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/custom_libs/subliminal_patch/providers/jimaku.py

419 lines
18 KiB

from __future__ import absolute_import
from datetime import timedelta
import logging
import os
import re
import time
from requests import Session
from subliminal import region, __short_version__
from subliminal.cache import REFINER_EXPIRATION_TIME
from subliminal.exceptions import ConfigurationError, AuthenticationError, ServiceUnavailable
from subliminal.utils import sanitize
from subliminal.video import Episode, Movie
from subliminal_patch.providers import Provider
from subliminal_patch.subtitle import Subtitle
from subliminal_patch.exceptions import APIThrottled
from subliminal_patch.providers.utils import get_subtitle_from_archive, get_archive_from_bytes
from urllib.parse import urlencode, urljoin
from guessit import guessit
from subzero.language import Language, FULL_LANGUAGE_LIST
logger = logging.getLogger(__name__)
# Unhandled formats, such files will always get filtered out
unhandled_archive_formats = (".7z",)
accepted_archive_formats = (".zip", ".rar")
class JimakuSubtitle(Subtitle):
'''Jimaku Subtitle.'''
provider_name = 'jimaku'
hash_verifiable = False
def __init__(self, language, video, download_url, filename):
super(JimakuSubtitle, self).__init__(language, page_link=download_url)
self.video = video
self.download_url = download_url
self.filename = filename
self.release_info = filename
self.is_archive = filename.endswith(accepted_archive_formats)
@property
def id(self):
return self.download_url
def get_matches(self, video):
matches = set()
# Episode/Movie specific matches
if isinstance(video, Episode):
if sanitize(video.series) and sanitize(self.video.series) in (
sanitize(name) for name in [video.series] + video.alternative_series):
matches.add('series')
if video.season and self.video.season is None or video.season and video.season == self.video.season:
matches.add('season')
elif isinstance(video, Movie):
if sanitize(video.title) and sanitize(self.video.title) in (
sanitize(name) for name in [video.title] + video.alternative_titles):
matches.add('title')
# General matches
if video.year and video.year == self.video.year:
matches.add('year')
video_type = 'movie' if isinstance(video, Movie) else 'episode'
matches.add(video_type)
guess = guessit(self.filename, {'type': video_type})
for g in guess:
if g[0] == "release_group" or "source":
if video.release_group == g[1]:
matches.add('release_group')
break
# Prioritize .srt by repurposing the audio_codec match
if self.filename.endswith(".srt"):
matches.add('audio_codec')
return matches
class JimakuProvider(Provider):
'''Jimaku Provider.'''
video_types = (Episode, Movie)
api_url = 'https://jimaku.cc/api'
api_ratelimit_max_delay_seconds = 5
api_ratelimit_backoff_limit = 3
corrupted_file_size_threshold = 500
languages = {Language.fromietf("ja")}
def __init__(self, enable_name_search_fallback, enable_archives_download, enable_ai_subs, api_key):
if api_key:
self.api_key = api_key
else:
raise ConfigurationError('Missing api_key.')
self.enable_name_search_fallback = enable_name_search_fallback
self.download_archives = enable_archives_download
self.enable_ai_subs = enable_ai_subs
self.session = None
def initialize(self):
self.session = Session()
self.session.headers['Content-Type'] = 'application/json'
self.session.headers['Authorization'] = self.api_key
self.session.headers['User-Agent'] = os.environ.get("SZ_USER_AGENT")
def terminate(self):
self.session.close()
def _query(self, video):
if isinstance(video, Movie):
media_name = video.title.lower()
elif isinstance(video, Episode):
media_name = video.series.lower()
# With entries that have a season larger than 1, Jimaku appends the corresponding season number to the name.
# We'll reassemble media_name here to account for cases where we can only search by name alone.
season_addendum = str(video.season) if video.season > 1 else None
media_name = f"{media_name} {season_addendum}" if season_addendum else media_name
# Search for entry
searching_for_entry_attempts = 0
additional_url_params = {}
while searching_for_entry_attempts < 2:
searching_for_entry_attempts += 1
url = self._assemble_jimaku_search_url(video, media_name, additional_url_params)
if not url:
return None
searching_for_entry = "query" in url
data = self._search_for_entry(url)
if not data:
if searching_for_entry and searching_for_entry_attempts < 2:
logger.info("Maybe this is live action media? Will retry search without anime parameter...")
additional_url_params = {'anime': "false"}
else:
return None
else:
break
# We only go for the first entry
entry = data[0]
entry_id = entry.get('id')
anilist_id = entry.get('anilist_id', None)
entry_name = entry.get('name')
is_movie = entry.get('flags', {}).get('movie', False)
if isinstance(video, Episode) and is_movie:
logger.warn("Bazarr thinks this is a series, but Jimaku says this is a movie! May not be able to match subtitles...")
logger.info(f"Matched entry: ID: '{entry_id}', anilist_id: '{anilist_id}', name: '{entry_name}', english_name: '{entry.get('english_name')}', movie: {is_movie}")
if entry.get("flags").get("unverified"):
logger.warning(f"This entry '{entry_id}' is unverified, subtitles might be incomplete or have quality issues!")
# Get a list of subtitles for entry
episode_number = video.episode if "episode" in dir(video) else None
url_params = {'episode': episode_number} if isinstance(video, Episode) and not is_movie else {}
only_look_for_archives = False
has_offset = isinstance(video, Episode) and video.series_anidb_season_episode_offset is not None
retry_count = 0
adjusted_ep_num = None
while retry_count <= 1:
# Account for positive episode offset first
if isinstance(video, Episode) and not is_movie and retry_count < 1:
if video.season > 1 and has_offset:
offset_value = video.series_anidb_season_episode_offset
offset_value = offset_value if offset_value > 0 else -offset_value
if episode_number < offset_value:
adjusted_ep_num = episode_number + offset_value
logger.warning(f"Will try using adjusted episode number {adjusted_ep_num} first")
url_params = {'episode': adjusted_ep_num}
url = f"entries/{entry_id}/files"
data = self._search_for_subtitles(url, url_params)
if not data:
if isinstance(video, Episode) and not is_movie and has_offset and retry_count < 1:
logger.warning(f"Found no subtitles for adjusted episode number, but will retry with normal episode number {episode_number}")
url_params = {'episode': episode_number}
elif isinstance(video, Episode) and not is_movie and retry_count < 1:
logger.warning(f"Found no subtitles for episode number {episode_number}, but will retry without 'episode' parameter")
url_params = {}
only_look_for_archives = True
else:
return None
retry_count += 1
else:
if adjusted_ep_num:
video.episode = adjusted_ep_num
logger.debug(f"This videos episode attribute has been updated to: {video.episode}")
break
# Filter subtitles
list_of_subtitles = []
data = [item for item in data if not item['name'].endswith(unhandled_archive_formats)]
# Detect only archives being uploaded
archive_entries = [item for item in data if item['name'].endswith(accepted_archive_formats)]
subtitle_entries = [item for item in data if not item['name'].endswith(accepted_archive_formats)]
has_only_archives = len(archive_entries) > 0 and len(subtitle_entries) == 0
if has_only_archives:
logger.warning("Have only found archived subtitles")
elif only_look_for_archives:
data = [item for item in data if item['name'].endswith(accepted_archive_formats)]
for item in data:
filename = item.get('name')
download_url = item.get('url')
is_archive = filename.endswith(accepted_archive_formats)
# Archives will still be considered if they're the only files available, as is mostly the case for movies.
if is_archive and not has_only_archives and not self.download_archives:
logger.warning(f"Skipping archive '{filename}' because normal subtitles are available instead")
continue
if not self.enable_ai_subs:
p = re.compile(r'[\[\(]?(whisperai)[\]\)]?|[\[\(]whisper[\]\)]', re.IGNORECASE)
if p.search(filename):
logger.warning(f"Skipping subtitle '{filename}' as it's suspected of being AI generated")
continue
sub_languages = self._try_determine_subtitle_languages(filename)
if len(sub_languages) > 1:
logger.warning(f"Skipping subtitle '{filename}' as it's suspected of containing multiple languages")
continue
# Check if file is obviously corrupt. If no size is returned, assume OK
filesize = item.get('size', self.corrupted_file_size_threshold)
if filesize < self.corrupted_file_size_threshold:
logger.warning(f"Skipping possibly corrupt file '{filename}': Filesize is just {filesize} bytes")
continue
if not filename.endswith(unhandled_archive_formats):
lang = sub_languages[0] if len(sub_languages) > 1 else Language("jpn")
list_of_subtitles.append(JimakuSubtitle(lang, video, download_url, filename))
else:
logger.debug(f"Skipping archive '{filename}' as it's not a supported format")
return list_of_subtitles
def list_subtitles(self, video, languages=None):
subtitles = self._query(video)
if not subtitles:
return []
return [s for s in subtitles]
def download_subtitle(self, subtitle: JimakuSubtitle):
target_url = subtitle.download_url
response = self.session.get(target_url, timeout=10)
response.raise_for_status()
if subtitle.is_archive:
archive = get_archive_from_bytes(response.content)
if archive:
if isinstance(subtitle.video, Episode):
subtitle.content = get_subtitle_from_archive(
archive,
episode=subtitle.video.episode,
episode_title=subtitle.video.title
)
else:
subtitle.content = get_subtitle_from_archive(
archive
)
else:
logger.warning("Archive seems to not be an archive! File possibly corrupt?")
return None
else:
subtitle.content = response.content
def _do_jimaku_request(self, url_path, url_params={}):
url = urljoin(f"{self.api_url}/{url_path}", '?' + urlencode(url_params))
retry_count = 0
while retry_count < self.api_ratelimit_backoff_limit:
response = self.session.get(url, timeout=10)
if response.status_code == 429:
reset_time = 5
retry_count + 1
logger.warning(f"Jimaku ratelimit hit, waiting for '{reset_time}' seconds ({retry_count}/{self.api_ratelimit_backoff_limit} tries)")
time.sleep(reset_time)
continue
elif response.status_code == 401:
raise AuthenticationError("Unauthorized. API key possibly invalid")
else:
response.raise_for_status()
data = response.json()
logger.debug(f"Length of response on {url}: {len(data)}")
if len(data) == 0:
logger.error(f"Jimaku returned no items for our our query: {url}")
return None
elif 'error' in data:
raise ServiceUnavailable(f"Jimaku returned an error: '{data.get('error')}', Code: '{data.get('code')}'")
else:
return data
raise APIThrottled(f"Jimaku ratelimit max backoff limit of {self.api_ratelimit_backoff_limit} reached, aborting")
# Wrapper functions to indirectly call _do_jimaku_request with different cache configs
@region.cache_on_arguments(expiration_time=REFINER_EXPIRATION_TIME)
def _search_for_entry(self, url_path, url_params={}):
return self._do_jimaku_request(url_path, url_params)
@region.cache_on_arguments(expiration_time=timedelta(minutes=1).total_seconds())
def _search_for_subtitles(self, url_path, url_params={}):
return self._do_jimaku_request(url_path, url_params)
@staticmethod
def _try_determine_subtitle_languages(filename):
# This is more like a guess and not a 100% fool-proof way of detecting multi-lang subs:
# It assumes that language codes, if present, are in the last metadata group of the subs filename.
# If such codes are not present, or we failed to match any at all, then we'll just assume that the sub is purely Japanese.
default_language = Language("jpn")
dot_delimit = filename.split(".")
bracket_delimit = re.split(r'[\[\]\(\)]+', filename)
candidate_list = list()
if len(dot_delimit) > 2:
candidate_list = dot_delimit[-2]
elif len(bracket_delimit) > 2:
candidate_list = bracket_delimit[-2]
candidates = [] if len(candidate_list) == 0 else re.split(r'[,\-\+\& ]+', candidate_list)
# Discard match group if any candidate...
# ...contains any numbers, as the group is likely encoding information
if any(re.compile(r'\d').search(string) for string in candidates):
return [default_language]
# ...is >= 5 chars long, as the group is likely other unrelated metadata
if any(len(string) >= 5 for string in candidates):
return [default_language]
languages = list()
for candidate in candidates:
candidate = candidate.lower()
if candidate in ["ass", "srt"]:
continue
# Sometimes, languages are hidden in 4 character blocks, i.e. "JPSC"
if len(candidate) == 4:
for addendum in [candidate[:2], candidate[2:]]:
candidates.append(addendum)
continue
# Sometimes, language codes can have additional info such as 'cc' or 'sdh'. For example: "ja[cc]"
if len(dot_delimit) > 2 and any(c in candidate for c in '[]()'):
candidate = re.split(r'[\[\]\(\)]+', candidate)[0]
try:
language_squash = {
"jp": "ja",
"jap": "ja",
"chs": "zho",
"cht": "zho",
"zhi": "zho",
"cn": "zho"
}
candidate = language_squash[candidate] if candidate in language_squash else candidate
if len(candidate) > 2:
language = Language(candidate)
else:
language = Language.fromietf(candidate)
if not any(l.alpha3 == language.alpha3 for l in languages):
languages.append(language)
except:
if candidate in FULL_LANGUAGE_LIST:
# Create a dummy for the unknown language
languages.append(Language("zul"))
if len(languages) > 1:
# Sometimes a metadata group that actually contains info about codecs gets processed as valid languages.
# To prevent false positives, we'll check if Japanese language codes are in the processed languages list.
# If not, then it's likely that we didn't actually match language codes -> Assume Japanese only subtitle.
contains_jpn = any([l for l in languages if l.alpha3 == "jpn"])
return languages if contains_jpn else [Language("jpn")]
else:
return [default_language]
def _assemble_jimaku_search_url(self, video, media_name, additional_params={}):
endpoint = "entries/search"
anilist_id = video.anilist_id
params = {}
if anilist_id:
params = {'anilist_id': anilist_id}
else:
if self.enable_name_search_fallback or isinstance(video, Movie):
params = {'query': media_name}
else:
logger.error(f"Skipping '{media_name}': Got no AniList ID and fuzzy matching using name is disabled")
return None
if additional_params:
params.update(additional_params)
logger.info(f"Will search for entry based on params: {params}")
return urljoin(endpoint, '?' + urlencode(params))