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(self
, percent
, data_len_str
, speed
, eta
):
208 """Report download progress."""
209 if self
.params
.get('noprogress', False):
211 clear_line
= (u
'\x1b[K' if sys
.stderr
.isatty() and os
.name
!= 'nt' else u
'')
213 eta_str
= self
.format_eta(eta
)
215 eta_str
= 'Unknown ETA'
216 if percent
is not None:
217 percent_str
= self
.format_percent(percent
)
219 percent_str
= 'Unknown %'
220 speed_str
= self
.format_speed(speed
)
221 if self
.params
.get('progress_with_newline', False):
222 self
.to_screen(u
'[download] %s of %s at %s ETA %s' %
223 (percent_str
, data_len_str
, speed_str
, eta_str
))
225 self
.to_screen(u
'\r%s[download] %s of %s at %s ETA %s' %
226 (clear_line
, percent_str
, data_len_str
, speed_str
, eta_str
), skip_eol
=True)
227 self
.to_console_title(u
'youtube-dl - %s of %s at %s ETA %s' %
228 (percent_str
.strip(), data_len_str
.strip(), speed_str
.strip(), eta_str
.strip()))
230 def report_resuming_byte(self
, resume_len
):
231 """Report attempt to resume at given byte."""
232 self
.to_screen(u
'[download] Resuming download at byte %s' % resume_len
)
234 def report_retry(self
, count
, retries
):
235 """Report retry in case of HTTP error 5xx"""
236 self
.to_screen(u
'[download] Got server HTTP error. Retrying (attempt %d of %d)...' % (count
, retries
))
238 def report_file_already_downloaded(self
, file_name
):
239 """Report file has already been fully downloaded."""
241 self
.to_screen(u
'[download] %s has already been downloaded' % file_name
)
242 except UnicodeEncodeError:
243 self
.to_screen(u
'[download] The file has already been downloaded')
245 def report_unable_to_resume(self
):
246 """Report it was impossible to resume download."""
247 self
.to_screen(u
'[download] Unable to resume')
249 def report_finish(self
, data_len_str
, tot_time
):
250 """Report download finished."""
251 if self
.params
.get('noprogress', False):
252 self
.to_screen(u
'[download] Download completed')
254 clear_line
= (u
'\x1b[K' if sys
.stderr
.isatty() and os
.name
!= 'nt' else u
'')
255 self
.to_screen(u
'\r%s[download] 100%% of %s in %s' %
256 (clear_line
, data_len_str
, self
.format_seconds(tot_time
)))
258 def _download_with_rtmpdump(self
, filename
, url
, player_url
, page_url
, play_path
, tc_url
, live
):
259 def run_rtmpdump(args
):
261 resume_percent
= None
262 resume_downloaded_data_len
= None
263 proc
= subprocess
.Popen(args
, stderr
=subprocess
.PIPE
)
264 cursor_in_new_line
= True
265 proc_stderr_closed
= False
266 while not proc_stderr_closed
:
267 # read line from stderr
270 char
= proc
.stderr
.read(1)
272 proc_stderr_closed
= True
274 if char
in [b
'\r', b
'\n']:
276 line
+= char
.decode('ascii', 'replace')
278 # proc_stderr_closed is True
280 mobj
= re
.search(r
'([0-9]+\.[0-9]{3}) kB / [0-9]+\.[0-9]{2} sec \(([0-9]{1,2}\.[0-9])%\)', line
)
282 downloaded_data_len
= int(float(mobj
.group(1))*1024)
283 percent
= float(mobj
.group(2))
284 if not resume_percent
:
285 resume_percent
= percent
286 resume_downloaded_data_len
= downloaded_data_len
287 eta
= self
.calc_eta(start
, time
.time(), 100-resume_percent
, percent
-resume_percent
)
288 speed
= self
.calc_speed(start
, time
.time(), downloaded_data_len
-resume_downloaded_data_len
)
291 data_len
= int(downloaded_data_len
* 100 / percent
)
292 data_len_str
= u
'~' + format_bytes(data_len
)
293 self
.report_progress(percent
, data_len_str
, speed
, eta
)
294 cursor_in_new_line
= False
295 self
._hook
_progress
({
296 'downloaded_bytes': downloaded_data_len
,
297 'total_bytes': data_len
,
298 'tmpfilename': tmpfilename
,
299 'filename': filename
,
300 'status': 'downloading',
304 elif self
.params
.get('verbose', False):
305 if not cursor_in_new_line
:
307 cursor_in_new_line
= True
308 self
.to_screen(u
'[rtmpdump] '+line
)
310 if not cursor_in_new_line
:
312 return proc
.returncode
314 self
.report_destination(filename
)
315 tmpfilename
= self
.temp_name(filename
)
316 test
= self
.params
.get('test', False)
318 # Check for rtmpdump first
320 subprocess
.call(['rtmpdump', '-h'], stdout
=(open(os
.path
.devnull
, 'w')), stderr
=subprocess
.STDOUT
)
321 except (OSError, IOError):
322 self
.report_error(u
'RTMP download detected but "rtmpdump" could not be run')
325 # Download using rtmpdump. rtmpdump returns exit code 2 when
326 # the connection was interrumpted and resuming appears to be
327 # possible. This is part of rtmpdump's normal usage, AFAIK.
328 basic_args
= ['rtmpdump', '--verbose', '-r', url
, '-o', tmpfilename
]
329 if player_url
is not None:
330 basic_args
+= ['--swfVfy', player_url
]
331 if page_url
is not None:
332 basic_args
+= ['--pageUrl', page_url
]
333 if play_path
is not None:
334 basic_args
+= ['--playpath', play_path
]
335 if tc_url
is not None:
336 basic_args
+= ['--tcUrl', url
]
338 basic_args
+= ['--stop', '1']
340 basic_args
+= ['--live']
341 args
= basic_args
+ [[], ['--resume', '--skip', '1']][self
.params
.get('continuedl', False)]
343 if sys
.platform
== 'win32' and sys
.version_info
< (3, 0):
344 # Windows subprocess module does not actually support Unicode
346 # See http://stackoverflow.com/a/9951851/35070
347 subprocess_encoding
= sys
.getfilesystemencoding()
348 args
= [a
.encode(subprocess_encoding
, 'ignore') for a
in args
]
350 subprocess_encoding
= None
352 if self
.params
.get('verbose', False):
353 if subprocess_encoding
:
355 a
.decode(subprocess_encoding
) if isinstance(a
, bytes) else a
361 shell_quote
= lambda args
: ' '.join(map(pipes
.quote
, str_args
))
364 self
.to_screen(u
'[debug] rtmpdump command line: ' + shell_quote(str_args
))
366 retval
= run_rtmpdump(args
)
368 while (retval
== 2 or retval
== 1) and not test
:
369 prevsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
370 self
.to_screen(u
'[rtmpdump] %s bytes' % prevsize
)
371 time
.sleep(5.0) # This seems to be needed
372 retval
= run_rtmpdump(basic_args
+ ['-e'] + [[], ['-k', '1']][retval
== 1])
373 cursize
= os
.path
.getsize(encodeFilename(tmpfilename
))
374 if prevsize
== cursize
and retval
== 1:
376 # Some rtmp streams seem abort after ~ 99.8%. Don't complain for those
377 if prevsize
== cursize
and retval
== 2 and cursize
> 1024:
378 self
.to_screen(u
'[rtmpdump] Could not download the whole video. This can happen for some advertisements.')
381 if retval
== 0 or (test
and retval
== 2):
382 fsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
383 self
.to_screen(u
'[rtmpdump] %s bytes' % fsize
)
384 self
.try_rename(tmpfilename
, filename
)
385 self
._hook
_progress
({
386 'downloaded_bytes': fsize
,
387 'total_bytes': fsize
,
388 'filename': filename
,
389 'status': 'finished',
393 self
.to_stderr(u
"\n")
394 self
.report_error(u
'rtmpdump exited with code %d' % retval
)
397 def _download_with_mplayer(self
, filename
, url
):
398 self
.report_destination(filename
)
399 tmpfilename
= self
.temp_name(filename
)
401 args
= ['mplayer', '-really-quiet', '-vo', 'null', '-vc', 'dummy', '-dumpstream', '-dumpfile', tmpfilename
, url
]
402 # Check for mplayer first
404 subprocess
.call(['mplayer', '-h'], stdout
=(open(os
.path
.devnull
, 'w')), stderr
=subprocess
.STDOUT
)
405 except (OSError, IOError):
406 self
.report_error(u
'MMS or RTSP download detected but "%s" could not be run' % args
[0] )
409 # Download using mplayer.
410 retval
= subprocess
.call(args
)
412 fsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
413 self
.to_screen(u
'\r[%s] %s bytes' % (args
[0], fsize
))
414 self
.try_rename(tmpfilename
, filename
)
415 self
._hook
_progress
({
416 'downloaded_bytes': fsize
,
417 'total_bytes': fsize
,
418 'filename': filename
,
419 'status': 'finished',
423 self
.to_stderr(u
"\n")
424 self
.report_error(u
'mplayer exited with code %d' % retval
)
427 def _download_m3u8_with_ffmpeg(self
, filename
, url
):
428 self
.report_destination(filename
)
429 tmpfilename
= self
.temp_name(filename
)
431 args
= ['-y', '-i', url
, '-f', 'mp4', '-c', 'copy',
432 '-bsf:a', 'aac_adtstoasc', tmpfilename
]
434 for program
in ['avconv', 'ffmpeg']:
436 subprocess
.call([program
, '-version'], stdout
=(open(os
.path
.devnull
, 'w')), stderr
=subprocess
.STDOUT
)
438 except (OSError, IOError):
441 self
.report_error(u
'm3u8 download detected but ffmpeg or avconv could not be found')
442 cmd
= [program
] + args
444 retval
= subprocess
.call(cmd
)
446 fsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
447 self
.to_screen(u
'\r[%s] %s bytes' % (args
[0], fsize
))
448 self
.try_rename(tmpfilename
, filename
)
449 self
._hook
_progress
({
450 'downloaded_bytes': fsize
,
451 'total_bytes': fsize
,
452 'filename': filename
,
453 'status': 'finished',
457 self
.to_stderr(u
"\n")
458 self
.report_error(u
'ffmpeg exited with code %d' % retval
)
462 def _do_download(self
, filename
, info_dict
):
463 url
= info_dict
['url']
465 # Check file already present
466 if self
.params
.get('continuedl', False) and os
.path
.isfile(encodeFilename(filename
)) and not self
.params
.get('nopart', False):
467 self
.report_file_already_downloaded(filename
)
468 self
._hook
_progress
({
469 'filename': filename
,
470 'status': 'finished',
471 'total_bytes': os
.path
.getsize(encodeFilename(filename
)),
475 # Attempt to download using rtmpdump
476 if url
.startswith('rtmp'):
477 return self
._download
_with
_rtmpdump
(filename
, url
,
478 info_dict
.get('player_url', None),
479 info_dict
.get('page_url', None),
480 info_dict
.get('play_path', None),
481 info_dict
.get('tc_url', None),
482 info_dict
.get('rtmp_live', False))
484 # Attempt to download using mplayer
485 if url
.startswith('mms') or url
.startswith('rtsp'):
486 return self
._download
_with
_mplayer
(filename
, url
)
488 # m3u8 manifest are downloaded with ffmpeg
489 if determine_ext(url
) == u
'm3u8':
490 return self
._download
_m
3u8_with
_ffmpeg
(filename
, url
)
492 tmpfilename
= self
.temp_name(filename
)
495 # Do not include the Accept-Encoding header
496 headers
= {'Youtubedl-no-compression': 'True'}
497 if 'user_agent' in info_dict
:
498 headers
['Youtubedl-user-agent'] = info_dict
['user_agent']
499 basic_request
= compat_urllib_request
.Request(url
, None, headers
)
500 request
= compat_urllib_request
.Request(url
, None, headers
)
502 if self
.params
.get('test', False):
503 request
.add_header('Range','bytes=0-10240')
505 # Establish possible resume length
506 if os
.path
.isfile(encodeFilename(tmpfilename
)):
507 resume_len
= os
.path
.getsize(encodeFilename(tmpfilename
))
513 if self
.params
.get('continuedl', False):
514 self
.report_resuming_byte(resume_len
)
515 request
.add_header('Range','bytes=%d-' % resume_len
)
521 retries
= self
.params
.get('retries', 0)
522 while count
<= retries
:
523 # Establish connection
525 if count
== 0 and 'urlhandle' in info_dict
:
526 data
= info_dict
['urlhandle']
527 data
= compat_urllib_request
.urlopen(request
)
529 except (compat_urllib_error
.HTTPError
, ) as err
:
530 if (err
.code
< 500 or err
.code
>= 600) and err
.code
!= 416:
531 # Unexpected HTTP error
533 elif err
.code
== 416:
534 # Unable to resume (requested range not satisfiable)
536 # Open the connection again without the range header
537 data
= compat_urllib_request
.urlopen(basic_request
)
538 content_length
= data
.info()['Content-Length']
539 except (compat_urllib_error
.HTTPError
, ) as err
:
540 if err
.code
< 500 or err
.code
>= 600:
543 # Examine the reported length
544 if (content_length
is not None and
545 (resume_len
- 100 < int(content_length
) < resume_len
+ 100)):
546 # The file had already been fully downloaded.
547 # Explanation to the above condition: in issue #175 it was revealed that
548 # YouTube sometimes adds or removes a few bytes from the end of the file,
549 # changing the file size slightly and causing problems for some users. So
550 # I decided to implement a suggested change and consider the file
551 # completely downloaded if the file size differs less than 100 bytes from
552 # the one in the hard drive.
553 self
.report_file_already_downloaded(filename
)
554 self
.try_rename(tmpfilename
, filename
)
555 self
._hook
_progress
({
556 'filename': filename
,
557 'status': 'finished',
561 # The length does not match, we start the download over
562 self
.report_unable_to_resume()
568 self
.report_retry(count
, retries
)
571 self
.report_error(u
'giving up after %s retries' % retries
)
574 data_len
= data
.info().get('Content-length', None)
575 if data_len
is not None:
576 data_len
= int(data_len
) + resume_len
577 min_data_len
= self
.params
.get("min_filesize", None)
578 max_data_len
= self
.params
.get("max_filesize", None)
579 if min_data_len
is not None and data_len
< min_data_len
:
580 self
.to_screen(u
'\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len
, min_data_len
))
582 if max_data_len
is not None and data_len
> max_data_len
:
583 self
.to_screen(u
'\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len
, max_data_len
))
586 data_len_str
= format_bytes(data_len
)
587 byte_counter
= 0 + resume_len
588 block_size
= self
.params
.get('buffersize', 1024)
593 data_block
= data
.read(block_size
)
595 if len(data_block
) == 0:
597 byte_counter
+= len(data_block
)
599 # Open file just in time
602 (stream
, tmpfilename
) = sanitize_open(tmpfilename
, open_mode
)
603 assert stream
is not None
604 filename
= self
.undo_temp_name(tmpfilename
)
605 self
.report_destination(filename
)
606 except (OSError, IOError) as err
:
607 self
.report_error(u
'unable to open for writing: %s' % str(err
))
610 stream
.write(data_block
)
611 except (IOError, OSError) as err
:
612 self
.to_stderr(u
"\n")
613 self
.report_error(u
'unable to write data: %s' % str(err
))
615 if not self
.params
.get('noresizebuffer', False):
616 block_size
= self
.best_block_size(after
- before
, len(data_block
))
619 speed
= self
.calc_speed(start
, time
.time(), byte_counter
- resume_len
)
623 percent
= self
.calc_percent(byte_counter
, data_len
)
624 eta
= self
.calc_eta(start
, time
.time(), data_len
- resume_len
, byte_counter
- resume_len
)
625 self
.report_progress(percent
, data_len_str
, speed
, eta
)
627 self
._hook
_progress
({
628 'downloaded_bytes': byte_counter
,
629 'total_bytes': data_len
,
630 'tmpfilename': tmpfilename
,
631 'filename': filename
,
632 'status': 'downloading',
638 self
.slow_down(start
, byte_counter
- resume_len
)
641 self
.to_stderr(u
"\n")
642 self
.report_error(u
'Did not get any data blocks')
645 self
.report_finish(data_len_str
, (time
.time() - start
))
646 if data_len
is not None and byte_counter
!= data_len
:
647 raise ContentTooShortError(byte_counter
, int(data_len
))
648 self
.try_rename(tmpfilename
, filename
)
650 # Update file modification time
651 if self
.params
.get('updatetime', True):
652 info_dict
['filetime'] = self
.try_utime(filename
, data
.info().get('last-modified', None))
654 self
._hook
_progress
({
655 'downloaded_bytes': byte_counter
,
656 'total_bytes': byte_counter
,
657 'filename': filename
,
658 'status': 'finished',
663 def _hook_progress(self
, status
):
664 for ph
in self
._progress
_hooks
:
667 def add_progress_hook(self
, ph
):
668 """ ph gets called on download progress, with a dictionary with the entries
669 * filename: The final filename
670 * status: One of "downloading" and "finished"
672 It can also have some of the following entries:
674 * downloaded_bytes: Bytes on disks
675 * total_bytes: Total bytes, None if unknown
676 * tmpfilename: The filename we're currently writing to
677 * eta: The estimated time in seconds, None if unknown
678 * speed: The download speed in bytes/second, None if unknown
680 Hooks are guaranteed to be called at least once (with status "finished")
681 if the download is successful.
683 self
._progress
_hooks
.append(ph
)