bazarr/custom_libs/subliminal_patch/providers/avistaz_network.py

374 lines
9.1 KiB
Python
Raw Normal View History

import logging
import time
from http.cookies import SimpleCookie
from random import randint
import pycountry
from requests.cookies import RequestsCookieJar
from subliminal.exceptions import AuthenticationError
from subliminal.providers import ParserBeautifulSoup
from subliminal_patch.http import RetryingCFSession
from subliminal_patch.pitcher import store_verification
from subliminal_patch.providers import Provider
from subliminal_patch.subtitle import Subtitle
from subzero.language import Language
from .utils import get_archive_from_bytes, get_subtitle_from_archive, FIRST_THOUSAND_OR_SO_USER_AGENTS as AGENT_LIST
logger = logging.getLogger(__name__)
supported_languages_names = [
"Abkhazian",
"Afar",
"Afrikaans",
"Akan",
"Albanian",
"Amharic",
"Arabic",
"Aragonese",
"Armenian",
"Assamese",
"Avaric",
"Avestan",
"Aymara",
"Azerbaijani",
"Bambara",
"Bashkir",
"Basque",
"Belarusian",
"Bengali",
"Bihari languages",
"Bislama",
"Bokmål, Norwegian",
"Bosnian",
"Brazilian Portuguese",
"Breton",
"Bulgarian",
"Burmese",
"Cantonese",
"Catalan",
"Central Khmer",
"Chamorro",
"Chechen",
"Chichewa",
"Chinese",
"Church Slavic",
"Chuvash",
"Cornish",
"Corsican",
"Cree",
"Croatian",
"Czech",
"Danish",
"Dhivehi",
"Dutch",
"Dzongkha",
"English",
"Esperanto",
"Estonian",
"Ewe",
"Faroese",
"Fijian",
"Filipino",
"Finnish",
"French",
"Fulah",
"Gaelic",
"Galician",
"Ganda",
"Georgian",
"German",
"Greek",
"Guarani",
"Gujarati",
"Haitian",
"Hausa",
"Hebrew",
"Herero",
"Hindi",
"Hiri Motu",
"Hungarian",
"Icelandic",
"Ido",
"Igbo",
"Indonesian",
"Interlingua",
"Interlingue",
"Inuktitut",
"Inupiaq",
"Irish",
"Italian",
"Japanese",
"Javanese",
"Kalaallisut",
"Kannada",
"Kanuri",
"Kashmiri",
"Kazakh",
"Kikuyu",
"Kinyarwanda",
"Kirghiz",
"Komi",
"Kongo",
"Korean",
"Kuanyama",
"Kurdish",
"Lao",
"Latin",
"Latvian",
"Limburgan",
"Lingala",
"Lithuanian",
"Luba-Katanga",
"Luxembourgish",
"Macedonian",
"Malagasy",
"Malay",
"Malayalam",
"Maltese",
"Mandarin",
"Manx",
"Maori",
"Marathi",
"Marshallese",
"Mongolian",
"Moore",
"Nauru",
"Navajo",
"Ndebele, North",
"Ndebele, South",
"Ndonga",
"Nepali",
"Northern Sami",
"Norwegian",
"Norwegian Nynorsk",
"Occitan (post 1500)",
"Ojibwa",
"Oriya",
"Oromo",
"Ossetian",
"Pali",
"Panjabi",
"Persian",
"Polish",
"Portuguese",
"Pushto",
"Quechua",
"Romanian",
"Romansh",
"Rundi",
"Russian",
"Samoan",
"Sango",
"Sanskrit",
"Sardinian",
"Serbian",
"Shona",
"Sichuan Yi",
"Sindhi",
"Sinhala",
"Slovak",
"Slovenian",
"Somali",
"Sotho, Southern",
"Spanish",
"Sundanese",
"Swahili",
"Swati",
"Swedish",
"Tagalog",
"Tahitian",
"Tajik",
"Tamil",
"Tatar",
"Telugu",
"Thai",
"Tibetan",
"Tigrinya",
"Tongan",
"Tsonga",
"Tswana",
"Turkish",
"Turkmen",
"Twi",
"Uighur",
"Ukrainian",
"Urdu",
"Uzbek",
"Venda",
"Vietnamese",
"Volapük",
"Walloon",
"Welsh",
"Western Frisian",
"Wolof",
"Xhosa",
"Yiddish",
"Yoruba",
"Zhuang",
"Zulu"
]
class AvistazNetworkSubtitle(Subtitle):
"""AvistaZ.to Subtitle."""
provider_name = None
def __init__(self, provider_name, page_link, download_link, language, video, filename, release, uploader):
super().__init__(language, page_link=page_link)
self.provider_name = provider_name
self.hearing_impaired = None
self.language = language
self.filename = filename
self.release_info = release
self.page_link = page_link
self.download_link = download_link
self.video = video
self.matches = None
self.content = None
self.uploader = uploader
self.encoding = None
@property
def id(self):
return self.filename
def get_matches(self, video):
# we download subtitles directly from the
# release page, so it's always a perfect match
self.matches = {'hash'}
return self.matches
def lookup_lang(name):
try:
return Language(pycountry.languages.lookup(name).alpha_3)
except:
return None
class AvistazNetworkProviderBase(Provider):
"""AvistaZ Network base provider"""
subtitle_class = AvistazNetworkSubtitle
languages = set(filter(None, map(lookup_lang, supported_languages_names)))
languages.update(set(Language.rebuild(L, hi=True) for L in languages))
server_url = None
provider_name = None
hash_verifiable = True
def __init__(self, cookies, user_agent=None):
self.session = None
self.cookies = cookies
self.user_agent = user_agent
def initialize(self):
self.session = RetryingCFSession()
if self.user_agent:
self.session.headers['User-Agent'] = self.user_agent
else:
self.session.headers['User-Agent'] = AGENT_LIST[randint(0, len(AGENT_LIST) - 1)]
if self.cookies:
self.session.cookies = RequestsCookieJar()
simple_cookie = SimpleCookie()
simple_cookie.load(self.cookies)
for k, v in simple_cookie.items():
self.session.cookies.set(k, v.value)
rr = self.session.get(self.server_url + 'rules', allow_redirects=False, timeout=10,
headers={"Referer": self.server_url})
if rr.status_code in [302, 404, 403]:
logger.info('Cookies expired')
raise AuthenticationError("cookies not valid anymore")
store_verification(self.provider_name, self.session)
logger.debug('Cookies valid')
time.sleep(2)
return True
def terminate(self):
self.session.close()
def list_subtitles(self, video, languages):
if video.info_url is None or not video.info_url.startswith(self.server_url):
logger.debug('%s not downloaded from %s. Skipped', video, self.server_url)
return []
html = self._query_info_url(video.info_url)
if html is None:
logger.debug('%s release page not found. Release might have been removed', video)
return []
release = self._parse_release_table(html)
if release['Subtitles'].table is None:
logger.debug('No subtitles found for %s', video)
return []
subtitle_columns = list(map(lambda x: x.get_text(), release['Subtitles'].thead.find_all('th')))
subtitles = []
for row in release['Subtitles'].tbody.find_all('tr', recursive=False):
subtitle_cols = self._parse_subtitle_row(row, subtitle_columns)
release_name = release['Title'].get_text().strip()
lang = lookup_lang(subtitle_cols['Language'].get_text().strip())
download_link = subtitle_cols['Download'].a['href']
uploader_name = subtitle_cols['Uploader'].get_text().strip()
if lang not in languages:
continue
subtitles.append(self.subtitle_class(
provider_name=self.provider_name,
page_link=video.info_url,
download_link=download_link,
language=lang,
video=video,
filename=download_link.split('/')[-1],
release=release_name,
uploader=uploader_name,
))
return subtitles
def _query_info_url(self, info_url):
response = self.session.get(info_url, timeout=30)
if response.status_code == 404:
return None
else:
response.raise_for_status()
return response.content.decode('utf-8', 'ignore')
def _parse_subtitle_row(self, row, subtitle_columns):
columns = {}
for i, data in enumerate(row.find_all('td', recursive=False)):
columns[subtitle_columns[i]] = data
return columns
def _parse_release_table(self, html):
release_data_table = (ParserBeautifulSoup(html, ['html.parser'])
.select_one('#content-area > div:nth-child(4) > div.table-responsive > table > tbody'))
rows = {}
for tr in release_data_table.find_all('tr', recursive=False):
rows[tr.td.get_text()] = tr.select_one('td:nth-child(2)', recursive=False)
return rows
def download_subtitle(self, subtitle):
response = self.session.get(subtitle.download_link)
response.raise_for_status()
if subtitle.filename.endswith((".zip", ".rar")):
archive = get_archive_from_bytes(response.content)
subtitle.content = get_subtitle_from_archive(
archive, episode=subtitle.video.episode
)
else:
subtitle.content = response.content