False):
358 if subprocess_encoding:
360 a.decode(subprocess_encoding) if isinstance(a, bytes) else a
366 shell_quote = lambda args: ' '.join(map(pipes.quote, str_args))
369 self.to_screen(u'[debug] rtmpdump command line: ' + shell_quote(str_args))
371 retval = run_rtmpdump(args)
373 while (retval == 2 or retval == 1) and not test:
374 prevsize = os.path.getsize(encodeFilename(tmpfilename))
375 self.to_screen(u'[rtmpdump] %s bytes' % prevsize)
376 time.sleep(5.0) # This seems to be needed
377 retval = run_rtmpdump(basic_args + ['-e'] + [[], ['-k', '1']][retval == 1])
378 cursize = os.path.getsize(encodeFilename(tmpfilename))
379 if prevsize == cursize and retval == 1:
381 # Some rtmp streams seem abort after ~ 99.8%. Don't complain for those
382 if prevsize == cursize and retval == 2 and cursize > 1024:
383 self.to_screen(u'[rtmpdump] Could not download the whole video. This can happen for some advertisements.')
386 if retval == 0 or (test and retval == 2):
387 fsize = os.path.getsize(encodeFilename(tmpfilename))
388 self.to_screen(u'[rtmpdump] %s bytes' % fsize)
389 self.try_rename(tmpfilename, filename)
390 self._hook_progress({
391 'downloaded_bytes': fsize,
392 'total_bytes': fsize,
393 'filename': filename,
394 'status': 'finished',
398 self.to_stderr(u"\n")
399 self.report_error(u'rtmpdump exited with code %d' % retval)
402 def _download_with_mplayer(self, filename, url):
403 self.report_destination(filename)
404 tmpfilename = self.temp_name(filename)
406 args = ['mplayer', '-really-quiet', '-vo', 'null', '-vc', 'dummy', '-dumpstream', '-dumpfile', tmpfilename, url]
407 # Check for mplayer first
409 subprocess.call(['mplayer', '-h'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
410 except (OSError, IOError):
411 self.report_error(u'MMS or RTSP download detected but "%s" could not be run' % args[0] )
414 # Download using mplayer.
415 retval = subprocess.call(args)
417 fsize = os.path.getsize(encodeFilename(tmpfilename))
418 self.to_screen(u'\r[%s] %s bytes' % (args[0], fsize))
419 self.try_rename(tmpfilename, filename)
420 self._hook_progress({
421 'downloaded_bytes': fsize,
422 'total_bytes': fsize,
423 'filename': filename,
424 'status': 'finished',
428 self.to_stderr(u"\n")
429 self.report_error(u'mplayer exited with code %d' % retval)
432 def _download_m3u8_with_ffmpeg(self, filename, url):
433 self.report_destination(filename)
434 tmpfilename = self.temp_name(filename)
436 args = ['-y', '-i', url, '-f', 'mp4', '-c', 'copy',
437 '-bsf:a', 'aac_adtstoasc', tmpfilename]
439 for program in ['avconv', 'ffmpeg']:
441 subprocess.call([program, '-version'], stdout=(open(os.path.devnull, 'w')), stderr=subprocess.STDOUT)
443 except (OSError, IOError):
446 self.report_error(u'm3u8 download detected but ffmpeg or avconv could not be found')
447 cmd = [program] + args
449 retval = subprocess.call(cmd)
451 fsize = os.path.getsize(encodeFilename(tmpfilename))
452 self.to_screen(u'\r[%s] %s bytes' % (args[0], fsize))
453 self.try_rename(tmpfilename, filename)
454 self._hook_progress({
455 'downloaded_bytes': fsize,
456 'total_bytes': fsize,
457 'filename': filename,
458 'status': 'finished',
462 self.to_stderr(u"\n")
463 self.report_error(u'ffmpeg exited with code %d' % retval)
467 def _do_download(self, filename, info_dict):
468 url = info_dict['url']
470 # Check file already present
471 if self.params.get('continuedl', False) and os.path.isfile(encodeFilename(filename)) and not self.params.get('nopart', False):
472 self.report_file_already_downloaded(filename)
473 self._hook_progress({
474 'filename': filename,
475 'status': 'finished',
476 'total_bytes': os.path.getsize(encodeFilename(filename)),
480 # Attempt to download using rtmpdump
481 if url.startswith('rtmp'):
482 return self._download_with_rtmpdump(filename, url,
483 info_dict.get('player_url', None),
484 info_dict.get('page_url', None),
485 info_dict.get('play_path', None),
486 info_dict.get('tc_url', None),
487 info_dict.get('rtmp_live', False))
489 # Attempt to download using mplayer
490 if url.startswith('mms') or url.startswith('rtsp'):
491 return self._download_with_mplayer(filename, url)
493 # m3u8 manifest are downloaded with ffmpeg
494 if determine_ext(url) == u'm3u8':
495 return self._download_m3u8_with_ffmpeg(filename, url)
497 tmpfilename = self.temp_name(filename)
500 # Do not include the Accept-Encoding header
501 headers = {'Youtubedl-no-compression': 'True'}
502 if 'user_agent' in info_dict:
503 headers['Youtubedl-user-agent'] = info_dict['user_agent']
504 basic_request = compat_urllib_request.Request(url, None, headers)
505 request = compat_urllib_request.Request(url, None, headers)
507 if self.params.get('test', False):
508 request.add_header('Range','bytes=0-10240')
510 # Establish possible resume length
511 if os.path.isfile(encodeFilename(tmpfilename)):
512 resume_len = os.path.getsize(encodeFilename(tmpfilename))
518 if self.params.get('continuedl', False):
519 self.report_resuming_byte(resume_len)
520 request.add_header('Range','bytes=%d-' % resume_len)
526 retries = self.params.get('retries', 0)
527 while count <= retries:
528 # Establish connection
530 if count == 0 and 'urlhandle' in info_dict:
531 data = info_dict['urlhandle']
532 data = compat_urllib_request.urlopen(request)
534 except (compat_urllib_error.HTTPError, ) as err:
535 if (err.code < 500 or err.code >= 600) and err.code != 416:
536 # Unexpected HTTP error
538 elif err.code == 416:
539 # Unable to resume (requested range not satisfiable)
541 # Open the connection again without the range header
542 data = compat_urllib_request.urlopen(basic_request)
543 content_length = data.info()['Content-Length']
544 except (compat_urllib_error.HTTPError, ) as err:
545 if err.code < 500 or err.code >= 600:
548 # Examine the reported length
549 if (content_length is not None and
550 (resume_len - 100 < int(content_length) < resume_len + 100)):
551 # The file had already been fully downloaded.
552 # Explanation to the above condition: in issue #175 it was revealed that
553 # YouTube sometimes adds or removes a few bytes from the end of the file,
554 # changing the file size slightly and causing problems for some users. So
555 # I decided to implement a suggested change and consider the file
556 # completely downloaded if the file size differs less than 100 bytes from
557 # the one in the hard drive.
558 self.report_file_already_downloaded(filename)
559 self.try_rename(tmpfilename, filename)
560 self._hook_progress({
561 'filename': filename,
562 'status': 'finished',
566 # The length does not match, we start the download over
567 self.report_unable_to_resume()
573 self.report_retry(count, retries)
576 self.report_error(u'giving up after %s retries' % retries)
579 data_len = data.info().get('Content-length', None)
580 if data_len is not None:
581 data_len = int(data_len) + resume_len
582 min_data_len = self.params.get("min_filesize", None)
583 max_data_len = self.params.get("max_filesize", None)
584 if min_data_len is not None and data_len < min_data_len:
585 self.to_screen(u'\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len, min_data_len))
587 if max_data_len is not None and data_len > max_data_len:
588 self.to_screen(u'\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len, max_data_len))
591 data_len_str = format_bytes(data_len)
592 byte_counter = 0 + resume_len
593 block_size = self.params.get('buffersize', 1024)
598 data_block = data.read(block_size)
600 if len(data_block) == 0:
602 byte_counter += len(data_block)
604 # Open file just in time
607 (stream, tmpfilename) = sanitize_open(tmpfilename, open_mode)
608 assert stream is not None
609 filename = self.undo_temp_name(tmpfilename)
610 self.report_destination(filename)
611 except (OSError, IOError) as err:
612 self.report_error(u'unable to open for writing: %s' % str(err))
615 stream.write(data_block)
616 except (IOError, OSError) as err:
617 self.to_stderr(u"\n")
618 self.report_error(u'unable to write data: %s' % str(err))
620 if not self.params.get('noresizebuffer', False):
621 block_size = self.best_block_size(after - before, len(data_block))
624 speed = self.calc_speed(start, time.time(), byte_counter - resume_len)
628 percent = self.calc_percent(byte_counter, data_len)
629 eta = self.calc_eta(start, time.time(), data_len - resume_len, byte_counter - resume_len)
630 self.report_progress(percent, data_len_str, speed, eta)
632 self._hook_progress({
633 'downloaded_bytes': byte_counter,
634 'total_bytes': data_len,
635 'tmpfilename': tmpfilename,
636 'filename': filename,
637 'status': 'downloading',
643 self.slow_down(start, byte_counter - resume_len)
646 self.to_stderr(u"\n")
647 self.report_error(u'Did not get any data blocks')
650 self.report_finish(data_len_str, (time.time() - start))
651 if data_len is not None and byte_counter != data_len:
652 raise ContentTooShortError(byte_counter, int(data_len))
653 self.try_rename(tmpfilename, filename)
655 # Update file modification time
656 if self.params.get('updatetime', True):
657 info_dict['filetime'] = self.try_utime(filename, data.info().get('last-modified', None))
659 self._hook_progress({
660 'downloaded_bytes': byte_counter,
661 'total_bytes': byte_counter,
662 'filename': filename,
663 'status': 'finished',
668 def _hook_progress(self, status):
669 for ph in self._progress_hooks:
672 def add_progress_hook(self, ph):
673 """ ph gets called on download progress, with a dictionary with the entries
674 * filename: The final filename
675 * status: One of "downloading" and "finished"
677 It can also have some of the following entries:
679 * downloaded_bytes: Bytes on disks
680 * total_bytes: Total bytes, None if unknown
681 * tmpfilename: The filename we're currently writing to
682 * eta: The estimated time in seconds, None if unknown
683 * speed: The download speed in bytes/second, None if unknown
685 Hooks are guaranteed to be called at least once (with status "finished")
686 if the download is successful.
688 self._progress_hooks.append(ph)