mirror of https://github.com/morpheus65535/bazarr
426 lines
18 KiB
Python
426 lines
18 KiB
Python
# coding=utf-8
|
||
import io
|
||
import logging
|
||
import re
|
||
import os
|
||
import time
|
||
|
||
from babelfish import language_converters
|
||
from subzero.language import Language
|
||
from requests import Session
|
||
import urllib.parse
|
||
|
||
from subliminal.subtitle import fix_line_ending
|
||
from subliminal_patch.providers import Provider
|
||
from subliminal_patch.providers.mixins import ProviderSubtitleArchiveMixin
|
||
from subliminal.providers import ParserBeautifulSoup
|
||
from bs4.element import Tag, NavigableString
|
||
from subliminal.score import get_equivalent_release_groups
|
||
from subliminal_patch.subtitle import Subtitle, guess_matches
|
||
from subliminal.utils import sanitize, sanitize_release_group
|
||
from subliminal.video import Episode, Movie
|
||
from zipfile import ZipFile
|
||
from rarfile import RarFile, is_rarfile
|
||
from subliminal_patch.utils import sanitize
|
||
from guessit import guessit
|
||
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
language_converters.register('supersubtitles = subliminal_patch.converters.supersubtitles:SuperSubtitlesConverter')
|
||
|
||
|
||
class SuperSubtitlesSubtitle(Subtitle):
|
||
"""SuperSubtitles Subtitle."""
|
||
provider_name = 'supersubtitles'
|
||
|
||
def __str__(self):
|
||
subtit = "Subtitle id: " + str(self.subtitle_id) \
|
||
+ " Series: " + self.series \
|
||
+ " Season: " + str(self.season) \
|
||
+ " Episode: " + str(self.episode) \
|
||
+ " Version: " + str(self.version) \
|
||
+ " Releases: " + str(self.releases) \
|
||
+ " DownloadLink: " + str(self.page_link) \
|
||
+ " Matches: " + str(self.matches)
|
||
if self.year:
|
||
subtit = subtit + " Year: " + str(self.year)
|
||
return subtit.encode('utf-8')
|
||
|
||
def __init__(self, language, page_link, subtitle_id, series, season, episode, version,
|
||
releases, year, imdb_id, uploader, asked_for_episode=None, asked_for_release_group=None):
|
||
super(SuperSubtitlesSubtitle, self).__init__(language, page_link=page_link)
|
||
self.subtitle_id = subtitle_id
|
||
self.series = series
|
||
self.season = season
|
||
self.episode = episode
|
||
self.version = version
|
||
self.releases = releases
|
||
self.year = year
|
||
self.uploader = uploader
|
||
if year:
|
||
self.year = int(year)
|
||
|
||
self.release_info = u", ".join(releases)
|
||
self.page_link = page_link
|
||
self.asked_for_release_group = asked_for_release_group
|
||
self.asked_for_episode = asked_for_episode
|
||
self.imdb_id = imdb_id
|
||
self.is_pack = True
|
||
self.matches = None
|
||
|
||
def numeric_id(self):
|
||
return self.subtitle_id
|
||
|
||
def __repr__(self):
|
||
ep_addon = (" S%02dE%02d" % (self.season, self.episode)) if self.episode else ""
|
||
return '<%s %r [%s]>' % (
|
||
self.__class__.__name__, u"%s%s%s [%s]" % (self.series, " (%s)" % self.year if self.year else "", ep_addon,
|
||
self.release_info), self.language)
|
||
|
||
@property
|
||
def id(self):
|
||
return str(self.subtitle_id)
|
||
|
||
def get_matches(self, video):
|
||
matches = guess_matches(video, guessit(self.release_info))
|
||
|
||
# episode
|
||
if isinstance(video, Episode):
|
||
# series
|
||
if video.series and sanitize(self.series) == sanitize(video.series):
|
||
matches.add('series')
|
||
# season
|
||
if video.season and self.season == video.season:
|
||
matches.add('season')
|
||
# episode
|
||
if video.episode and self.episode == video.episode:
|
||
matches.add('episode')
|
||
# imdb_id
|
||
if video.series_imdb_id and self.imdb_id and str(self.imdb_id) == str(video.series_imdb_id):
|
||
matches.add('series_imdb_id')
|
||
matches.add('series')
|
||
matches.add('year')
|
||
# year
|
||
if ('series' in matches and video.original_series and self.year is None or
|
||
video.year and video.year == self.year):
|
||
matches.add('year')
|
||
# movie
|
||
elif isinstance(video, Movie):
|
||
# title
|
||
if video.title and (sanitize(self.series) in (
|
||
sanitize(name) for name in [video.title] + video.alternative_titles)):
|
||
matches.add('title')
|
||
# imdb_id
|
||
if video.imdb_id and self.imdb_id == video.imdb_id:
|
||
matches.add('imdb_id')
|
||
matches.add('title')
|
||
matches.add('year')
|
||
# year
|
||
if video.year and self.year == video.year:
|
||
matches.add('year')
|
||
|
||
# release_group
|
||
if (video.release_group and self.version and
|
||
any(r in sanitize_release_group(self.version)
|
||
for r in get_equivalent_release_groups(sanitize_release_group(video.release_group)))):
|
||
matches.add('release_group')
|
||
# resolution
|
||
if video.resolution and self.version and video.resolution in self.version.lower():
|
||
matches.add('resolution')
|
||
# source
|
||
if video.source and self.version and video.source.lower() in self.version.lower():
|
||
matches.add('source')
|
||
|
||
self.matches = matches
|
||
return matches
|
||
|
||
|
||
class SuperSubtitlesProvider(Provider, ProviderSubtitleArchiveMixin):
|
||
"""SuperSubtitles Provider."""
|
||
languages = {Language('hun', 'HU')} | {Language(lang) for lang in [
|
||
'hun', 'eng'
|
||
]}
|
||
video_types = (Episode, Movie)
|
||
# https://www.feliratok.info/?search=&soriSorszam=&nyelv=&sorozatnev=The+Flash+%282014%29&sid=3212&complexsearch=true&knyelv=0&evad=4&epizod1=1&cimke=0&minoseg=0&rlsr=0&tab=all
|
||
server_url = 'https://www.feliratok.info/'
|
||
hearing_impaired_verifiable = False
|
||
multi_result_throttle = 2 # seconds
|
||
|
||
def __init__(self):
|
||
self.session = None
|
||
|
||
def initialize(self):
|
||
self.session = Session()
|
||
self.session.headers = {'User-Agent': os.environ.get("SZ_USER_AGENT", "Sub-Zero/2")}
|
||
|
||
def terminate(self):
|
||
self.session.close()
|
||
|
||
@staticmethod
|
||
def get_language(text):
|
||
if text == 'Magyar':
|
||
return Language.fromsupersubtitles('hu')
|
||
if text == 'Angol':
|
||
return Language.fromsupersubtitles('en')
|
||
return None
|
||
|
||
def find_imdb_id(self, sub_id):
|
||
"""
|
||
|
||
"""
|
||
|
||
url = self.server_url + "index.php?tipus=adatlap&azon=a_" + sub_id
|
||
# url = https://www.feliratok.info/index.php?tipus=adatlap&azon=a_1518600916
|
||
logger.info('Get IMDB id from URL %s', url)
|
||
r = self.session.get(url, timeout=10).content
|
||
|
||
soup = ParserBeautifulSoup(r, ['lxml'])
|
||
links = soup.find_all("a")
|
||
|
||
for value in links:
|
||
if "imdb.com" in str(value):
|
||
# <a alt="iMDB" href="http://www.imdb.com/title/tt2357547/" target="_blank"><img alt="iMDB"
|
||
# src="img/adatlap/imdb.png"/></a>
|
||
imdb_id = re.search(r'(?<=www\.imdb\.com/title/).*(?=/")', str(value))
|
||
imdb_id = imdb_id.group() if imdb_id else ''
|
||
return imdb_id
|
||
|
||
return None
|
||
|
||
def find_id(self, series, year, original_title):
|
||
"""
|
||
We need to find the id of the series at the following url:
|
||
https://www.feliratok.info/index.php?term=SERIESNAME&nyelv=0&action=autoname
|
||
Where SERIESNAME is a searchable string.
|
||
The result will be something like this:
|
||
[{"name":"DC\u2019s Legends of Tomorrow (2016)","ID":"3725"},{"name":"Miles from Tomorrowland (2015)",
|
||
"ID":"3789"},{"name":"No Tomorrow (2016)","ID":"4179"}]
|
||
|
||
"""
|
||
|
||
# Search for exact name
|
||
url = self.server_url + "index.php?term=" + series + "&nyelv=0&action=autoname"
|
||
# url = self.server_url + "index.php?term=" + "fla"+ "&nyelv=0&action=autoname"
|
||
logger.info('Get series id from URL %s', url)
|
||
r = self.session.get(url, timeout=10)
|
||
|
||
# r is something like this:
|
||
# [{"name":"DC\u2019s Legends of Tomorrow (2016)","ID":"3725"},{"name":"Miles from Tomorrowland (2015)",
|
||
# "ID":"3789"},{"name":"No Tomorrow (2016)","ID":"4179"}]
|
||
|
||
results = r.json()
|
||
|
||
# check all of the results:
|
||
for result in results:
|
||
try:
|
||
# "name":"Miles from Tomorrowland (2015)","ID":"3789"
|
||
result_year = re.search(r"(?<=\()\d\d\d\d(?=\))", result['name'])
|
||
result_year = result_year.group() if result_year else ''
|
||
except IndexError:
|
||
result_year = ""
|
||
|
||
try:
|
||
# "name":"Miles from Tomorrowland (2015)","ID":"3789"
|
||
result_title = re.search(r".*(?=\(\d\d\d\d\))", result['name'])
|
||
result_title = result_title.group() if result_title else ''
|
||
result_id = result['ID']
|
||
except IndexError:
|
||
continue
|
||
|
||
result_title = result_title.strip().replace("<EFBFBD>", "").replace(" ", ".")
|
||
if not result_title:
|
||
continue
|
||
|
||
guessable = result_title.strip() + ".s01e01." + result_year
|
||
guess = guessit(guessable, {'type': "episode"})
|
||
|
||
if sanitize(original_title) == sanitize(guess['title']) and year and guess['year'] and \
|
||
year == guess['year']:
|
||
# Return the founded id
|
||
return result_id
|
||
elif sanitize(original_title) == sanitize(guess['title']) and not year:
|
||
# Return the founded id
|
||
return result_id
|
||
|
||
return None
|
||
|
||
def query(self, series, languages, video=None):
|
||
year = video.year
|
||
subtitle = None
|
||
if isinstance(video, Episode):
|
||
series = video.series
|
||
season = video.season
|
||
episode = video.episode
|
||
# seriesa = series.replace(' ', '+')
|
||
|
||
# Get ID of series with original name
|
||
series_id = self.find_id(series, year, series)
|
||
if not series_id:
|
||
# If not founded try without ' char
|
||
modified_series = urllib.parse.quote_plus(series.replace('\'', ''))
|
||
series_id = self.find_id(modified_series, year, series)
|
||
if not series_id and modified_series:
|
||
# If still not founded try with the longest word is series title
|
||
modified_series = modified_series.split('+')
|
||
modified_series = max(modified_series, key=len)
|
||
series_id = self.find_id(modified_series, year, series)
|
||
|
||
if not series_id:
|
||
return None
|
||
|
||
# https://www.feliratok.info/index.php?search=&soriSorszam=&nyelv=&sorozatnev=&sid=2075&complexsearch=true&
|
||
# knyelv=0&evad=6&epizod1=16&cimke=0&minoseg=0&rlsr=0&tab=all
|
||
url = self.server_url + "index.php?search=&soriSorszam=&nyelv=&sorozatnev=&sid=" + \
|
||
str(series_id) + "&complexsearch=true&knyelv=0&evad=" + str(season) + "&epizod1=" + \
|
||
str(episode) + "&cimke=0&minoseg=0&rlsr=0&tab=all"
|
||
subtitle = self.process_subs(languages, video, url)
|
||
|
||
if not subtitle:
|
||
# No Subtitle found. Maybe already archived to season pack
|
||
url = self.server_url + "index.php?search=&soriSorszam=&nyelv=&sorozatnev=&sid=" + \
|
||
str(series_id) + "&complexsearch=true&knyelv=0&evad=" + \
|
||
str(season) + "&epizod1=&evadpakk=on&cimke=0&minoseg=0&rlsr=0&tab=all"
|
||
subtitle = self.process_subs(languages, video, url)
|
||
|
||
if isinstance(video, Movie):
|
||
title = urllib.parse.quote_plus(series)
|
||
|
||
# https://www.feliratok.info/index.php?search=The+Hitman%27s+BodyGuard&soriSorszam=&nyelv=&tab=film
|
||
url = self.server_url + "index.php?search=" + title + "&soriSorszam=&nyelv=&tab=film"
|
||
subtitle = self.process_subs(languages, video, url)
|
||
|
||
return subtitle
|
||
|
||
def process_subs(self, languages, video, url):
|
||
|
||
subtitles = []
|
||
|
||
logger.info('URL for subtitles %s', url)
|
||
r = self.session.get(url, timeout=10).content
|
||
|
||
soup = ParserBeautifulSoup(r, ['lxml'])
|
||
tables = soup.find_all("table")
|
||
tables = tables[0].find_all("tr")
|
||
i = 0
|
||
series_imdb_id = None
|
||
for table in tables:
|
||
if "vilagit" in str(table) and i > 1:
|
||
asked_for_episode = None
|
||
sub_season = None
|
||
sub_episode = None
|
||
sub_english = table.findAll("div", {"class": "eredeti"})
|
||
sub_english_name = None
|
||
if isinstance(video, Episode):
|
||
asked_for_episode = video.episode
|
||
if "Season" not in str(sub_english):
|
||
# [<div class="eredeti">Gossip Girl (Season 3) (DVDRip-REWARD)</div>]
|
||
sub_english_name = re.search(r'(?<=<div class="eredeti">).*?(?= -)',
|
||
str(sub_english))
|
||
sub_english_name = sub_english_name.group() if sub_english_name else ''
|
||
|
||
sub_season = re.search(r"(?<=- ).*?(?= - )", str(sub_english))
|
||
sub_season = sub_season.group() if sub_season else ''
|
||
sub_season = int((sub_season.split('x')[0]).strip())
|
||
|
||
sub_episode = re.search(r"(?<=- ).*?(?= - )", str(sub_english))
|
||
sub_episode = sub_episode.group() if sub_episode else ''
|
||
sub_episode = int((sub_episode.split('x')[1]).strip())
|
||
|
||
else:
|
||
# [<div class="eredeti">DC's Legends of Tomorrow - 3x11 - Here I Go Again (HDTV-AFG, HDTV-RMX,
|
||
# 720p-SVA, 720p-PSA </div>]
|
||
sub_english_name = \
|
||
re.search(r'(?<=<div class="eredeti">).*?(?=\(Season)', str(sub_english))
|
||
sub_english_name = sub_english_name.group() if sub_english_name else ''
|
||
sub_season = re.search(r"(?<=Season )\d+(?=\))", str(sub_english))
|
||
sub_season = int(sub_season.group()) if sub_season else None
|
||
sub_episode = int(video.episode)
|
||
if isinstance(video, Movie):
|
||
sub_english_name = re.search(r'(?<=<div class="eredeti">).*?(?=</div>)', str(sub_english))
|
||
sub_english_name = sub_english_name.group() if sub_english_name else ''
|
||
sub_english_name = sub_english_name.split(' (')[0]
|
||
|
||
sub_version = 'n/a'
|
||
if len(str(sub_english).split('(')) > 1:
|
||
sub_version = (str(sub_english).split('(')[len(str(sub_english).split('(')) - 1]).split(')')[0]
|
||
# <small>Angol</small>
|
||
lang = table.find("small")
|
||
sub_language = re.search(r"(?<=<small>).*(?=</small>)", str(lang))
|
||
sub_language = sub_language.group() if sub_language else ''
|
||
sub_language = self.get_language(sub_language)
|
||
|
||
# <a href="/index.php?action=letolt&fnev=DCs Legends of Tomorrow - 03x11 - Here I Go Again.SVA.
|
||
# English.C.orig.Addic7ed.com.srt&felirat=1519162191">
|
||
link = str(table.findAll("a")[len(table.findAll("a")) - 1]).replace("amp;", "")
|
||
sub_downloadlink = re.search(r'(?<=href="/).*(?=">)', link)
|
||
sub_downloadlink = sub_downloadlink.group() if sub_downloadlink else ''
|
||
sub_downloadlink = self.server_url + sub_downloadlink
|
||
|
||
sub_id = re.search(r"(?<=felirat=).*(?=\">)", link)
|
||
sub_id = sub_id.group() if sub_id else ''
|
||
sub_year = video.year
|
||
sub_releases = [s.strip() for s in sub_version.split(',')]
|
||
|
||
uploader = ''
|
||
for item in table.contents[7].contents:
|
||
if isinstance(item, Tag):
|
||
uploader = item.text.lstrip('\r\n\t\t\t\t\t').rstrip('\r\n\t\t\t\t')
|
||
elif isinstance(item, NavigableString):
|
||
uploader = item.lstrip('\r\n\t\t\t\t\t').rstrip('\r\n\t\t\t\t')
|
||
|
||
# For episodes we open the series page so all subtitles imdb_id must be the same. no need to check all
|
||
if isinstance(video, Episode) and series_imdb_id is not None:
|
||
sub_imdb_id = series_imdb_id
|
||
else:
|
||
sub_imdb_id = self.find_imdb_id(sub_id)
|
||
series_imdb_id = sub_imdb_id
|
||
|
||
subtitle = SuperSubtitlesSubtitle(sub_language, sub_downloadlink, sub_id, sub_english_name.strip(),
|
||
sub_season, sub_episode, sub_version, sub_releases, sub_year,
|
||
sub_imdb_id, uploader, asked_for_episode,
|
||
asked_for_release_group=video.release_group)
|
||
if subtitle.language in languages:
|
||
subtitles.append(subtitle)
|
||
i = i + 1
|
||
return subtitles
|
||
|
||
def list_subtitles(self, video, languages):
|
||
titles = []
|
||
if isinstance(video, Episode):
|
||
titles = [video.series] + video.alternative_series
|
||
elif isinstance(video, Movie):
|
||
titles = [video.title] + video.alternative_titles
|
||
|
||
subtitles = []
|
||
|
||
for title in titles:
|
||
subs = self.query(title, languages, video=video)
|
||
if subs:
|
||
for item in subs:
|
||
if item.series in titles:
|
||
subtitles.append(item)
|
||
|
||
time.sleep(self.multi_result_throttle)
|
||
return subtitles
|
||
|
||
def download_subtitle(self, subtitle):
|
||
|
||
# download as a zip
|
||
logger.info('Downloading subtitle %r', subtitle.subtitle_id)
|
||
r = self.session.get(subtitle.page_link, timeout=10)
|
||
r.raise_for_status()
|
||
|
||
if ".rar" in subtitle.page_link:
|
||
logger.debug('Archive identified as rar')
|
||
archive_stream = io.BytesIO(r.content)
|
||
archive = RarFile(archive_stream)
|
||
subtitle.content = self.get_subtitle_from_archive(subtitle, archive)
|
||
elif ".zip" in subtitle.page_link:
|
||
logger.debug('Archive identified as zip')
|
||
archive_stream = io.BytesIO(r.content)
|
||
archive = ZipFile(archive_stream)
|
||
subtitle.content = self.get_subtitle_from_archive(subtitle, archive)
|
||
else:
|
||
subtitle.content = fix_line_ending(r.content)
|