You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
bazarr/libs/subliminal_patch/providers/napisy24.py

119 lines
3.7 KiB

import logging
import os
from io import BytesIO
from zipfile import ZipFile
from babelfish import Language
from requests import Session
from subliminal.providers import Provider
from subliminal import __short_version__
from subliminal.exceptions import AuthenticationError, ConfigurationError
from subliminal.subtitle import Subtitle, fix_line_ending
logger = logging.getLogger(__name__)
class Napisy24Subtitle(Subtitle):
'''Napisy24 Subtitle.'''
provider_name = 'napisy24'
def __init__(self, language, hash, imdb_id, napis_id):
super(Napisy24Subtitle, self).__init__(language)
self.hash = hash
self.imdb_id = imdb_id
self.napis_id = napis_id
@property
def id(self):
return self.hash
def get_matches(self, video):
matches = set()
# hash
if 'opensubtitles' in video.hashes and video.hashes['opensubtitles'] == self.hash:
matches.add('hash')
# imdb_id
if video.imdb_id and self.imdb_id == video.imdb_id:
matches.add('imdb_id')
return matches
class Napisy24Provider(Provider):
'''Napisy24 Provider.'''
languages = {Language.fromalpha2(l) for l in ['pl']}
required_hash = 'opensubtitles'
api_url = 'http://napisy24.pl/run/CheckSubAgent.php'
def __init__(self):
self.username = 'subliminal'
self.password = 'lanimilbus'
self.session = None
def initialize(self):
self.session = Session()
self.session.headers['User-Agent'] = 'Subliminal/%s' % __short_version__
self.session.headers['Content-Type'] = 'application/x-www-form-urlencoded'
def terminate(self):
self.session.close()
def query(self, language, size, name, hash):
params = {
'postAction': 'CheckSub',
'ua': self.username,
'ap': self.password,
'fs': size,
'fh': hash,
'fn': os.path.basename(name),
'n24pref': 1
}
response = self.session.post(self.api_url, data=params, timeout=10)
response.raise_for_status()
response_content = response.content.split(b'||', 1)
n24_data = response_content[0].decode()
if n24_data[:2] != 'OK':
if n24_data[:11] == 'login error':
raise AuthenticationError('Login failed')
logger.error('Unknown response: %s', response.content)
return None
n24_status = n24_data[:4]
if n24_status == 'OK-0':
logger.info('No subtitles found')
return None
subtitle_info = dict(p.split(':', 1) for p in n24_data.split('|')[1:])
logger.debug('Subtitle info: %s', subtitle_info)
if n24_status == 'OK-1':
logger.info('No subtitles found but got video info')
return None
elif n24_status == 'OK-2':
logger.info('Found subtitles')
elif n24_status == 'OK-3':
logger.info('Found subtitles but not from Napisy24 database')
return None
subtitle_content = response_content[1]
subtitle = Napisy24Subtitle(language, hash, 'tt%s' % subtitle_info['imdb'].zfill(7), subtitle_info['napisId'])
with ZipFile(BytesIO(subtitle_content)) as zf:
subtitle.content = fix_line_ending(zf.open(zf.namelist()[0]).read())
return subtitle
def list_subtitles(self, video, languages):
subtitles = [self.query(l, video.size, video.name, video.hashes['opensubtitles']) for l in languages]
return [s for s in subtitles if s is not None]
def download_subtitle(self, subtitle):
# there is no download step, content is already filled from listing subtitles
pass