mirror of
https://github.com/moparisthebest/SickRage
synced 2024-11-14 05:15:12 -05:00
c945726f05
Added ignore/required words option to bet set individually for each show. Fixed issue with global ignore words not properly matching against releases. Fixed issue with
125 lines
7.6 KiB
Python
125 lines
7.6 KiB
Python
# -*- coding: utf-8 -*-
|
|
# Copyright 2011-2012 Antoine Bertin <diaoulael@gmail.com>
|
|
#
|
|
# This file is part of subliminal.
|
|
#
|
|
# subliminal is free software; you can redistribute it and/or modify it under
|
|
# the terms of the GNU Lesser General Public License as published by
|
|
# the Free Software Foundation; either version 3 of the License, or
|
|
# (at your option) any later version.
|
|
#
|
|
# subliminal is distributed in the hope that it will be useful,
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
# GNU Lesser General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU Lesser General Public License
|
|
# along with subliminal. If not, see <http://www.gnu.org/licenses/>.
|
|
from . import ServiceBase
|
|
from ..exceptions import DownloadFailedError
|
|
from ..language import Language, language_set
|
|
from ..subtitles import ResultSubtitle
|
|
from ..utils import get_keywords
|
|
from ..videos import Episode, Movie
|
|
from bs4 import BeautifulSoup
|
|
import guessit
|
|
import logging
|
|
import re
|
|
from subliminal.subtitles import get_subtitle_path
|
|
|
|
|
|
logger = logging.getLogger("subliminal")
|
|
|
|
|
|
class PodnapisiWeb(ServiceBase):
|
|
server_url = 'http://simple.podnapisi.net'
|
|
site_url = 'http://www.podnapisi.net'
|
|
api_based = True
|
|
user_agent = 'Subliminal/0.6'
|
|
videos = [Episode, Movie]
|
|
require_video = False
|
|
required_features = ['xml']
|
|
languages = language_set(['Albanian', 'Arabic', 'Spanish (Argentina)', 'Belarusian', 'Bosnian', 'Portuguese (Brazil)', 'Bulgarian', 'Catalan',
|
|
'Chinese', 'Croatian', 'Czech', 'Danish', 'Dutch', 'English', 'Estonian', 'Persian',
|
|
'Finnish', 'French', 'German', 'gre', 'Kalaallisut', 'Hebrew', 'Hindi', 'Hungarian',
|
|
'Icelandic', 'Indonesian', 'Irish', 'Italian', 'Japanese', 'Kazakh', 'Korean', 'Latvian',
|
|
'Lithuanian', 'Macedonian', 'Malay', 'Norwegian', 'Polish', 'Portuguese', 'Romanian',
|
|
'Russian', 'Serbian', 'Sinhala', 'Slovak', 'Slovenian', 'Spanish', 'Swedish', 'Thai',
|
|
'Turkish', 'Ukrainian', 'Vietnamese'])
|
|
language_map = {Language('Albanian'): 29, Language('Arabic'): 12, Language('Spanish (Argentina)'): 14, Language('Belarusian'): 50,
|
|
Language('Bosnian'): 10, Language('Portuguese (Brazil)'): 48, Language('Bulgarian'): 33, Language('Catalan'): 53,
|
|
Language('Chinese'): 17, Language('Croatian'): 38, Language('Czech'): 7, Language('Danish'): 24,
|
|
Language('Dutch'): 23, Language('English'): 2, Language('Estonian'): 20, Language('Persian'): 52,
|
|
Language('Finnish'): 31, Language('French'): 8, Language('German'): 5, Language('gre'): 16,
|
|
Language('Kalaallisut'): 57, Language('Hebrew'): 22, Language('Hindi'): 42, Language('Hungarian'): 15,
|
|
Language('Icelandic'): 6, Language('Indonesian'): 54, Language('Irish'): 49, Language('Italian'): 9,
|
|
Language('Japanese'): 11, Language('Kazakh'): 58, Language('Korean'): 4, Language('Latvian'): 21,
|
|
Language('Lithuanian'): 19, Language('Macedonian'): 35, Language('Malay'): 55,
|
|
Language('Norwegian'): 3, Language('Polish'): 26, Language('Portuguese'): 32, Language('Romanian'): 13,
|
|
Language('Russian'): 27, Language('Serbian'): 36, Language('Sinhala'): 56, Language('Slovak'): 37,
|
|
Language('Slovenian'): 1, Language('Spanish'): 28, Language('Swedish'): 25, Language('Thai'): 44,
|
|
Language('Turkish'): 30, Language('Ukrainian'): 46, Language('Vietnamese'): 51,
|
|
29: Language('Albanian'), 12: Language('Arabic'), 14: Language('Spanish (Argentina)'), 50: Language('Belarusian'),
|
|
10: Language('Bosnian'), 48: Language('Portuguese (Brazil)'), 33: Language('Bulgarian'), 53: Language('Catalan'),
|
|
17: Language('Chinese'), 38: Language('Croatian'), 7: Language('Czech'), 24: Language('Danish'),
|
|
23: Language('Dutch'), 2: Language('English'), 20: Language('Estonian'), 52: Language('Persian'),
|
|
31: Language('Finnish'), 8: Language('French'), 5: Language('German'), 16: Language('gre'),
|
|
57: Language('Kalaallisut'), 22: Language('Hebrew'), 42: Language('Hindi'), 15: Language('Hungarian'),
|
|
6: Language('Icelandic'), 54: Language('Indonesian'), 49: Language('Irish'), 9: Language('Italian'),
|
|
11: Language('Japanese'), 58: Language('Kazakh'), 4: Language('Korean'), 21: Language('Latvian'),
|
|
19: Language('Lithuanian'), 35: Language('Macedonian'), 55: Language('Malay'), 40: Language('Chinese'),
|
|
3: Language('Norwegian'), 26: Language('Polish'), 32: Language('Portuguese'), 13: Language('Romanian'),
|
|
27: Language('Russian'), 36: Language('Serbian'), 47: Language('Serbian'), 56: Language('Sinhala'),
|
|
37: Language('Slovak'), 1: Language('Slovenian'), 28: Language('Spanish'), 25: Language('Swedish'),
|
|
44: Language('Thai'), 30: Language('Turkish'), 46: Language('Ukrainian'), Language('Vietnamese'): 51}
|
|
|
|
def list_checked(self, video, languages):
|
|
if isinstance(video, Movie):
|
|
return self.query(video.path or video.release, languages, video.title, year=video.year,
|
|
keywords=get_keywords(video.guess))
|
|
if isinstance(video, Episode):
|
|
return self.query(video.path or video.release, languages, video.series, season=video.season,
|
|
episode=video.episode, keywords=get_keywords(video.guess))
|
|
|
|
def query(self, filepath, languages, title, season=None, episode=None, year=None, keywords=None):
|
|
params = {'sXML': 1, 'sK': title, 'sJ': ','.join([str(self.get_code(l)) for l in languages])}
|
|
if season is not None:
|
|
params['sTS'] = season
|
|
if episode is not None:
|
|
params['sTE'] = episode
|
|
if year is not None:
|
|
params['sY'] = year
|
|
if keywords is not None:
|
|
params['sR'] = keywords
|
|
r = self.session.get(self.server_url + '/ppodnapisi/search', params=params)
|
|
if r.status_code != 200:
|
|
logger.error(u'Request %s returned status code %d' % (r.url, r.status_code))
|
|
return []
|
|
subtitles = []
|
|
soup = BeautifulSoup(r.content, self.required_features)
|
|
for sub in soup('subtitle'):
|
|
if 'n' in sub.flags:
|
|
logger.debug(u'Skipping hearing impaired')
|
|
continue
|
|
language = self.get_language(sub.languageId.text)
|
|
confidence = float(sub.rating.text) / 5.0
|
|
sub_keywords = set()
|
|
for release in sub.release.text.split():
|
|
sub_keywords |= get_keywords(guessit.guess_file_info(release + '.srt', 'autodetect'))
|
|
sub_path = get_subtitle_path(filepath, language, self.config.multi)
|
|
subtitle = ResultSubtitle(sub_path, language, self.__class__.__name__.lower(),
|
|
sub.url.text, confidence=confidence, keywords=sub_keywords)
|
|
subtitles.append(subtitle)
|
|
return subtitles
|
|
|
|
def download(self, subtitle):
|
|
r = self.session.get(subtitle.link)
|
|
if r.status_code != 200:
|
|
raise DownloadFailedError()
|
|
soup = BeautifulSoup(r.content)
|
|
self.download_zip_file(self.server_url + soup.find('a', href=re.compile('download'))['href'], subtitle.path)
|
|
return subtitle
|
|
|
|
|
|
Service = PodnapisiWeb
|