mirror of
https://github.com/SickGear/SickGear.git
synced 2024-11-27 15:13:38 +00:00
467 lines
16 KiB
Python
467 lines
16 KiB
Python
#
|
|
# This file is part of SickGear.
|
|
#
|
|
# SickGear is free software: you can redistribute it and/or modify
|
|
# it under the terms of the GNU General Public License as published by
|
|
# the Free Software Foundation, either version 3 of the License, or
|
|
# (at your option) any later version.
|
|
#
|
|
# SickGear 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 General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU General Public License
|
|
# along with SickGear. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
import datetime
|
|
import itertools
|
|
import time
|
|
|
|
from exceptions_helper import AuthException, ex, MultipleShowObjectsException
|
|
|
|
from . import db, helpers, logger, show_name_helpers
|
|
from .classes import SearchResult
|
|
from .common import Quality
|
|
from .name_parser.parser import InvalidNameException, InvalidShowException, NameParser, ParseResult
|
|
from .rssfeeds import RSSFeeds
|
|
from .sgdatetime import SGDatetime
|
|
from .tv import TVEpisode
|
|
|
|
# noinspection PyUnreachableCode
|
|
if False:
|
|
from typing import Any, AnyStr, Dict, List, Tuple, Union
|
|
from providers.generic import GenericProvider, NZBProvider, TorrentProvider
|
|
|
|
|
|
class CacheDBConnection(db.DBConnection):
|
|
def __init__(self):
|
|
db.DBConnection.__init__(self, 'cache.db')
|
|
|
|
# Create the table if it's not already there
|
|
try:
|
|
if not self.has_table('lastUpdate'):
|
|
self.action('CREATE TABLE lastUpdate (provider TEXT, time NUMERIC)')
|
|
except (BaseException, Exception) as e:
|
|
if ex(e) != 'table lastUpdate already exists':
|
|
raise e
|
|
|
|
|
|
class TVCache(object):
|
|
def __init__(self, provider, interval=10):
|
|
# type: (Union[GenericProvider, NZBProvider, TorrentProvider], int) -> None
|
|
self.provider = provider
|
|
self.providerID = self.provider.get_id()
|
|
self.providerDB = None
|
|
self.update_iv = interval
|
|
|
|
@staticmethod
|
|
def get_db():
|
|
return CacheDBConnection()
|
|
|
|
def clear_cache(self):
|
|
if self.should_clear_cache():
|
|
my_db = self.get_db()
|
|
my_db.action('DELETE FROM provider_cache WHERE provider = ?', [self.providerID])
|
|
|
|
def _title_and_url(self, item):
|
|
"""
|
|
|
|
:param item:
|
|
:type item:
|
|
:return:
|
|
:rtype: Tuple[AnyStr, AnyStr] or Tuple[None, None]
|
|
"""
|
|
# override this in the provider if recent search has a different data layout to backlog searches
|
|
# noinspection PyProtectedMember
|
|
return self.provider._title_and_url(item)
|
|
|
|
def _cache_data(self, **kwargs):
|
|
data = None
|
|
return data
|
|
|
|
def check_auth(self):
|
|
# noinspection PyProtectedMember
|
|
return self.provider._check_auth()
|
|
|
|
def update_cache(self, **kwargs):
|
|
try:
|
|
self.check_auth()
|
|
except AuthException as e:
|
|
logger.error(f'Authentication error: {ex(e)}')
|
|
return []
|
|
|
|
if self.should_update():
|
|
data = self._cache_data(**kwargs)
|
|
|
|
# clear cache
|
|
if data:
|
|
self.clear_cache()
|
|
|
|
# parse data
|
|
cl = []
|
|
for item in data or []:
|
|
title, url = self._title_and_url(item)
|
|
ci = self.parse_item(title, url)
|
|
if None is not ci:
|
|
cl.append(ci)
|
|
|
|
if 0 < len(cl):
|
|
my_db = self.get_db()
|
|
try:
|
|
my_db.mass_action(cl)
|
|
except (BaseException, Exception) as e:
|
|
logger.log('Warning could not save cache value [%s], caught err: %s' % (cl, ex(e)))
|
|
|
|
# set updated as time the attempt to fetch data is
|
|
self.set_last_update()
|
|
|
|
def get_rss(self, url, **kwargs):
|
|
return RSSFeeds(self.provider).get_feed(url, **kwargs)
|
|
|
|
@staticmethod
|
|
def _translate_title(title):
|
|
"""
|
|
|
|
:param title: title
|
|
:type title: AnyStr
|
|
:return:
|
|
:rtype: AnyStr
|
|
"""
|
|
return f'{title.replace(" ", ".")}'
|
|
|
|
@staticmethod
|
|
def _translate_link_url(url):
|
|
"""
|
|
|
|
:param url: url
|
|
:type url: AnyStr
|
|
:return:
|
|
:rtype: AnyStr
|
|
"""
|
|
return url.replace('&', '&')
|
|
|
|
def parse_item(self, title, url):
|
|
"""
|
|
|
|
:param title: title
|
|
:type title: AnyStr
|
|
:param url: url
|
|
:type url: AnyStr
|
|
:return:
|
|
:rtype: None or List[AnyStr, List[Any]]
|
|
"""
|
|
if title and url:
|
|
title = self._translate_title(title)
|
|
url = self._translate_link_url(url)
|
|
|
|
return self.add_cache_entry(title, url)
|
|
|
|
logger.debug('Data returned from the %s feed is incomplete, this result is unusable' % self.provider.name)
|
|
|
|
def _get_last_update(self):
|
|
"""
|
|
|
|
:return:
|
|
:rtype: datetime.datetime
|
|
"""
|
|
my_db = self.get_db()
|
|
sql_result = my_db.select('SELECT time FROM lastUpdate WHERE provider = ?', [self.providerID])
|
|
|
|
if sql_result:
|
|
last_time = int(sql_result[0]['time'])
|
|
if last_time > SGDatetime.timestamp_near():
|
|
last_time = 0
|
|
else:
|
|
last_time = 0
|
|
|
|
return datetime.datetime.fromtimestamp(last_time)
|
|
|
|
def _get_last_search(self):
|
|
"""
|
|
|
|
:return:
|
|
:rtype: datetime.datetime
|
|
"""
|
|
my_db = self.get_db()
|
|
sql_result = my_db.select('SELECT time FROM lastSearch WHERE provider = ?', [self.providerID])
|
|
|
|
if sql_result:
|
|
last_time = int(sql_result[0]['time'])
|
|
if last_time > SGDatetime.timestamp_near():
|
|
last_time = 0
|
|
else:
|
|
last_time = 0
|
|
|
|
return datetime.datetime.fromtimestamp(last_time)
|
|
|
|
def set_last_update(self, to_date=None):
|
|
"""
|
|
|
|
:param to_date: date time
|
|
:type to_date: datetime.datetime or None
|
|
"""
|
|
if not to_date:
|
|
to_date = datetime.datetime.now()
|
|
|
|
my_db = self.get_db()
|
|
my_db.upsert('lastUpdate',
|
|
{'time': int(time.mktime(to_date.timetuple()))},
|
|
{'provider': self.providerID})
|
|
|
|
def _set_last_search(self, to_date=None):
|
|
"""
|
|
|
|
:param to_date: date time
|
|
:type to_date: datetime.datetime or None
|
|
"""
|
|
if not to_date:
|
|
to_date = datetime.datetime.now()
|
|
|
|
my_db = self.get_db()
|
|
my_db.upsert('lastSearch',
|
|
{'time': int(time.mktime(to_date.timetuple()))},
|
|
{'provider': self.providerID})
|
|
|
|
last_update = property(_get_last_update)
|
|
last_search = property(_get_last_search)
|
|
|
|
def should_update(self):
|
|
"""
|
|
|
|
:return:
|
|
:rtype: bool
|
|
"""
|
|
# if we've updated recently then skip the update
|
|
return datetime.datetime.now() - self.last_update >= datetime.timedelta(minutes=self.update_iv)
|
|
|
|
def should_clear_cache(self):
|
|
"""
|
|
|
|
:return:
|
|
:rtype: bool
|
|
"""
|
|
# if recent search hasn't used our previous results yet then don't clear the cache
|
|
return self.last_search >= self.last_update
|
|
|
|
def add_cache_entry(self,
|
|
name, # type: AnyStr
|
|
url, # type: AnyStr
|
|
parse_result=None, # type: ParseResult
|
|
tvid_prodid=None # type: Union[AnyStr, None]
|
|
): # type: (...) -> Union[List[AnyStr, List[Any]], None]
|
|
"""
|
|
|
|
:param name: name
|
|
:param url: url
|
|
:param parse_result: parse result
|
|
:param tvid_prodid: tvid_prodid
|
|
:return:
|
|
"""
|
|
# check if we passed in a parsed result or should we try and create one
|
|
if not parse_result:
|
|
|
|
# create show_obj from tvid_prodid if available
|
|
show_obj = None
|
|
if tvid_prodid:
|
|
try:
|
|
show_obj = helpers.find_show_by_id(tvid_prodid, no_mapped_ids=False, check_multishow=True)
|
|
except MultipleShowObjectsException:
|
|
return
|
|
|
|
try:
|
|
parser = NameParser(show_obj=show_obj, convert=True, indexer_lookup=False)
|
|
parse_result = parser.parse(name)
|
|
except InvalidNameException:
|
|
logger.debug('Unable to parse the filename %s into a valid episode' % name)
|
|
return
|
|
except InvalidShowException:
|
|
return
|
|
|
|
if not parse_result or not parse_result.series_name:
|
|
return
|
|
|
|
if None is show_obj and parse_result.show_obj.is_anime:
|
|
t_show_obj = helpers.get_show(parse_result.show_obj.name, True)
|
|
post_parser = NameParser(False, show_obj=t_show_obj, convert=True, indexer_lookup=False)
|
|
try:
|
|
parse_result = post_parser.parse(name, release_group=self.get_id())
|
|
except(BaseException, Exception):
|
|
return
|
|
|
|
# if we made it this far then lets add the parsed result to cache for usage later on
|
|
season_number = parse_result.season_number if parse_result.season_number else 1
|
|
episode_numbers = parse_result.episode_numbers
|
|
|
|
if season_number and episode_numbers:
|
|
# store episodes as a separated string
|
|
episode_text = '|%s|' % '|'.join(map(str, episode_numbers))
|
|
|
|
# get the current timestamp
|
|
cur_timestamp = SGDatetime.timestamp_near()
|
|
|
|
# get quality of release
|
|
quality = parse_result.quality
|
|
|
|
# get release group
|
|
release_group = parse_result.release_group
|
|
|
|
# get version
|
|
version = parse_result.version
|
|
|
|
logger.debug('Add to cache: [%s]' % name)
|
|
|
|
return [
|
|
'INSERT OR IGNORE INTO provider_cache'
|
|
' (provider, name, season, episodes,'
|
|
' indexerid,'
|
|
' url, time, quality, release_group, version,'
|
|
' indexer)'
|
|
' VALUES (?,?,?,?,?,?,?,?,?,?,?)',
|
|
[self.providerID, name, season_number, episode_text,
|
|
parse_result.show_obj.prodid,
|
|
url, cur_timestamp, quality, release_group, version,
|
|
parse_result.show_obj.tvid]]
|
|
|
|
def search_cache(self,
|
|
episode, # type: TVEpisode
|
|
manual_search=False # type: bool
|
|
): # type: (...) -> List[SearchResult]
|
|
"""
|
|
|
|
:param episode: episode object
|
|
:param manual_search: manual search
|
|
:return: found results or empty List
|
|
"""
|
|
needed_eps = self.find_needed_episodes(episode, manual_search)
|
|
if 0 != len(needed_eps):
|
|
return needed_eps[episode]
|
|
return []
|
|
|
|
def list_propers(self, date=None):
|
|
"""
|
|
|
|
:param date: date
|
|
:type date: datetime.date
|
|
:return:
|
|
:rtype:
|
|
"""
|
|
my_db = self.get_db()
|
|
sql = "SELECT * FROM provider_cache WHERE name LIKE '%.PROPER.%' OR name LIKE '%.REPACK.%' " \
|
|
"OR name LIKE '%.REAL.%' AND provider = ?"
|
|
|
|
if date:
|
|
sql += ' AND time >= ' + str(int(time.mktime(date.timetuple())))
|
|
|
|
return list(filter(lambda x: x['indexerid'] != 0, my_db.select(sql, [self.providerID])))
|
|
|
|
def find_needed_episodes(self, ep_obj_list, manual_search=False):
|
|
# type: (Union[TVEpisode, List[TVEpisode]], bool) -> Dict[TVEpisode, SearchResult]
|
|
"""
|
|
|
|
:param ep_obj_list: episode object or list of episode objects
|
|
:param manual_search: manual search
|
|
"""
|
|
needed_eps = {}
|
|
cl = []
|
|
|
|
my_db = self.get_db()
|
|
if type(ep_obj_list) != list:
|
|
ep_obj_list = [ep_obj_list]
|
|
|
|
for ep_obj in ep_obj_list:
|
|
cl.append([
|
|
'SELECT *'
|
|
+ ' FROM provider_cache'
|
|
+ ' WHERE provider = ?'
|
|
+ ' AND indexer = ? AND indexerid = ?'
|
|
+ ' AND season = ? AND episodes LIKE ?'
|
|
+ ' AND quality IN (%s)' % ','.join([str(x) for x in ep_obj.wanted_quality]),
|
|
[self.providerID,
|
|
ep_obj.show_obj.tvid, ep_obj.show_obj.prodid,
|
|
ep_obj.season, '%|' + str(ep_obj.episode) + '|%']])
|
|
sql_result = my_db.mass_action(cl)
|
|
if sql_result:
|
|
sql_result = list(itertools.chain(*sql_result))
|
|
|
|
if not sql_result:
|
|
self._set_last_search()
|
|
return needed_eps
|
|
|
|
# for each cache entry
|
|
for cur_result in sql_result:
|
|
|
|
# get the show object, or if it's not one of our shows then ignore it
|
|
show_obj = helpers.find_show_by_id({int(cur_result['indexer']): int(cur_result['indexerid'])})
|
|
if not show_obj:
|
|
continue
|
|
|
|
# skip non-tv crap
|
|
if not show_name_helpers.pass_wordlist_checks(cur_result['name'], parse=False, indexer_lookup=False,
|
|
show_obj=show_obj):
|
|
continue
|
|
|
|
# skip if provider is anime only and show is not anime
|
|
if self.provider.anime_only and not show_obj.is_anime:
|
|
logger.debug(f'{show_obj.unique_name} is not an anime, skipping')
|
|
continue
|
|
|
|
# get season and ep data (ignoring multi-eps for now)
|
|
season = int(cur_result['season'])
|
|
if -1 == season:
|
|
continue
|
|
ep_obj_list = cur_result['episodes'].split('|')[1]
|
|
if not ep_obj_list:
|
|
continue
|
|
ep_obj_list = int(ep_obj_list)
|
|
|
|
quality = int(cur_result['quality'])
|
|
release_group = cur_result['release_group']
|
|
version = cur_result['version']
|
|
|
|
# if the show says we want that episode then add it to the list
|
|
if not show_obj.want_episode(season, ep_obj_list, quality, manual_search):
|
|
logger.debug(f"Skipping {cur_result['name']}"
|
|
f" because we don't want an episode that's {Quality.qualityStrings[quality]}")
|
|
continue
|
|
|
|
ep_obj = show_obj.get_episode(season, ep_obj_list)
|
|
|
|
# build a result object
|
|
title = cur_result['name']
|
|
url = cur_result['url']
|
|
|
|
logger.log(f'Found result {title} at {url}')
|
|
|
|
result = self.provider.get_result([ep_obj], url)
|
|
if None is result:
|
|
continue
|
|
result.show_obj = show_obj
|
|
result.name = title
|
|
result.quality = quality
|
|
result.release_group = release_group
|
|
result.version = version
|
|
result.content = None
|
|
np = NameParser(False, show_obj=show_obj)
|
|
try:
|
|
parsed_result = np.parse(title)
|
|
extra_info_no_name = parsed_result.extra_info_no_name()
|
|
version = parsed_result.version
|
|
is_anime = parsed_result.is_anime
|
|
except (BaseException, Exception):
|
|
extra_info_no_name = None
|
|
version = -1
|
|
is_anime = False
|
|
result.is_repack, result.properlevel = Quality.get_proper_level(extra_info_no_name, version, is_anime,
|
|
check_is_repack=True)
|
|
|
|
# add it to the list
|
|
if ep_obj not in needed_eps:
|
|
needed_eps[ep_obj] = [result]
|
|
else:
|
|
needed_eps[ep_obj].append(result)
|
|
|
|
# datetime stamp this search so cache gets cleared
|
|
self._set_last_search()
|
|
|
|
return needed_eps
|