yt-dlc/youtube_dlc/postprocessor/sponskrub.py

87 lines
3.7 KiB
Python

from __future__ import unicode_literals
import os
import subprocess
from .common import PostProcessor
from ..compat import compat_shlex_split
from ..utils import (
check_executable,
encodeArgument,
shell_quote,
PostProcessingError,
)
class SponSkrubPP(PostProcessor):
_temp_ext = 'spons'
_def_args = []
_exe_name = 'sponskrub'
def __init__(self, downloader, path='', args=None, ignoreerror=False, cut=False, force=False):
PostProcessor.__init__(self, downloader)
self.force = force
self.cutout = cut
self.args = ['-chapter'] if not cut else []
self.args += self._def_args if args is None else compat_shlex_split(args)
self.path = self.get_exe(path)
if not ignoreerror and self.path is None:
if path:
raise PostProcessingError('sponskrub not found in "%s"' % path)
else:
raise PostProcessingError('sponskrub not found. Please install or provide the path using --sponskrub-path.')
def get_exe(self, path=''):
if not path or not check_executable(path, ['-h']):
path = os.path.join(path, self._exe_name)
if not check_executable(path, ['-h']):
return None
return path
def run(self, information):
if self.path is None:
return [], information
if information['extractor_key'].lower() != 'youtube':
self._downloader.to_screen('[sponskrub] Skipping sponskrub since it is not a YouTube video')
return [], information
if self.cutout and not self.force and not information.get('__real_download', False):
self._downloader.to_screen(
'[sponskrub] Skipping sponskrub since the video was already downloaded. '
'Use --sponskrub-force to run sponskrub anyway')
return [], information
self._downloader.to_screen('[sponskrub] Trying to %s sponsor sections' % ('remove' if self.cutout else 'mark'))
if self.cutout:
self._downloader.to_screen('WARNING: Cutting out sponsor segments will cause the subtitles to go out of sync.')
if not information.get('__real_download', False):
self._downloader.to_screen('WARNING: If sponskrub is run multiple times, unintended parts of the video could be cut out.')
filename = information['filepath']
temp_filename = filename + '.' + self._temp_ext + os.path.splitext(filename)[1]
if os.path.exists(temp_filename):
os.remove(temp_filename)
cmd = [self.path]
if self.args:
cmd += self.args
cmd += ['--', information['id'], filename, temp_filename]
cmd = [encodeArgument(i) for i in cmd]
if self._downloader.params.get('verbose', False):
self._downloader.to_screen('[debug] sponskrub command line: %s' % shell_quote(cmd))
p = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
stdout, stderr = p.communicate()
if p.returncode == 0:
os.remove(filename)
os.rename(temp_filename, filename)
self._downloader.to_screen('[sponskrub] Sponsor sections have been %s' % ('removed' if self.cutout else 'marked'))
elif p.returncode != 3: # error code 3 means there was no info about the video
stderr = stderr.decode('utf-8', 'replace')
msg = stderr.strip().split('\n')[-1]
raise PostProcessingError(msg if msg else 'sponskrub failed with error code %s!' % p.returncode)
else:
self._downloader.to_screen('[sponskrub] No segments in the SponsorBlock database')
return [], information