yt-dlp/yt_dlp/downloader/external.py

543 lines
22 KiB
Python
Raw Normal View History

import os.path
import re
import subprocess
import sys
import time
from .fragment import FragmentFD
from ..compat import functools
from ..compat import compat_setenv, compat_str
from ..postprocessor.ffmpeg import EXT_TO_OUT_FORMATS, FFmpegPostProcessor
from ..utils import (
Popen,
_configuration_args,
check_executable,
classproperty,
cli_bool_option,
cli_option,
cli_valueless_option,
determine_ext,
encodeArgument,
encodeFilename,
handle_youtubedl_headers,
remove_end,
traverse_obj,
)
class ExternalFD(FragmentFD):
SUPPORTED_PROTOCOLS = ('http', 'https', 'ftp', 'ftps')
can_download_to_stdout = False
def real_download(self, filename, info_dict):
self.report_destination(filename)
tmpfilename = self.temp_name(filename)
try:
started = time.time()
retval = self._call_downloader(tmpfilename, info_dict)
except KeyboardInterrupt:
if not info_dict.get('is_live'):
raise
# Live stream downloading cancellation should be considered as
# correct and expected termination thus all postprocessing
# should take place
retval = 0
self.to_screen('[%s] Interrupted by user' % self.get_basename())
if retval == 0:
status = {
'filename': filename,
'status': 'finished',
'elapsed': time.time() - started,
}
if filename != '-':
fsize = os.path.getsize(encodeFilename(tmpfilename))
self.to_screen(f'\r[{self.get_basename()}] Downloaded {fsize} bytes')
self.try_rename(tmpfilename, filename)
status.update({
'downloaded_bytes': fsize,
'total_bytes': fsize,
})
self._hook_progress(status, info_dict)
return True
else:
self.to_stderr('\n')
self.report_error('%s exited with code %d' % (
self.get_basename(), retval))
return False
@classmethod
def get_basename(cls):
return cls.__name__[:-2].lower()
@classproperty
def EXE_NAME(cls):
return cls.get_basename()
@functools.cached_property
def exe(self):
return self.EXE_NAME
@classmethod
def available(cls, path=None):
path = check_executable(
cls.EXE_NAME if path in (None, cls.get_basename()) else path,
[cls.AVAILABLE_OPT])
if not path:
return False
cls.exe = path
return path
@classmethod
def supports(cls, info_dict):
return (
(cls.can_download_to_stdout or not info_dict.get('to_stdout'))
and info_dict['protocol'] in cls.SUPPORTED_PROTOCOLS)
@classmethod
def can_download(cls, info_dict, path=None):
return cls.available(path) and cls.supports(info_dict)
def _option(self, command_option, param):
return cli_option(self.params, command_option, param)
def _bool_option(self, command_option, param, true_value='true', false_value='false', separator=None):
return cli_bool_option(self.params, command_option, param, true_value, false_value, separator)
def _valueless_option(self, command_option, param, expected_value=True):
return cli_valueless_option(self.params, command_option, param, expected_value)
def _configuration_args(self, keys=None, *args, **kwargs):
return _configuration_args(
self.get_basename(), self.params.get('external_downloader_args'), self.EXE_NAME,
keys, *args, **kwargs)
def _call_downloader(self, tmpfilename, info_dict):
""" Either overwrite this or implement _make_cmd """
cmd = [encodeArgument(a) for a in self._make_cmd(tmpfilename, info_dict)]
self._debug_cmd(cmd)
if 'fragments' not in info_dict:
p = Popen(cmd, stderr=subprocess.PIPE)
_, stderr = p.communicate_or_kill()
if p.returncode != 0:
self.to_stderr(stderr.decode('utf-8', 'replace'))
return p.returncode
fragment_retries = self.params.get('fragment_retries', 0)
skip_unavailable_fragments = self.params.get('skip_unavailable_fragments', True)
count = 0
while count <= fragment_retries:
p = Popen(cmd, stderr=subprocess.PIPE)
_, stderr = p.communicate_or_kill()
if p.returncode == 0:
break
# TODO: Decide whether to retry based on error code
# https://aria2.github.io/manual/en/html/aria2c.html#exit-status
self.to_stderr(stderr.decode('utf-8', 'replace'))
count += 1
if count <= fragment_retries:
self.to_screen(
'[%s] Got error. Retrying fragments (attempt %d of %s)...'
% (self.get_basename(), count, self.format_retries(fragment_retries)))
self.sleep_retry('fragment', count)
if count > fragment_retries:
if not skip_unavailable_fragments:
self.report_error('Giving up after %s fragment retries' % fragment_retries)
return -1
decrypt_fragment = self.decrypter(info_dict)
dest, _ = self.sanitize_open(tmpfilename, 'wb')
for frag_index, fragment in enumerate(info_dict['fragments']):
fragment_filename = '%s-Frag%d' % (tmpfilename, frag_index)
try:
src, _ = self.sanitize_open(fragment_filename, 'rb')
except OSError as err:
if skip_unavailable_fragments and frag_index > 1:
self.report_skip_fragment(frag_index, err)
continue
self.report_error(f'Unable to open fragment {frag_index}; {err}')
return -1
dest.write(decrypt_fragment(fragment, src.read()))
src.close()
if not self.params.get('keep_fragments', False):
self.try_remove(encodeFilename(fragment_filename))
dest.close()
self.try_remove(encodeFilename('%s.frag.urls' % tmpfilename))
return 0
class CurlFD(ExternalFD):
AVAILABLE_OPT = '-V'
def _make_cmd(self, tmpfilename, info_dict):
cmd = [self.exe, '--location', '-o', tmpfilename, '--compressed']
if info_dict.get('http_headers') is not None:
for key, val in info_dict['http_headers'].items():
cmd += ['--header', f'{key}: {val}']
cmd += self._bool_option('--continue-at', 'continuedl', '-', '0')
cmd += self._valueless_option('--silent', 'noprogress')
cmd += self._valueless_option('--verbose', 'verbose')
cmd += self._option('--limit-rate', 'ratelimit')
retry = self._option('--retry', 'retries')
if len(retry) == 2:
if retry[1] in ('inf', 'infinite'):
retry[1] = '2147483647'
cmd += retry
cmd += self._option('--max-filesize', 'max_filesize')
cmd += self._option('--interface', 'source_address')
cmd += self._option('--proxy', 'proxy')
cmd += self._valueless_option('--insecure', 'nocheckcertificate')
cmd += self._configuration_args()
cmd += ['--', info_dict['url']]
return cmd
def _call_downloader(self, tmpfilename, info_dict):
cmd = [encodeArgument(a) for a in self._make_cmd(tmpfilename, info_dict)]
self._debug_cmd(cmd)
# curl writes the progress to stderr so don't capture it.
p = Popen(cmd)
p.communicate_or_kill()
return p.returncode
class AxelFD(ExternalFD):
AVAILABLE_OPT = '-V'
def _make_cmd(self, tmpfilename, info_dict):
cmd = [self.exe, '-o', tmpfilename]
if info_dict.get('http_headers') is not None:
for key, val in info_dict['http_headers'].items():
cmd += ['-H', f'{key}: {val}']
cmd += self._configuration_args()
cmd += ['--', info_dict['url']]
return cmd
class WgetFD(ExternalFD):
AVAILABLE_OPT = '--version'
def _make_cmd(self, tmpfilename, info_dict):
cmd = [self.exe, '-O', tmpfilename, '-nv', '--no-cookies', '--compression=auto']
if info_dict.get('http_headers') is not None:
for key, val in info_dict['http_headers'].items():
cmd += ['--header', f'{key}: {val}']
cmd += self._option('--limit-rate', 'ratelimit')
retry = self._option('--tries', 'retries')
if len(retry) == 2:
if retry[1] in ('inf', 'infinite'):
retry[1] = '0'
cmd += retry
cmd += self._option('--bind-address', 'source_address')
proxy = self.params.get('proxy')
if proxy:
for var in ('http_proxy', 'https_proxy'):
cmd += ['--execute', f'{var}={proxy}']
cmd += self._valueless_option('--no-check-certificate', 'nocheckcertificate')
cmd += self._configuration_args()
cmd += ['--', info_dict['url']]
return cmd
class Aria2cFD(ExternalFD):
AVAILABLE_OPT = '-v'
SUPPORTED_PROTOCOLS = ('http', 'https', 'ftp', 'ftps', 'dash_frag_urls', 'm3u8_frag_urls')
@staticmethod
def supports_manifest(manifest):
UNSUPPORTED_FEATURES = [
r'#EXT-X-BYTERANGE', # playlists composed of byte ranges of media files [1]
# 1. https://tools.ietf.org/html/draft-pantos-http-live-streaming-17#section-4.3.2.2
]
check_results = (not re.search(feature, manifest) for feature in UNSUPPORTED_FEATURES)
return all(check_results)
def _make_cmd(self, tmpfilename, info_dict):
2021-03-20 03:50:24 +00:00
cmd = [self.exe, '-c',
'--console-log-level=warn', '--summary-interval=0', '--download-result=hide',
'--http-accept-gzip=true', '--file-allocation=none', '-x16', '-j16', '-s16']
2021-03-20 03:50:24 +00:00
if 'fragments' in info_dict:
cmd += ['--allow-overwrite=true', '--allow-piece-length-change=true']
else:
cmd += ['--min-split-size', '1M']
2021-03-20 03:50:24 +00:00
if info_dict.get('http_headers') is not None:
for key, val in info_dict['http_headers'].items():
cmd += ['--header', f'{key}: {val}']
2021-08-26 19:29:36 +00:00
cmd += self._option('--max-overall-download-limit', 'ratelimit')
cmd += self._option('--interface', 'source_address')
cmd += self._option('--all-proxy', 'proxy')
cmd += self._bool_option('--check-certificate', 'nocheckcertificate', 'false', 'true', '=')
cmd += self._bool_option('--remote-time', 'updatetime', 'true', 'false', '=')
cmd += self._bool_option('--show-console-readout', 'noprogress', 'false', 'true', '=')
2021-03-20 03:50:24 +00:00
cmd += self._configuration_args()
# aria2c strips out spaces from the beginning/end of filenames and paths.
# We work around this issue by adding a "./" to the beginning of the
# filename and relative path, and adding a "/" at the end of the path.
# See: https://github.com/yt-dlp/yt-dlp/issues/276
# https://github.com/ytdl-org/youtube-dl/issues/20312
# https://github.com/aria2/aria2/issues/1373
2021-03-20 03:50:24 +00:00
dn = os.path.dirname(tmpfilename)
if dn:
if not os.path.isabs(dn):
dn = f'.{os.path.sep}{dn}'
cmd += ['--dir', dn + os.path.sep]
2021-03-20 03:50:24 +00:00
if 'fragments' not in info_dict:
cmd += ['--out', f'.{os.path.sep}{os.path.basename(tmpfilename)}']
cmd += ['--auto-file-renaming=false']
2021-03-20 03:50:24 +00:00
if 'fragments' in info_dict:
cmd += ['--file-allocation=none', '--uri-selector=inorder']
url_list_file = '%s.frag.urls' % tmpfilename
url_list = []
for frag_index, fragment in enumerate(info_dict['fragments']):
fragment_filename = '%s-Frag%d' % (os.path.basename(tmpfilename), frag_index)
url_list.append('%s\n\tout=%s' % (fragment['url'], fragment_filename))
stream, _ = self.sanitize_open(url_list_file, 'wb')
stream.write('\n'.join(url_list).encode())
stream.close()
cmd += ['-i', url_list_file]
else:
cmd += ['--', info_dict['url']]
return cmd
class HttpieFD(ExternalFD):
AVAILABLE_OPT = '--version'
EXE_NAME = 'http'
def _make_cmd(self, tmpfilename, info_dict):
cmd = ['http', '--download', '--output', tmpfilename, info_dict['url']]
if info_dict.get('http_headers') is not None:
for key, val in info_dict['http_headers'].items():
cmd += [f'{key}:{val}']
return cmd
class FFmpegFD(ExternalFD):
SUPPORTED_PROTOCOLS = ('http', 'https', 'ftp', 'ftps', 'm3u8', 'm3u8_native', 'rtsp', 'rtmp', 'rtmp_ffmpeg', 'mms', 'http_dash_segments')
can_download_to_stdout = True
@classmethod
def available(cls, path=None):
# TODO: Fix path for ffmpeg
2021-07-31 10:51:01 +00:00
# Fixme: This may be wrong when --ffmpeg-location is used
return FFmpegPostProcessor().available
@classmethod
def supports(cls, info_dict):
return all(proto in cls.SUPPORTED_PROTOCOLS for proto in info_dict['protocol'].split('+'))
def on_process_started(self, proc, stdin):
""" Override this in subclasses """
pass
2021-07-31 10:51:01 +00:00
@classmethod
2021-09-11 12:17:26 +00:00
def can_merge_formats(cls, info_dict, params):
2021-07-31 10:51:01 +00:00
return (
info_dict.get('requested_formats')
and info_dict.get('protocol')
and not params.get('allow_unplayable_formats')
and 'no-direct-merge' not in params.get('compat_opts', [])
and cls.can_download(info_dict))
def _call_downloader(self, tmpfilename, info_dict):
urls = [f['url'] for f in info_dict.get('requested_formats', [])] or [info_dict['url']]
ffpp = FFmpegPostProcessor(downloader=self)
if not ffpp.available:
2021-01-30 11:13:54 +00:00
self.report_error('m3u8 download detected but ffmpeg could not be found. Please install')
return False
ffpp.check_version()
args = [ffpp.executable, '-y']
for log_level in ('quiet', 'verbose'):
if self.params.get(log_level, False):
args += ['-loglevel', log_level]
break
if not self.params.get('verbose'):
args += ['-hide_banner']
args += traverse_obj(info_dict, ('downloader_options', 'ffmpeg_args'), default=[])
# These exists only for compatibility. Extractors should use
# info_dict['downloader_options']['ffmpeg_args'] instead
args += info_dict.get('_ffmpeg_args') or []
seekable = info_dict.get('_seekable')
if seekable is not None:
# setting -seekable prevents ffmpeg from guessing if the server
# supports seeking(by adding the header `Range: bytes=0-`), which
# can cause problems in some cases
2019-03-09 12:14:41 +00:00
# https://github.com/ytdl-org/youtube-dl/issues/11800#issuecomment-275037127
# http://trac.ffmpeg.org/ticket/6125#comment:10
args += ['-seekable', '1' if seekable else '0']
# start_time = info_dict.get('start_time') or 0
# if start_time:
# args += ['-ss', compat_str(start_time)]
# end_time = info_dict.get('end_time')
# if end_time:
# args += ['-t', compat_str(end_time - start_time)]
http_headers = None
if info_dict.get('http_headers'):
youtubedl_headers = handle_youtubedl_headers(info_dict['http_headers'])
http_headers = [
# Trailing \r\n after each HTTP header is important to prevent warning from ffmpeg/avconv:
# [http @ 00000000003d2fa0] No trailing CRLF found in HTTP header.
'-headers',
''.join(f'{key}: {val}\r\n' for key, val in youtubedl_headers.items())
]
env = None
proxy = self.params.get('proxy')
if proxy:
if not re.match(r'^[\da-zA-Z]+://', proxy):
proxy = 'http://%s' % proxy
if proxy.startswith('socks'):
self.report_warning(
'%s does not support SOCKS proxies. Downloading is likely to fail. '
'Consider adding --hls-prefer-native to your command.' % self.get_basename())
# Since December 2015 ffmpeg supports -http_proxy option (see
# http://git.videolan.org/?p=ffmpeg.git;a=commit;h=b4eb1f29ebddd60c41a2eb39f5af701e38e0d3fd)
# We could switch to the following code if we are able to detect version properly
# args += ['-http_proxy', proxy]
env = os.environ.copy()
compat_setenv('HTTP_PROXY', proxy, env=env)
compat_setenv('http_proxy', proxy, env=env)
protocol = info_dict.get('protocol')
if protocol == 'rtmp':
player_url = info_dict.get('player_url')
page_url = info_dict.get('page_url')
app = info_dict.get('app')
play_path = info_dict.get('play_path')
tc_url = info_dict.get('tc_url')
flash_version = info_dict.get('flash_version')
live = info_dict.get('rtmp_live', False)
conn = info_dict.get('rtmp_conn')
if player_url is not None:
args += ['-rtmp_swfverify', player_url]
if page_url is not None:
args += ['-rtmp_pageurl', page_url]
if app is not None:
args += ['-rtmp_app', app]
if play_path is not None:
args += ['-rtmp_playpath', play_path]
if tc_url is not None:
args += ['-rtmp_tcurl', tc_url]
if flash_version is not None:
args += ['-rtmp_flashver', flash_version]
if live:
args += ['-rtmp_live', 'live']
if isinstance(conn, list):
for entry in conn:
args += ['-rtmp_conn', entry]
elif isinstance(conn, compat_str):
args += ['-rtmp_conn', conn]
for i, url in enumerate(urls):
# We need to specify headers for each http input stream
# otherwise, it will only be applied to the first.
# https://github.com/yt-dlp/yt-dlp/issues/2696
if http_headers is not None and re.match(r'^https?://', url):
args += http_headers
args += self._configuration_args((f'_i{i + 1}', '_i')) + ['-i', url]
args += ['-c', 'copy']
if info_dict.get('requested_formats') or protocol == 'http_dash_segments':
for (i, fmt) in enumerate(info_dict.get('requested_formats') or [info_dict]):
stream_number = fmt.get('manifest_stream_number', 0)
args.extend(['-map', f'{i}:{stream_number}'])
if self.params.get('test', False):
args += ['-fs', compat_str(self._TEST_FILE_SIZE)]
ext = info_dict['ext']
if protocol in ('m3u8', 'm3u8_native'):
use_mpegts = (tmpfilename == '-') or self.params.get('hls_use_mpegts')
if use_mpegts is None:
use_mpegts = info_dict.get('is_live')
if use_mpegts:
args += ['-f', 'mpegts']
else:
args += ['-f', 'mp4']
if (ffpp.basename == 'ffmpeg' and ffpp._features.get('needs_adtstoasc')) and (not info_dict.get('acodec') or info_dict['acodec'].split('.')[0] in ('aac', 'mp4a')):
args += ['-bsf:a', 'aac_adtstoasc']
elif protocol == 'rtmp':
args += ['-f', 'flv']
elif ext == 'mp4' and tmpfilename == '-':
args += ['-f', 'mpegts']
elif ext == 'unknown_video':
ext = determine_ext(remove_end(tmpfilename, '.part'))
if ext == 'unknown_video':
self.report_warning(
'The video format is unknown and cannot be downloaded by ffmpeg. '
'Explicitly set the extension in the filename to attempt download in that format')
else:
self.report_warning(f'The video format is unknown. Trying to download as {ext} according to the filename')
args += ['-f', EXT_TO_OUT_FORMATS.get(ext, ext)]
else:
args += ['-f', EXT_TO_OUT_FORMATS.get(ext, ext)]
args += self._configuration_args(('_o1', '_o', ''))
args = [encodeArgument(opt) for opt in args]
args.append(encodeFilename(ffpp._ffmpeg_filename_argument(tmpfilename), True))
self._debug_cmd(args)
proc = Popen(args, stdin=subprocess.PIPE, env=env)
if url in ('-', 'pipe:'):
self.on_process_started(proc, proc.stdin)
try:
retval = proc.wait()
except BaseException as e:
# subprocces.run would send the SIGKILL signal to ffmpeg and the
# mp4 file couldn't be played, but if we ask ffmpeg to quit it
# produces a file that is playable (this is mostly useful for live
# streams). Note that Windows is not affected and produces playable
2019-03-09 12:14:41 +00:00
# files (see https://github.com/ytdl-org/youtube-dl/issues/8300).
if isinstance(e, KeyboardInterrupt) and sys.platform != 'win32' and url not in ('-', 'pipe:'):
proc.communicate_or_kill(b'q')
else:
proc.kill()
proc.wait()
raise
return retval
class AVconvFD(FFmpegFD):
pass
_BY_NAME = {
klass.get_basename(): klass
for name, klass in globals().items()
if name.endswith('FD') and name not in ('ExternalFD', 'FragmentFD')
}
_BY_EXE = {klass.EXE_NAME: klass for klass in _BY_NAME.values()}
def list_external_downloaders():
return sorted(_BY_NAME.keys())
def get_external_downloader(external_downloader):
""" Given the name of the executable, see whether we support the given
downloader . """
# Drop .exe extension on Windows
bn = os.path.splitext(os.path.basename(external_downloader))[0]
return _BY_NAME.get(bn, _BY_EXE.get(bn))