15 class FileDownloader(object):
16 """File Downloader class.
18 File downloader objects are the ones responsible of downloading the
19 actual video file and writing it to disk.
21 File downloaders accept a lot of parameters. In order not to saturate
22 the object constructor with arguments, it receives a dictionary of
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
46 def __init__(self
, ydl
, params
):
47 """Create a FileDownloader object with the given options."""
49 self
._progress
_hooks
= []
53 def format_bytes(bytes):
56 if type(bytes) is str:
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
)
67 def format_seconds(seconds
):
68 (mins
, secs
) = divmod(seconds
, 60)
69 (hours
, mins
) = divmod(mins
, 60)
73 return '%02d:%02d' % (mins
, secs
)
75 return '%02d:%02d:%02d' % (hours
, mins
, secs
)
78 def calc_percent(byte_counter
, data_len
):
81 return float(byte_counter
) / float(data_len
) * 100.0
84 def format_percent(percent
):
87 return '%6s' % ('%3.1f%%' % percent
)
90 def calc_eta(start
, now
, total
, current
):
94 if current
== 0 or dif
< 0.001: # One millisecond
96 rate
= float(current
) / dif
97 return int((float(total
) - float(current
)) / rate
)
103 return FileDownloader
.format_seconds(eta
)
106 def calc_speed(start
, now
, bytes):
108 if bytes == 0 or dif
< 0.001: # One millisecond
110 return float(bytes) / dif
113 def format_speed(speed
):
115 return '%10s' % '---b/s'
116 return '%10s' % ('%s/s' % FileDownloader
.format_bytes(speed
))
119 def best_block_size(elapsed_time
, bytes):
120 new_min
= max(bytes / 2.0, 1.0)
121 new_max
= min(max(bytes * 2.0, 1.0), 4194304) # Do not surpass 4 MB
122 if elapsed_time
< 0.001:
124 rate
= bytes / elapsed_time
132 def parse_bytes(bytestr
):
133 """Parse a string indicating a byte quantity into an integer."""
134 matchobj
= re
.match(r
'(?i)^(\d+(?:\.\d+)?)([kMGTPEZY]?)$', bytestr
)
137 number
= float(matchobj
.group(1))
138 multiplier
= 1024.0 ** 'bkmgtpezy'.index(matchobj
.group(2).lower())
139 return int(round(number
* multiplier
))
141 def to_screen(self
, *args
, **kargs
):
142 self
.ydl
.to_screen(*args
, **kargs
)
144 def to_stderr(self
, message
):
145 self
.ydl
.to_screen(message
)
147 def to_cons_title(self
, message
):
148 """Set console/terminal window title to message."""
149 if not self
.params
.get('consoletitle', False):
151 if os
.name
== 'nt' and ctypes
.windll
.kernel32
.GetConsoleWindow():
152 # c_wchar_p() might not be necessary if `message` is
153 # already of type unicode()
154 ctypes
.windll
.kernel32
.SetConsoleTitleW(ctypes
.c_wchar_p(message
))
155 elif 'TERM' in os
.environ
:
156 self
.to_screen('\033]0;%s\007' % message
, skip_eol
=True)
158 def trouble(self
, *args
, **kargs
):
159 self
.ydl
.trouble(*args
, **kargs
)
161 def report_warning(self
, *args
, **kargs
):
162 self
.ydl
.report_warning(*args
, **kargs
)
164 def report_error(self
, *args
, **kargs
):
165 self
.ydl
.report_error(*args
, **kargs
)
167 def slow_down(self
, start_time
, byte_counter
):
168 """Sleep if the download speed is over the rate limit."""
169 rate_limit
= self
.params
.get('ratelimit', None)
170 if rate_limit
is None or byte_counter
== 0:
173 elapsed
= now
- start_time
176 speed
= float(byte_counter
) / elapsed
177 if speed
> rate_limit
:
178 time
.sleep((byte_counter
- rate_limit
* (now
- start_time
)) / rate_limit
)
180 def temp_name(self
, filename
):
181 """Returns a temporary filename for the given filename."""
182 if self
.params
.get('nopart', False) or filename
== u
'-' or \
183 (os
.path
.exists(encodeFilename(filename
)) and not os
.path
.isfile(encodeFilename(filename
))):
185 return filename
+ u
'.part'
187 def undo_temp_name(self
, filename
):
188 if filename
.endswith(u
'.part'):
189 return filename
[:-len(u
'.part')]
192 def try_rename(self
, old_filename
, new_filename
):
194 if old_filename
== new_filename
:
196 os
.rename(encodeFilename(old_filename
), encodeFilename(new_filename
))
197 except (IOError, OSError) as err
:
198 self
.report_error(u
'unable to rename file')
200 def try_utime(self
, filename
, last_modified_hdr
):
201 """Try to set the last-modified time of the given file."""
202 if last_modified_hdr
is None:
204 if not os
.path
.isfile(encodeFilename(filename
)):
206 timestr
= last_modified_hdr
209 filetime
= timeconvert(timestr
)
212 # Ignore obviously invalid dates
216 os
.utime(filename
, (time
.time(), filetime
))
221 def report_destination(self
, filename
):
222 """Report destination filename."""
223 self
.to_screen(u
'[download] Destination: ' + filename
)
225 def report_progress(self
, percent
, data_len_str
, speed
, eta
):
226 """Report download progress."""
227 if self
.params
.get('noprogress', False):
229 clear_line
= (u
'\x1b[K' if sys
.stderr
.isatty() and os
.name
!= 'nt' else u
'')
230 eta_str
= self
.format_eta(eta
)
231 percent_str
= self
.format_percent(percent
)
232 speed_str
= self
.format_speed(speed
)
233 if self
.params
.get('progress_with_newline', False):
234 self
.to_screen(u
'[download] %s of %s at %s ETA %s' %
235 (percent_str
, data_len_str
, speed_str
, eta_str
))
237 self
.to_screen(u
'\r%s[download] %s of %s at %s ETA %s' %
238 (clear_line
, percent_str
, data_len_str
, speed_str
, eta_str
), skip_eol
=True)
239 self
.to_cons_title(u
'youtube-dl - %s of %s at %s ETA %s' %
240 (percent_str
.strip(), data_len_str
.strip(), speed_str
.strip(), eta_str
.strip()))
242 def report_resuming_byte(self
, resume_len
):
243 """Report attempt to resume at given byte."""
244 self
.to_screen(u
'[download] Resuming download at byte %s' % resume_len
)
246 def report_retry(self
, count
, retries
):
247 """Report retry in case of HTTP error 5xx"""
248 self
.to_screen(u
'[download] Got server HTTP error. Retrying (attempt %d of %d)...' % (count
, retries
))
250 def report_file_already_downloaded(self
, file_name
):
251 """Report file has already been fully downloaded."""
253 self
.to_screen(u
'[download] %s has already been downloaded' % file_name
)
254 except (UnicodeEncodeError) as err
:
255 self
.to_screen(u
'[download] The file has already been downloaded')
257 def report_unable_to_resume(self
):
258 """Report it was impossible to resume download."""
259 self
.to_screen(u
'[download] Unable to resume')
261 def report_finish(self
, data_len_str
, tot_time
):
262 """Report download finished."""
263 if self
.params
.get('noprogress', False):
264 self
.to_screen(u
'[download] Download completed')
266 clear_line
= (u
'\x1b[K' if sys
.stderr
.isatty() and os
.name
!= 'nt' else u
'')
267 self
.to_screen(u
'\r%s[download] 100%% of %s in %s' %
268 (clear_line
, data_len_str
, self
.format_seconds(tot_time
)))
270 def _download_with_rtmpdump(self
, filename
, url
, player_url
, page_url
, play_path
, tc_url
):
271 self
.report_destination(filename
)
272 tmpfilename
= self
.temp_name(filename
)
274 # Check for rtmpdump first
276 subprocess
.call(['rtmpdump', '-h'], stdout
=(open(os
.path
.devnull
, 'w')), stderr
=subprocess
.STDOUT
)
277 except (OSError, IOError):
278 self
.report_error(u
'RTMP download detected but "rtmpdump" could not be run')
280 verbosity_option
= '--verbose' if self
.params
.get('verbose', False) else '--quiet'
282 # Download using rtmpdump. rtmpdump returns exit code 2 when
283 # the connection was interrumpted and resuming appears to be
284 # possible. This is part of rtmpdump's normal usage, AFAIK.
285 basic_args
= ['rtmpdump', verbosity_option
, '-r', url
, '-o', tmpfilename
]
286 if player_url
is not None:
287 basic_args
+= ['--swfVfy', player_url
]
288 if page_url
is not None:
289 basic_args
+= ['--pageUrl', page_url
]
290 if play_path
is not None:
291 basic_args
+= ['--playpath', play_path
]
292 if tc_url
is not None:
293 basic_args
+= ['--tcUrl', url
]
294 if self
.params
.get('test', False):
295 basic_args
+= ['--stop', '1']
296 args
= basic_args
+ [[], ['--resume', '--skip', '1']][self
.params
.get('continuedl', False)]
297 if self
.params
.get('verbose', False):
300 shell_quote
= lambda args
: ' '.join(map(pipes
.quote
, args
))
303 self
.to_screen(u
'[debug] rtmpdump command line: ' + shell_quote(args
))
304 retval
= subprocess
.call(args
)
305 while retval
== 2 or retval
== 1:
306 prevsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
307 self
.to_screen(u
'\r[rtmpdump] %s bytes' % prevsize
, skip_eol
=True)
308 time
.sleep(5.0) # This seems to be needed
309 retval
= subprocess
.call(basic_args
+ ['-e'] + [[], ['-k', '1']][retval
== 1])
310 cursize
= os
.path
.getsize(encodeFilename(tmpfilename
))
311 if prevsize
== cursize
and retval
== 1:
313 # Some rtmp streams seem abort after ~ 99.8%. Don't complain for those
314 if prevsize
== cursize
and retval
== 2 and cursize
> 1024:
315 self
.to_screen(u
'\r[rtmpdump] Could not download the whole video. This can happen for some advertisements.')
319 fsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
320 self
.to_screen(u
'\r[rtmpdump] %s bytes' % fsize
)
321 self
.try_rename(tmpfilename
, filename
)
322 self
._hook
_progress
({
323 'downloaded_bytes': fsize
,
324 'total_bytes': fsize
,
325 'filename': filename
,
326 'status': 'finished',
330 self
.to_stderr(u
"\n")
331 self
.report_error(u
'rtmpdump exited with code %d' % retval
)
334 def _download_with_mplayer(self
, filename
, url
):
335 self
.report_destination(filename
)
336 tmpfilename
= self
.temp_name(filename
)
338 args
= ['mplayer', '-really-quiet', '-vo', 'null', '-vc', 'dummy', '-dumpstream', '-dumpfile', tmpfilename
, url
]
339 # Check for mplayer first
341 subprocess
.call(['mplayer', '-h'], stdout
=(open(os
.path
.devnull
, 'w')), stderr
=subprocess
.STDOUT
)
342 except (OSError, IOError):
343 self
.report_error(u
'MMS or RTSP download detected but "%s" could not be run' % args
[0] )
346 # Download using mplayer.
347 retval
= subprocess
.call(args
)
349 fsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
350 self
.to_screen(u
'\r[%s] %s bytes' % (args
[0], fsize
))
351 self
.try_rename(tmpfilename
, filename
)
352 self
._hook
_progress
({
353 'downloaded_bytes': fsize
,
354 'total_bytes': fsize
,
355 'filename': filename
,
356 'status': 'finished',
360 self
.to_stderr(u
"\n")
361 self
.report_error(u
'mplayer exited with code %d' % retval
)
364 def _download_m3u8_with_ffmpeg(self
, filename
, url
):
365 self
.report_destination(filename
)
366 tmpfilename
= self
.temp_name(filename
)
368 args
= ['ffmpeg', '-y', '-i', url
, '-f', 'mp4', tmpfilename
]
369 # Check for ffmpeg first
371 subprocess
.call(['ffmpeg', '-h'], stdout
=(open(os
.path
.devnull
, 'w')), stderr
=subprocess
.STDOUT
)
372 except (OSError, IOError):
373 self
.report_error(u
'm3u8 download detected but "%s" could not be run' % args
[0] )
376 retval
= subprocess
.call(args
)
378 fsize
= os
.path
.getsize(encodeFilename(tmpfilename
))
379 self
.to_screen(u
'\r[%s] %s bytes' % (args
[0], fsize
))
380 self
.try_rename(tmpfilename
, filename
)
381 self
._hook
_progress
({
382 'downloaded_bytes': fsize
,
383 'total_bytes': fsize
,
384 'filename': filename
,
385 'status': 'finished',
389 self
.to_stderr(u
"\n")
390 self
.report_error(u
'ffmpeg exited with code %d' % retval
)
394 def _do_download(self
, filename
, info_dict
):
395 url
= info_dict
['url']
397 # Check file already present
398 if self
.params
.get('continuedl', False) and os
.path
.isfile(encodeFilename(filename
)) and not self
.params
.get('nopart', False):
399 self
.report_file_already_downloaded(filename
)
400 self
._hook
_progress
({
401 'filename': filename
,
402 'status': 'finished',
403 'total_bytes': os
.path
.getsize(encodeFilename(filename
)),
407 # Attempt to download using rtmpdump
408 if url
.startswith('rtmp'):
409 return self
._download
_with
_rtmpdump
(filename
, url
,
410 info_dict
.get('player_url', None),
411 info_dict
.get('page_url', None),
412 info_dict
.get('play_path', None),
413 info_dict
.get('tc_url', None))
415 # Attempt to download using mplayer
416 if url
.startswith('mms') or url
.startswith('rtsp'):
417 return self
._download
_with
_mplayer
(filename
, url
)
419 # m3u8 manifest are downloaded with ffmpeg
420 if determine_ext(url
) == u
'm3u8':
421 return self
._download
_m
3u8_with
_ffmpeg
(filename
, url
)
423 tmpfilename
= self
.temp_name(filename
)
426 # Do not include the Accept-Encoding header
427 headers
= {'Youtubedl-no-compression': 'True'}
428 if 'user_agent' in info_dict
:
429 headers
['Youtubedl-user-agent'] = info_dict
['user_agent']
430 basic_request
= compat_urllib_request
.Request(url
, None, headers
)
431 request
= compat_urllib_request
.Request(url
, None, headers
)
433 if self
.params
.get('test', False):
434 request
.add_header('Range','bytes=0-10240')
436 # Establish possible resume length
437 if os
.path
.isfile(encodeFilename(tmpfilename
)):
438 resume_len
= os
.path
.getsize(encodeFilename(tmpfilename
))
444 if self
.params
.get('continuedl', False):
445 self
.report_resuming_byte(resume_len
)
446 request
.add_header('Range','bytes=%d-' % resume_len
)
452 retries
= self
.params
.get('retries', 0)
453 while count
<= retries
:
454 # Establish connection
456 if count
== 0 and 'urlhandle' in info_dict
:
457 data
= info_dict
['urlhandle']
458 data
= compat_urllib_request
.urlopen(request
)
460 except (compat_urllib_error
.HTTPError
, ) as err
:
461 if (err
.code
< 500 or err
.code
>= 600) and err
.code
!= 416:
462 # Unexpected HTTP error
464 elif err
.code
== 416:
465 # Unable to resume (requested range not satisfiable)
467 # Open the connection again without the range header
468 data
= compat_urllib_request
.urlopen(basic_request
)
469 content_length
= data
.info()['Content-Length']
470 except (compat_urllib_error
.HTTPError
, ) as err
:
471 if err
.code
< 500 or err
.code
>= 600:
474 # Examine the reported length
475 if (content_length
is not None and
476 (resume_len
- 100 < int(content_length
) < resume_len
+ 100)):
477 # The file had already been fully downloaded.
478 # Explanation to the above condition: in issue #175 it was revealed that
479 # YouTube sometimes adds or removes a few bytes from the end of the file,
480 # changing the file size slightly and causing problems for some users. So
481 # I decided to implement a suggested change and consider the file
482 # completely downloaded if the file size differs less than 100 bytes from
483 # the one in the hard drive.
484 self
.report_file_already_downloaded(filename
)
485 self
.try_rename(tmpfilename
, filename
)
486 self
._hook
_progress
({
487 'filename': filename
,
488 'status': 'finished',
492 # The length does not match, we start the download over
493 self
.report_unable_to_resume()
499 self
.report_retry(count
, retries
)
502 self
.report_error(u
'giving up after %s retries' % retries
)
505 data_len
= data
.info().get('Content-length', None)
506 if data_len
is not None:
507 data_len
= int(data_len
) + resume_len
508 min_data_len
= self
.params
.get("min_filesize", None)
509 max_data_len
= self
.params
.get("max_filesize", None)
510 if min_data_len
is not None and data_len
< min_data_len
:
511 self
.to_screen(u
'\r[download] File is smaller than min-filesize (%s bytes < %s bytes). Aborting.' % (data_len
, min_data_len
))
513 if max_data_len
is not None and data_len
> max_data_len
:
514 self
.to_screen(u
'\r[download] File is larger than max-filesize (%s bytes > %s bytes). Aborting.' % (data_len
, max_data_len
))
517 data_len_str
= self
.format_bytes(data_len
)
518 byte_counter
= 0 + resume_len
519 block_size
= self
.params
.get('buffersize', 1024)
524 data_block
= data
.read(block_size
)
526 if len(data_block
) == 0:
528 byte_counter
+= len(data_block
)
530 # Open file just in time
533 (stream
, tmpfilename
) = sanitize_open(tmpfilename
, open_mode
)
534 assert stream
is not None
535 filename
= self
.undo_temp_name(tmpfilename
)
536 self
.report_destination(filename
)
537 except (OSError, IOError) as err
:
538 self
.report_error(u
'unable to open for writing: %s' % str(err
))
541 stream
.write(data_block
)
542 except (IOError, OSError) as err
:
543 self
.to_stderr(u
"\n")
544 self
.report_error(u
'unable to write data: %s' % str(err
))
546 if not self
.params
.get('noresizebuffer', False):
547 block_size
= self
.best_block_size(after
- before
, len(data_block
))
550 speed
= self
.calc_speed(start
, time
.time(), byte_counter
- resume_len
)
552 self
.report_progress('Unknown %', data_len_str
, speed_str
, 'Unknown ETA')
555 percent
= self
.calc_percent(byte_counter
, data_len
)
556 eta
= self
.calc_eta(start
, time
.time(), data_len
- resume_len
, byte_counter
- resume_len
)
557 self
.report_progress(percent
, data_len_str
, speed
, eta
)
559 self
._hook
_progress
({
560 'downloaded_bytes': byte_counter
,
561 'total_bytes': data_len
,
562 'tmpfilename': tmpfilename
,
563 'filename': filename
,
564 'status': 'downloading',
570 self
.slow_down(start
, byte_counter
- resume_len
)
573 self
.to_stderr(u
"\n")
574 self
.report_error(u
'Did not get any data blocks')
577 self
.report_finish(data_len_str
, (time
.time() - start
))
578 if data_len
is not None and byte_counter
!= data_len
:
579 raise ContentTooShortError(byte_counter
, int(data_len
))
580 self
.try_rename(tmpfilename
, filename
)
582 # Update file modification time
583 if self
.params
.get('updatetime', True):
584 info_dict
['filetime'] = self
.try_utime(filename
, data
.info().get('last-modified', None))
586 self
._hook
_progress
({
587 'downloaded_bytes': byte_counter
,
588 'total_bytes': byte_counter
,
589 'filename': filename
,
590 'status': 'finished',
595 def _hook_progress(self
, status
):
596 for ph
in self
._progress
_hooks
:
599 def add_progress_hook(self
, ph
):
600 """ ph gets called on download progress, with a dictionary with the entries
601 * filename: The final filename
602 * status: One of "downloading" and "finished"
604 It can also have some of the following entries:
606 * downloaded_bytes: Bytes on disks
607 * total_bytes: Total bytes, None if unknown
608 * tmpfilename: The filename we're currently writing to
609 * eta: The estimated time in seconds, None if unknown
610 * speed: The download speed in bytes/second, None if unknown
612 Hooks are guaranteed to be called at least once (with status "finished")
613 if the download is successful.
615 self
._progress
_hooks
.append(ph
)