# # 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 . from __future__ import with_statement import random import weakref from collections import Counter, OrderedDict from functools import reduce from itertools import chain from weakref import ReferenceType import datetime import glob import inspect import os.path import re import requests import stat import threading import traceback from imdbpie import ImdbAPIError from lxml_etree import etree # noinspection PyPep8Naming import encodingKludge as ek import exceptions_helper from exceptions_helper import ex import sickgear from . import db, helpers, history, image_cache, indexermapper, logger, \ name_cache, network_timezones, notifiers, postProcessor, subtitles from .anime import AniGroupList from .classes import weakList from .common import Quality, statusStrings, \ ARCHIVED, DOWNLOADED, FAILED, IGNORED, SKIPPED, SNATCHED, SNATCHED_ANY, SNATCHED_PROPER, UNAIRED, UNKNOWN, WANTED, \ NAMING_DUPLICATE, NAMING_EXTEND, NAMING_LIMITED_EXTEND, NAMING_LIMITED_EXTEND_E_PREFIXED, NAMING_SEPARATED_REPEAT from .generic_queue import QueuePriorities from .helpers import try_float, try_int from .indexermapper import del_mapping, MapStatus, save_mapping from .indexers.indexer_config import TVINFO_IMDB, TVINFO_TMDB, TVINFO_TRAKT, TVINFO_TVDB, TVINFO_TVMAZE, TVINFO_TVRAGE from .name_parser.parser import InvalidNameException, InvalidShowException, NameParser from .sgdatetime import SGDatetime, timestamp_near from .tv_base import TVEpisodeBase, TVShowBase from lib import imdbpie, subliminal from lib.dateutil import tz from lib.dateutil.parser import parser as du_parser from lib.fuzzywuzzy import fuzz from lib.tvinfo_base import RoleTypes, TVINFO_FACEBOOK, TVINFO_INSTAGRAM, TVINFO_SLUG, TVINFO_TWITTER, TVINFO_WIKIPEDIA from lib.tvinfo_base.exceptions import * from sg_helpers import calc_age, int_to_time, remove_file_perm, time_to_int from _23 import filter_iter, filter_list, list_keys from six import integer_types, iteritems, itervalues, moves, PY2, string_types # noinspection PyUnreachableCode if False: from typing import Any, AnyStr, Dict, List, Optional, Set, Text, Tuple, Union from sqlite3 import Row from lib.tvinfo_base import CastList, Character as TVINFO_Character, Person as TVINFO_Person, \ TVInfoEpisode, TVInfoShow coreid_warnings = False if coreid_warnings: import warnings warnings.simplefilter('always', DeprecationWarning) tz_p = du_parser() invalid_date_limit = datetime.date(1900, 1, 1) # status codes for switching tv show source TVSWITCH_DUPLICATE_SHOW = 0 TVSWITCH_ID_CONFLICT = 1 TVSWITCH_NO_NEW_ID = 2 TVSWITCH_NORMAL = 3 TVSWITCH_NOT_FOUND_ERROR = 4 TVSWITCH_SAME_ID = 5 TVSWITCH_SOURCE_NOT_FOUND_ERROR = 6 TVSWITCH_VERIFY_ERROR = 7 tvswitch_names = { TVSWITCH_DUPLICATE_SHOW: 'Duplicate Show', TVSWITCH_ID_CONFLICT: 'new id conflicts with existing show', TVSWITCH_NO_NEW_ID: 'no new id found', TVSWITCH_NORMAL: 'Normal', TVSWITCH_NOT_FOUND_ERROR: 'Not found on new tv source', TVSWITCH_SAME_ID: 'unchanged ids given', TVSWITCH_SOURCE_NOT_FOUND_ERROR: 'source show not found', TVSWITCH_VERIFY_ERROR: 'Verification error', } TVSWITCH_EP_DELETED = 1 TVSWITCH_EP_RENAMED = 2 tvswitch_ep_names = { TVSWITCH_EP_DELETED: 'deleted', TVSWITCH_EP_RENAMED: 'renamed' } concurrent_show_not_found_days = 7 show_not_found_retry_days = 7 prodid_bitshift = 4 tvid_bitmask = (1 << prodid_bitshift) - 1 class TVidProdid(object): """ Helper class to standardise the use of a TV info source id with its associated TV show id Examples of setting and using on the one action, tvid and prodid are numbers of type int or string; TVidProdid({tvid: prodid}) TVidProdid({tvid: prodid})([]) TVidProdid({tvid: prodid})(list) TVidProdid({tvid: prodid})('list') TVidProdid({tvid: prodid}).list Whitespace may exist between the number value and in these strings TVidProdid('tvid prodid')({}) TVidProdid('tvid prodid')(dict) TVidProdid('tvid prodid')('dict') TVidProdid('tvid prodid').dict """ glue = ':' def __init__(self, tvid_prodid=None): """ :param tvid_prodid: TV info data source ID, and TV show ID :type tvid_prodid: Dict or String """ self.tvid = None self.prodid = None self.sid_int = None if isinstance(tvid_prodid, dict) and 1 == len(tvid_prodid): try: for (tvid, prodid) in iteritems(tvid_prodid): self.tvid, self.prodid = int(tvid), int(prodid) except ValueError: pass elif isinstance(tvid_prodid, string_types): if self.glue in tvid_prodid: try: for (tvid, prodid) in [re.findall(r'(\d+)\s*%s\s*(\d+)' % self.glue, tvid_prodid)[0]]: self.tvid, self.prodid = int(tvid), int(prodid) except IndexError: pass else: try: legacy_showid = int(re.findall(r'(?i)\d+', tvid_prodid)[0]) show_obj = (helpers.find_show_by_id({TVINFO_TVDB: legacy_showid}) or helpers.find_show_by_id({TVINFO_TVRAGE: legacy_showid})) pre_msg = 'Todo: Deprecate show_id used without a tvid' if None is show_obj: pre_msg += ' and show_obj not found' else: self.tvid, self.prodid = show_obj.tvid, legacy_showid if coreid_warnings: logger.log('%s\n' % pre_msg + '|>%s^-- Note: Bootstrap & Tornado startup functions stripped from traceback log.' % '|>'.join(filter_iter(lambda text: not re.search(r'(?i)bootstrap|traceback\.' r'format_stack|pydevd|tornado' r'|webserveinit', text), traceback.format_stack(inspect.currentframe())))) except IndexError: pass elif isinstance(tvid_prodid, integer_types): self.tvid = tvid_prodid & tvid_bitmask self.prodid = tvid_prodid >> prodid_bitshift self.sid_int = tvid_prodid return if None not in (self.prodid, self.tvid): self.sid_int = self.prodid << prodid_bitshift | self.tvid def __call__(self, *args, **kwargs): return self._get(*args, **kwargs) def __repr__(self): return self._get() def __str__(self): return self._get() def __tuple__(self): return self._get(tuple) def __int__(self): return self._get(int) @property def list(self): return self._get(list) @property def dict(self): return self._get(dict) @property def tuple(self): return self._get(tuple) @property def int(self): return self._get(int) @staticmethod def _checktype(value, t): if isinstance(value, string_types) and not isinstance(value, type) and value: if value == getattr(t, '__name__', None): return True elif (isinstance(value, type) and value == t) or isinstance(value, t): return True return False def _get(self, kind=None): if None is not kind: if self._checktype(kind, int): return self.sid_int elif self._checktype(kind, dict): return {self.tvid: self.prodid} elif self._checktype(kind, tuple): return self.tvid, self.prodid if None is kind or self._checktype(kind, string_types): return '%s%s%s' % (self.tvid, self.glue, self.prodid) return [self.tvid, self.prodid] @property def _tvid(self): return self.tvid @property def _prodid(self): return self.prodid class PersonGenders(object): UNKNOWN = 0 male = 1 female = 2 possible_values = [0, 1, 2] names = {UNKNOWN: 'Unknown', male: 'Male', female: 'Female'} tmdb_map = {0: UNKNOWN, 1: female, 2: male} def _make_date(dt): # type: (datetime.date) -> integer_types if not dt: return None return dt.toordinal() def usable_id(value): # type: (Union[AnyStr, int]) -> Optional[AnyStr, int] """ return value as int if value is numerical, or value as string if value is valid id:format otherwise None """ value_id = try_int(value, None) if None is not value_id: return value_id return usable_rid(value) def usable_rid(value): # type: (Union[AnyStr]) -> Optional[AnyStr] """ return value if is a id:format is valid otherwise None if value fails basic id format validation """ if isinstance(value, string_types) and ':' in value: temp = value.split(':') if 2 == len(temp) and None not in [try_int(_x, None) for _x in temp]: return value class Referential(object): """ This class superimposes handling for a string based ID named `ref_id`. __init__ will pass back control untouched to the object deriving from this one if an integer is passed so that integer ID can perform """ def __init__(self, sid=None): self.id = None # type: integer_types self.ids = {} # type: Dict[int, integer_types] self._rid = usable_rid(sid) def has_ref_id(self, item): if isinstance(item, integer_types): return self.id == item tvid, src_id = self.ref_id(rid=item, string=False) return self.ids.get(tvid) == src_id def ref_id(self, rid=None, # type: Optional[AnyStr, int] string=True # type: Optional[bool] ): # type: (...) -> Union[tuple[Optional[AnyStr, int], Optional[AnyStr, int]], AnyStr] """ return either, 1) a preferred external unique id for a consistent reliable `reference`, or the internal row id as fallback 2) convert a consistent `reference` to src and src_id to use for matching to an internal row id uses param rid or self._rid as reference id to convert if either is not Nonetype for example, 1) if string is False, preferred static tuple[tvid, id], or tuple[None, id] if ids not set 2) otherwise, preferred static 'tvid:id', or 'id' if ids not set 3) if string is False and self._rid contains ':', list[src, src_id] 4) otherwise if self._rid contains ':', list['src', 'src_id'] reason, a row id is highly volatile, but a preferred ID is more reliable. use cases, show removed and readded, or src switched, then refreshing any character/person views will be consistent. can share/save a character/person link/bookmark as it will be consistent across instances. cast glide startAt is unreliable using glide index or row id. """ # if self._rid is a temporary type string containing ':' (use case during init) # return it as a tuple[int, int] -> src, src_id rid = rid or self._rid if usable_rid(rid): self._rid = None # consume once parts = [try_int(_x, None) for _x in rid.split(':')] # if string: no usage for this case # ? return rid # ? return '%s' % parts[0], '%s' % parts[1] return parts[0], parts[1] # get an available id from an order of preferred ids for use if self.ids: for cur_tvid in [TVINFO_IMDB, TVINFO_TVMAZE, TVINFO_TMDB, TVINFO_TRAKT]: if self.ids.get(cur_tvid): if string: return '%s:%s' % (cur_tvid, self.ids.get(cur_tvid)) return cur_tvid, self.ids.get(cur_tvid) if string: return '%s' % self.id return None, self.id # default internal id has no slug class Person(Referential): def __init__( self, name=None, # type: AnyStr gender=PersonGenders.UNKNOWN, # type: int birthday=None, # type: datetime.date deathday=None, # type: datetime.date bio=None, # type: AnyStr ids=None, # type: Dict[int, integer_types] sid=None, # type: integer_types birthplace=None, # type: AnyStr homepage=None, # type: AnyStr image_url=None, # type: AnyStr thumb_url=None, # type: AnyStr show_obj=None, # type: TVShow updated=1, # type: integer_types deathplace=None, # type: AnyStr height=None, # type: Union[integer_types, float] real_name=None, # type: AnyStr nicknames=None, # type: Set[AnyStr] akas=None, # type: Set[AnyStr] character_obj=None, # type: Character tmp_character_obj=None # type: Character ): # type: (...) -> Person super(Person, self).__init__(sid) self.updated = updated # type: integer_types self._data_failure = False # type: bool self._data_fetched = False # type: bool self.name = name # type: AnyStr self.character_obj = character_obj # type: Optional[Character] self._tmp_character_obj = tmp_character_obj # type: Optional[Character] self.gender = (PersonGenders.UNKNOWN, gender)[gender in PersonGenders.possible_values] # type: int self.birthday = birthday # type: Optional[datetime.date] self.deathday = deathday # type: Optional[datetime.date] self.biography = bio # type: Optional[AnyStr] self.birthplace = birthplace # type: Optional[AnyStr] self.homepage = homepage # type: Optional[AnyStr] self.image_url = image_url # type: Optional[AnyStr] self.thumb_url = thumb_url # type: Optional[AnyStr] self.deathplace = deathplace # type: Optional[AnyStr] self.height = height # type: Optional[Union[integer_types, float]] self.real_name = real_name # type: Optional[AnyStr] self.nicknames = nicknames or set() # type: Set[AnyStr] self.akas = akas or set() # type: Set[AnyStr] self.ids = ids or {} # type: Dict[int, integer_types] if not self._rid: self.id = sid or self._get_sid(show_obj=show_obj) # type: integer_types new_dirty = not sid self.dirty_main = new_dirty # type: bool self.dirty_ids = new_dirty # type: bool if not sid and self.id: fetched = self._data_fetched cur_data = self._remember_properties() self.load_from_db() cur_data['ids'] = dict(chain.from_iterable(iteritems(d) for d in (self.ids, ids or {}))) self.update_properties(**cur_data) self._data_fetched = fetched elif not self.name: self.load_from_db() @staticmethod def _order_names(name_set): # type: (Set[AnyStr]) -> List[AnyStr] rc_aka = re.compile(r'[\x00-\x7f]+') aka_all_ascii = [] aka_non_ascii = [] for cur_aka in name_set or []: if rc_aka.match(cur_aka): aka_all_ascii += [cur_aka] else: aka_non_ascii += [cur_aka] return aka_all_ascii + aka_non_ascii @property def lang_ordered_akas(self): # type: (...) -> List[AnyStr] return self._order_names(self.akas) @property def lang_ordered_nicknames(self): # type: (...) -> List[AnyStr] return self._order_names(self.nicknames) def _remember_properties(self): # type: (...) -> Dict return {k: self.__dict__[k] for k in ['akas', 'biography', 'birthday', 'birthplace', 'deathday', 'deathplace', 'gender', 'height', 'homepage', 'ids', 'image_url', 'name', 'nicknames', 'real_name', 'thumb_url']} def reset(self, person_obj=None): # type: (TVINFO_Person) -> None """ reset all properties with the exception of: name, id, ids :param person_obj: TVInfo Person object to reset to """ self._data_failure = False self._data_fetched = False self.akas = person_obj.akas or set() self.biography = person_obj.bio or None self.birthday = person_obj.birthdate or None self.birthplace = person_obj.birthplace or None self.deathday = person_obj.deathdate or None self.deathplace = person_obj.deathplace or None self.dirty_main = True self.gender = person_obj.gender or PersonGenders.UNKNOWN self.height = person_obj.height or None self.homepage = person_obj.homepage or None self.image_url = person_obj.image or None self.nicknames = person_obj.nicknames or set() self.real_name = person_obj.real_name or None self.thumb_url = person_obj.thumb_url or None self.updated = 1 def _get_sid(self, mass_action_result=None, show_obj=None): # type: (List, TVShow) -> Optional[integer_types] if getattr(self, 'id', None): return self.id def _get_from_ids(): my_db = db.DBConnection() sql_result = my_db.select( """ SELECT person_id FROM person_ids WHERE %s """ % ' OR '.join(['( src = ? AND src_id = ? )'] * len(self.ids)), list(reduce(lambda a, b: a + b, iteritems(self.ids)))) for cur_id in sql_result or []: return cur_id['person_id'] if mass_action_result: try: return mass_action_result[-1:][0][0]['last_rowid'] except (BaseException, Exception): return elif self.ids: i = _get_from_ids() if i: return i self.get_missing_data(show_obj=show_obj) if self.ids: return _get_from_ids() def _set_updated(self): # type: (...) -> integer_types """ set and return new update date """ self.updated = datetime.date.today().toordinal() - (0, 120)[self._data_failure] return self.updated def remove_img(self): """ removes image of person """ try: remove_file_perm(image_cache.ImageCache().person_path(self)) except (BaseException, Exception): pass def remove_thumb(self): """ removes thumb image of person """ try: remove_file_perm(image_cache.ImageCache().person_thumb_path(self)) except (BaseException, Exception): pass def remove_all_img(self): """ remove all images of person """ try: for r in image_cache.ImageCache().person_both_paths(self): try: remove_file_perm(r) except (BaseException, Exception): pass except (BaseException, Exception): pass def update_properties(self, **kwargs): """ updates properties of class, if changed sets dirty :param kwargs: properties of class """ if (not self.image_url and not self.thumb_url) or (not self.image_url and kwargs.get('image_url')): if kwargs.get('image_url'): self.dirty_main = True self.remove_img() self.image_url = kwargs.get('image_url') # thumb changed, remove old image if kwargs.get('thumb_url'): self.remove_thumb() self.thumb_url = kwargs.get('thumb_url') else: self.remove_thumb() self.thumb_url = None elif kwargs.get('thumb_url'): self.dirty_main = True self.thumb_url = kwargs.get('thumb_url') # image changed remove old image self.remove_img() if kwargs.get('image_url'): self.remove_img() self.image_url = kwargs.get('image_url') else: self.remove_img() self.image_url = None for k, v in iteritems(kwargs): if k in ('image_url', 'thumb_url'): continue if k not in self.__dict__: raise Exception('Person has no property [%s]' % k) if None is not v: if 'akas' == k: v.update(self.akas) elif 'nicknames' == k: v.update(self.nicknames) if v != self.__dict__[k]: if 'biography' == k: if len(v) > len(self.biography or ''): self.biography = v self.dirty_main = True elif 'gender' == k and v not in (PersonGenders.male, PersonGenders.female): continue else: self.__dict__[k] = v if 'ids' == k: self.dirty_ids = True else: self.dirty_main = True def load_from_db(self): if self.id or usable_rid(self._rid): my_db = db.DBConnection() src, src_id = self.ref_id(string=False) sql_result = my_db.select( """ SELECT persons.*, (SELECT group_concat(person_ids.src || ':' || person_ids.src_id, ';;;') FROM person_ids WHERE person_ids.person_id = persons.id) AS p_ids FROM persons LEFT JOIN person_ids pi ON persons.id = pi.person_id WHERE %s """ % ('pi.src = ? and pi.src_id = ?', 'persons.id = ?')[not src], ([src, src_id], [self.id])[not src]) for cur_person in sql_result or []: birthdate = try_int(cur_person['birthdate'], None) birthdate = birthdate and datetime.date.fromordinal(cur_person['birthdate']) deathdate = try_int(cur_person['deathdate'], None) deathdate = deathdate and datetime.date.fromordinal(cur_person['deathdate']) p_ids = {} for cur_ids in (cur_person['p_ids'] and cur_person['p_ids'].split(';;;')) or []: k, v = cur_ids.split(':') k = try_int(k, None) if v and None is not k: p_ids[k] = v if k in (TVINFO_FACEBOOK, TVINFO_INSTAGRAM, TVINFO_TWITTER, TVINFO_WIKIPEDIA) \ else try_int(v, None) (self._data_failure, self._data_fetched, self.akas, self.biography, self.birthday, self.birthplace, self.deathday, self.deathplace, self.gender, self.height, self.homepage, self.id, self.ids, self.image_url, self.name, self.nicknames, self.real_name, self.thumb_url, self.updated) = \ (False, False, set((cur_person['akas'] and cur_person['akas'].split(';;;')) or []), cur_person['bio'], birthdate, cur_person['birthplace'], deathdate, cur_person['deathplace'], cur_person['gender'], cur_person['height'], cur_person['homepage'], cur_person['id'], p_ids, cur_person['image_url'], cur_person['name'], set((cur_person['nicknames'] and cur_person['nicknames'].split(';;;')) or []), cur_person['realname'], cur_person['thumb_url'], cur_person['updated']) break def update_prop_from_tvinfo_person(self, person_obj): # type: (TVINFO_Person) -> None """ update person with tvinfo person object info Note: doesn't change: name, id, image_url, thumb_url :param person_obj: TVInfo Person object """ if person_obj: char_obj = self.character_obj or self._tmp_character_obj if person_obj.characters and char_obj and char_obj.show_obj \ and char_obj.show_obj.ids.get(TVINFO_IMDB, {}).get('id'): p_char = [pc for pc in person_obj.characters if pc.show.ids.imdb == char_obj.show_obj.ids.get(TVINFO_IMDB, {}).get('id')] p_count = len(p_char) if 1 == p_count: if (p_char[0].start_year or p_char[0].end_year) and getattr(self, 'id', None): char_obj.combine_start_end_years( {self.id: {'start': p_char[0].start_year, 'end': p_char[0].end_year}}) elif 1 < p_count: start, end, found, p_p = None, None, False, None for c_p in p_char: if not start or start > (c_p.start_year or 3000): start = c_p.start_year if not end or end < (c_p.end_year or 0): end = c_p.end_year if c_p.name == char_obj.name: start = c_p.start_year end = c_p.end_year found = True break if not found: try: p_p = sorted([ (fuzz.UWRatio(char_obj.name, p.name), p.end_year and p.start_year and p.end_year - p.start_year, p) for p in p_char], key=lambda a: (a[0], a[1]), reverse=True)[0][2] except (BaseException, Exception): p_p = None if None is not p_p and (p_p.start_year or p_p.end_year): start = p_p.start_year end = p_p.end_year if (start or end) and getattr(self, 'id', None): char_obj.combine_start_end_years({self.id: {'start': start, 'end': end}}) self.update_properties( gender=person_obj.gender, homepage=person_obj.homepage, ids=dict(chain.from_iterable( iteritems(d) for d in (self.ids, {k: v for k, v in iteritems(person_obj.ids) if v and TVINFO_SLUG != k} or {}))), birthday=person_obj.birthdate, deathday=person_obj.deathdate, biography=person_obj.bio, birthplace=person_obj.birthplace, deathplace=person_obj.deathplace, height=person_obj.height, real_name=person_obj.real_name, nicknames=person_obj.nicknames, akas=person_obj.akas ) def get_missing_data(self, show_obj=None, character_obj=None, stop_event=None, force_id=False): # type: (TVShow, Character, threading.Event, bool) -> None """ load missing data for person from trakt + tmdb :param show_obj: :param character_obj: :param stop_event: :param force_id: refetch all external ids, normally only source ids and missing are fetched """ if character_obj and not self.character_obj: self.character_obj = character_obj if self._data_failure: self.dirty_main = True self._data_failure = False self._data_fetched = True tvsrc_result, found_persons, found_on_src, search_sources, found_ids, ids_to_check, tv_loop, imdb_confirmed, \ source_confirmed = \ None, {}, set(), [TVINFO_TRAKT, TVINFO_TMDB, TVINFO_IMDB], \ set([k for k, v in iteritems(self.ids) if v] + ['text']), {}, 0, False, {} # confirmed_character = False max_search_src = len(search_sources) logger.debug('Getting extra data for: %s' % self.name) for tv_loop in moves.xrange(0, max_search_src): search_sources = [s for s in search_sources if s not in found_on_src] for cur_tv_info_src in search_sources: if stop_event and stop_event.is_set(): return rp, confirmed_on_src = None, False tvinfo_config = sickgear.TVInfoAPI(cur_tv_info_src).api_params.copy() t = sickgear.TVInfoAPI(cur_tv_info_src).setup(**tvinfo_config) if 0 == tv_loop or cur_tv_info_src not in ids_to_check: ids_to_check[cur_tv_info_src] = t.supported_person_id_searches + ['text'] found_persons[cur_tv_info_src] = set() for tv_src in ids_to_check[cur_tv_info_src]: if 'text' != tv_src and not self.ids.get(tv_src): continue try: if 'text' != tv_src: kw = {'ids': {tv_src: self.ids[tv_src]}} else: kw = {'name': self.name} tvsrc_result = t.search_person(**kw) except (BaseException, Exception): self._data_failure = True continue if tvsrc_result: # verify we have the correct person for cur_person in tvsrc_result: # type: TVINFO_Person if None is not rp: break if not (imdb_confirmed and TVINFO_IMDB == tv_src) \ and cur_person.id in found_persons[cur_tv_info_src]: continue found_persons[cur_tv_info_src].add(cur_person.id) try: pd = t.get_person(cur_person.ids.get(cur_tv_info_src), get_show_credits=True, get_images=True) except (BaseException, Exception) as e: self._data_failure = True logger.warning('Error searching extra info for person: %s - %s' % (self.name, ex(e))) continue if None is not pd and imdb_confirmed and TVINFO_IMDB == tv_src: rp = pd break # noinspection PyUnresolvedReferences if show_obj and None is not pd and pd.characters: clean_show_name = indexermapper.clean_show_name(show_obj.name.lower()) for ch in pd.characters or []: # type: TVINFO_Character if clean_show_name == indexermapper.clean_show_name(ch.show.seriesname.lower()): rp = pd confirmed_on_src = True # confirmed_character = True break elif any(t_src == s_src and bool(t_id) and t_id == s_id['id'] for t_src, t_id in iteritems(ch.show.ids) for s_src, s_id in iteritems(show_obj.ids)): rp = pd confirmed_on_src = True # confirmed_character = True break if rp: if TVINFO_IMDB == cur_tv_info_src and confirmed_on_src: imdb_confirmed = True found_on_src.add(cur_tv_info_src) break ids_to_check[cur_tv_info_src] = [i for i in ids_to_check[cur_tv_info_src] if i not in found_ids] if None is not rp: if confirmed_on_src: for i in (TVINFO_TRAKT, TVINFO_IMDB, TVINFO_TMDB, TVINFO_TVMAZE, TVINFO_TVDB): # in case it's the current source use it's id and lock if from being changed if cur_tv_info_src == i and rp.ids.get(i): source_confirmed[i] = True if rp.ids.get(i) != self.ids.get(i): self.ids[i] = rp.ids[i] self.dirty_ids = True if not source_confirmed.get(i) and \ (rp.ids.get(i) and not self.ids.get(i) or (force_id and rp.ids.get(i) and rp.ids.get(i) != self.ids.get(i))): self.ids[i] = rp.ids[i] found_ids.add(i) self.dirty_ids = True for i in (TVINFO_INSTAGRAM, TVINFO_TWITTER, TVINFO_FACEBOOK, TVINFO_WIKIPEDIA): if rp.social_ids.get(i) and not self.ids.get(i) or \ (rp.social_ids.get(i) and rp.social_ids.get(i) != self.ids.get(i)): self.ids[i] = rp.social_ids[i] found_ids.add(i) self.dirty_ids = True self.update_prop_from_tvinfo_person(rp) if not self.image_url and rp.image: self.image_url = rp.image if rp.thumb_url: self.thumb_url = rp.thumb_url else: self.remove_thumb() self.thumb_url = None self.dirty_main = True elif not self.thumb_url and rp.thumb_url: self.thumb_url = rp.thumb_url if not rp.image: self.remove_img() self.image_url = None self.dirty_main = True def save_to_db(self, show_obj=None, character_obj=None, force=False, stop_event=None): # type: (TVShow, Character, bool, threading.Event) -> None if not self._data_fetched: self.get_missing_data(show_obj=show_obj, character_obj=character_obj, stop_event=stop_event) self._data_fetched = False if not any(d for d in (self.dirty_main, self.dirty_ids)): return my_db = db.DBConnection() if not self.id: cl = [[ """ INSERT INTO persons (akas, bio, birthdate, birthplace, deathdate, deathplace, gender, height, homepage, image_url, name, nicknames, realname, thumb_url, updated) VALUES (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?); """, [';;;'.join(self.akas or []), self.biography, _make_date(self.birthday), self.birthplace, _make_date(self.deathday), self.deathplace, self.gender, self.height, self.homepage, self.image_url, self.name, ';;;'.join(self.nicknames or []), self.real_name, self.thumb_url, self._set_updated()] ]] c_ids = len(self.ids) if c_ids: f_d = next(iter(self.ids)) cl.extend([[ """ INSERT INTO person_ids (src, src_id, person_id) VALUES (?,?,last_insert_rowid()) """, [f_d, self.ids[f_d]]] ]) if 1 < c_ids: cl.extend([[ """ INSERT INTO person_ids (src, src_id, person_id) VALUES (?,?, (SELECT person_ids.person_id FROM person_ids WHERE person_ids.id = last_insert_rowid()) ); """, [s, self.ids[s]]] for s in list(self.ids)[1:] ]) cl.extend([[ """ SELECT person_ids.person_id AS last_rowid FROM person_ids WHERE person_ids.id = last_insert_rowid() """] ]) else: cl.extend([['SELECT last_insert_rowid() as last_rowid;']]) else: cl = [] if force or self.dirty_main: cl = [[ """ UPDATE persons SET name = ?, gender = ?, birthdate = ?, deathdate = ?, bio = ?, birthplace = ?, homepage = ?, image_url = ?, thumb_url = ?, updated = ?, deathplace = ?, height = ?, realname = ?, nicknames = ?, akas = ? WHERE id = ? """, [self.name, self.gender, _make_date(self.birthday), _make_date(self.deathday), self.biography, self.birthplace, self.homepage, self.image_url, self.thumb_url, self._set_updated(), self.deathplace, self.height, self.real_name, ';;;'.join(self.nicknames or []), ';;;'.join(self.akas or []), self.id]], ] if force or self.dirty_ids: for s, v in iteritems(self.ids): cl.extend([ ['UPDATE person_ids SET src_id = ? WHERE person_id = ? AND src = ?', [v, self.id, s]], ["INSERT INTO person_ids (src, src_id, person_id) SELECT %s, '%s', %s WHERE changes() == 0" % (s, v, self.id)] ]) if cl: r_id = my_db.mass_action(cl) if r_id and r_id[-1:][0]: self.id = self._get_sid(r_id) self.dirty_main, self.dirty_ids = False, False def calc_age(self, date=None): # type: (Optional[datetime.date]) -> Optional[int] """ returns age based on current date or given date :param date: """ return calc_age(self.birthday, self.deathday, date) @property def age(self): # type: (...) -> Optional[int] """ :return: age of person if birthdate is known, in case of deathdate is known return age of death """ return self.calc_age() def __bool__(self): return bool(self.name) def __eq__(self, other): return (self.id not in [None, 0] and other.id == self.id) \ and any(self.ids[o] == v for o, v in iteritems(other.ids)) def __str__(self): lived, id_str, id_list = '', '', [] if self.birthday: lived += ' %s' % self.birthday if self.deathday: lived += ' - %s' % self.deathday for i, d in iteritems(self.ids): id_list.append('%s: %s' % (sickgear.TVInfoAPI(i).name, d)) if id_list: id_str = ' (%s)' % ', '.join(id_list) return '' % (self.name, lived, id_str) def __repr__(self): return self.__str__() __nonzero__ = __bool__ class Character(Referential): def __init__( self, name=None, # type: AnyStr person=None, # type: List[Person] ids=None, # type: Dict[int, integer_types] bio=None, # type: AnyStr sid=None, # type: integer_types image_url=None, # type: AnyStr thumb_url=None, # type: AnyStr show_obj=None, # type: TVShow updated=1, # type: integer_types persons_years=None, # type: Dict[integer_types, Dict[AnyStr, int]] tmp=False # type:bool ): # type: (...) -> Character super(Character, self).__init__(sid) self.updated = updated # type: integer_types self.name = name # type: AnyStr self.persons_years = persons_years or {} # type: Dict[integer_types, Dict[AnyStr, int]] self.person = person or [] # type: List[Person] self.biography = bio # type: Optional[AnyStr] self.ids = ids or {} # type: Dict[int, integer_types] self.image_url = image_url # type: Optional[AnyStr] self.thumb_url = thumb_url # type: Optional[AnyStr] self.show_obj = show_obj # type: TVShow if not tmp: if not self._rid: self.id = sid or self._get_sid() # type: integer_types new_dirty = not sid self.dirty_main = new_dirty # type: bool self.dirty_ids = new_dirty # type: bool self.dirty_person = new_dirty # type: bool self.dirty_years = new_dirty # type: bool if not sid and self.id: self.load_from_db() self.update_properties(name=name, person=person, biography=bio, ids=dict(chain.from_iterable(iteritems(d) for d in (self.ids, ids or {}))), image_url=image_url, thumb_url=thumb_url) elif not self.name: self.load_from_db() def _get_sid(self, mass_action_result=None): # type: (List) -> Optional[integer_types] if getattr(self, 'id', None): return self.id if mass_action_result: try: return mass_action_result[-1:][0][0]['last_rowid'] except (BaseException, Exception): return elif self.ids: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT character_id FROM character_ids WHERE %s """ % ' OR '.join(['( src = ? AND src_id = ? )'] * len(self.ids)), list(reduce(lambda a, b: a+b, iteritems(self.ids)))) if sql_result: return sql_result[0]['character_id'] elif self.person: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT characters.id AS c_id FROM characters LEFT JOIN character_person_map cpm ON characters.id = cpm.character_id WHERE name = ? AND cpm.person_id IN (%s) """ % ','.join(['?' * len(self.person)]), [self.name] + [p.id for p in self.person]) if sql_result: return sql_result[0]['c_id'] def _set_updated(self): # type: (...) -> integer_types """ set and return new update date """ self.updated = datetime.date.today().toordinal() return self.updated def remove_img(self): """ removes image of person """ try: remove_file_perm(image_cache.ImageCache().character_path(self, self.show_obj)) except (BaseException, Exception): pass def remove_thumb(self): """ removes thumb image of person """ try: remove_file_perm(image_cache.ImageCache().character_thumb_path(self, self.show_obj)) except (BaseException, Exception): pass def remove_all_img(self, include_person=False, tvid=None, proid=None): # type: (bool, integer_types, integer_types) -> None """ remove all images of person """ try: for r in image_cache.ImageCache().character_both_path(self, show_obj=self.show_obj, tvid=tvid, proid=proid): try: remove_file_perm(r) except (BaseException, Exception): pass except (BaseException, Exception): pass if include_person: for p in self.person or []: try: p.remove_all_img() except (BaseException, Exception): pass def update_properties(self, **kwargs): for k, v in iteritems(kwargs): if k not in self.__dict__: raise Exception('Character has no property [%s]' % k) if None is not v and v != self.__dict__[k]: if 'image_url' == k: self.dirty_main = True self.remove_img() if not kwargs.get('thumb_url'): self.remove_thumb() self.thumb_url = None elif 'thumb_url' == k: self.dirty_main = True self.remove_thumb() if not kwargs.get('image_url'): self.remove_img() self.image_url = None self.__dict__[k] = v if 'ids' == k: self.dirty_ids = True elif 'person' == k: self.dirty_person = True else: self.dirty_main = True def combine_start_end_years(self, new_years): # type: (Dict[integer_types, Dict[AnyStr, int]]) -> None new_dict = dict(chain.from_iterable( iteritems(d) for d in (self.persons_years, {k: v for k, v in iteritems(new_years) if v} or {}))) if new_dict != self.persons_years: self.persons_years = new_dict self.dirty_years = True def load_from_db(self): if self.id or usable_rid(self._rid): my_db = db.DBConnection() src, src_id = self.ref_id(string=False) sql_result = my_db.select( """ SELECT characters.name AS name, characters.bio AS bio, characters.thumb_url AS thumb_url, characters.image_url AS image_url, characters.updated AS c_updated, characters.id AS c_id, (SELECT group_concat(character_ids.src || ':' || character_ids.src_id, ';;;') FROM character_ids WHERE character_ids.character_id = characters.id) AS c_ids, (SELECT group_concat(character_person_map.person_id, ';;;') FROM character_person_map WHERE character_person_map.character_id = characters.id) AS person_ids, (SELECT group_concat(character_person_years.person_id || ':' || character_person_years.start_year || '-' || character_person_years.end_year, ';;;') FROM character_person_years WHERE character_person_years.character_id = characters.id) AS p_years FROM characters LEFT JOIN character_ids ci ON characters.id = ci.character_id WHERE %s """ % ('ci.src = ? and ci.src_id = ?', 'characters.id = ?')[not src], ([src, src_id], [self.id])[not src]) for cur_row in (sql_result or []): c_ids = {} for cur_ids in (cur_row['c_ids'] and cur_row['c_ids'].split(';;;')) or []: k, v = cur_ids.split(':') v = try_int(v, None) if v: c_ids[int(k)] = try_int(v, None) p_years = {} for cur_years in (cur_row['p_years'] and cur_row['p_years'].split(';;;')) or []: p_id, py = cur_years.split(':') start, end = py.split('-') p_years[int(p_id)] = {'start': try_int(start, None), 'end': try_int(end, None)} (self.biography, self.id, self.ids, self.image_url, self.name, self.person, self.persons_years, self.thumb_url, self.updated) = \ (cur_row['bio'], cur_row['c_id'], c_ids, cur_row['image_url'], cur_row['name'], [Person(sid=int(p), character_obj=self) for p in (cur_row['person_ids'] and cur_row['person_ids'].split(';;;')) or []], p_years, cur_row['thumb_url'], cur_row['c_updated']) self.dirty_main, self.dirty_ids, self.dirty_person = False, False, False break def _get_start_end_year(self): if self.person: tvinfo_config = sickgear.TVInfoAPI(TVINFO_IMDB).api_params.copy() t = sickgear.TVInfoAPI(TVINFO_IMDB).setup(**tvinfo_config) for cur_per in self.person: if not cur_per.id: continue if cur_per.ids.get(TVINFO_IMDB): try: res = t.get_person(cur_per.ids.get(TVINFO_IMDB), get_show_credits=True) except (BaseException, Exception): continue if res and res.characters: cur_per.character_obj = self cur_per.update_prop_from_tvinfo_person(res) def save_to_db(self, show_obj=None, force=False, stop_event=None): # type: (TVShow, bool, threading.Event) -> None if self.person: for cur_per in self.person: if stop_event and stop_event.is_set(): return if force or self.dirty_person or not cur_per.id or cur_per.dirty_main or cur_per.dirty_ids or \ 14 < (datetime.date.today().toordinal() - cur_per.updated): cur_per.save_to_db(show_obj=show_obj, character_obj=self, force=force, stop_event=stop_event) self.dirty_person = True if ((None is self.id and not self.persons_years) or self.dirty_main or 14 < datetime.date.today().toordinal() - self.updated) and self.person: self._get_start_end_year() my_db = db.DBConnection() if not self.id: cl = [[ """ INSERT INTO characters (bio, image_url, name, thumb_url, updated) VALUES (?,?,?,?,?) """, [self.biography, self.image_url, self.name, self.thumb_url, self._set_updated()]], ] c_ids = len(self.ids) if c_ids: f_d = next(iter(self.ids)) cl.extend([[ """ INSERT INTO character_ids (src, src_id, character_id) VALUES (?,?,last_insert_rowid()) """, [f_d, self.ids[f_d]]] ]) if 1 < c_ids: cl.extend([[ """ INSERT INTO character_ids (src, src_id, character_id) VALUES (?,?, (SELECT character_ids.character_id FROM character_ids WHERE character_ids.id = last_insert_rowid()) ); """, [s, self.ids[s]]] for s in list(self.ids)[1:] ]) cl.extend([[ """ SELECT character_ids.character_id AS last_rowid FROM character_ids WHERE character_ids.id = last_insert_rowid() """] ]) else: cl.extend([['SELECT last_insert_rowid() AS last_rowid;']]) else: cl = [] if force or self.dirty_main: cl = [[ """ UPDATE characters SET bio = ?, image_url = ?, name = ?, thumb_url = ?, updated = ? WHERE id = ? """, [self.biography, self.image_url, self.name, self.thumb_url, self._set_updated(), self.id]], ] if force or self.dirty_ids: for cur_tvid, cur_src_id in iteritems(self.ids): cl.extend([[ """ UPDATE character_ids SET src_id = ? WHERE src = ? AND character_id = ? """, [cur_src_id, cur_tvid, self.id] ], [ """ INSERT INTO character_ids (src, src_id, character_id) SELECT %s, %s, %s WHERE changes() == 0 """ % (cur_tvid, cur_src_id, self.id)] ]) # in case we don't have a character id yet, we need to fetch it for the next step if not self.id and cl: r_id = my_db.mass_action(cl) cl = [] if r_id and r_id[-1:][0]: # r_id = list(itertools.chain.from_iterable(r_id)) if r_id: self.id = self._get_sid(r_id) if (force or self.dirty_person) and self.person: cl.append([ """ DELETE FROM character_person_map WHERE character_id = ? AND person_id NOT IN (%s) """ % ','.join(['?'] * len(self.person)), [self.id] + [p.id for p in self.person]]) for cur_per in self.person: cl.extend([[ """ INSERT OR IGNORE INTO character_person_map (character_id, person_id) VALUES (?,?) """, [self.id, cur_per.id]] ]) if (force or self.dirty_years) and self.person: cl.append([ """ DELETE FROM character_person_years WHERE character_id = ? AND person_id NOT IN (%s) """ % ','.join(['?'] * len(self.person)), [self.id] + [p.id for p in self.person]]) for cur_per in self.person: if cur_per.id and any(1 for v in itervalues(self.persons_years.get(cur_per.id, {})) if v): p_years = self.persons_years.get(cur_per.id, {}) cl.append([ """ REPLACE INTO character_person_years (character_id, person_id, start_year, end_year) VALUES (?,?,?,?) """, [self.id, cur_per.id, p_years.get('start'), p_years.get('end')]] ) if cl: my_db.mass_action(cl) self.dirty_ids, self.dirty_main, self.dirty_person, self.dirty_years = False, False, False, False def __bool__(self): return bool(self.name) or bool(self.person) or bool(self.ids) def __eq__(self, other): return other.person == self.person and ((self.id not in [None, 0] and other.id == self.id) or any(self.ids[o] == v for o, v in iteritems(other.ids)) or (not other.ids and other.name == self.name)) def __str__(self): id_str, id_list = '', [] for i, d in iteritems(self.ids): id_list.append('%s: %s' % (sickgear.TVInfoAPI(i).name, d)) if id_list: id_str = ' (%s)' % ', '.join(id_list) return '' % (self.name, ', '.join(p.name for p in self.person), id_str) def __repr__(self): return self.__str__() __nonzero__ = __bool__ class TVShow(TVShowBase): __slots__ = ( 'path', 'unique_name', ) def __init__(self, tvid, prodid, lang='', show_result=None, imdb_info_result=None): # type: (int, int, Text, Optional[Row], Optional[Union[Row, Dict]]) -> None super(TVShow, self).__init__(tvid, prodid, lang) self.unique_name = '' self.tvid = int(tvid) self.prodid = int(prodid) self.sid_int = self.create_sid(self.tvid, self.prodid) if None is not helpers.find_show_by_id(self.sid_int, check_multishow=True): raise exceptions_helper.MultipleShowObjectsException('Can\'t create a show if it already exists') self._airtime = None # type: Optional[datetime.time] self._cast_list = None # type: Optional[ReferenceType[List[Character]]] self._last_found_on_indexer = -1 # type: int self._location = self.path = '' # type: AnyStr # self._is_location_good = None self._network_country = None # type: Optional[AnyStr] self._network_country_code = None # type: Optional[AnyStr] self._network_id = None # type: Optional[int] self._network_is_stream = None # type: Optional[bool] self._not_found_count = None # type: None or int self._paused = 0 self._src_update_time = None # type: Optional[integer_types] self.internal_ids = {} # type: Dict self.internal_timezone = None # type: Optional[AnyStr] self.lock = threading.RLock() self.nextaired = '' # type: AnyStr # noinspection added so that None _can_ be excluded from type annotation # so that this property evaluates directly to the class on ctrl+hover instead of "multiple implementations" # noinspection PyTypeChecker self.release_groups = None # type: AniGroupList self.sxe_ep_obj = {} # type: Dict self.load_from_db(show_result=show_result, imdb_info_result=imdb_info_result) def _get_end_episode(self, last=False, exclude_specials=False): # type: (bool, bool) -> Optional[TVEpisode] eps = self.get_all_episodes() if eps: ed = datetime.date(1900, 1, 1) return next(iter(sorted((ep for ep in eps if ep.airdate > ed and (not exclude_specials or ep.season != 0)), key=lambda a: a.airdate, reverse=last)), None) @property def first_aired_episode(self): # type: (...) -> Optional[TVEpisode] """ returns first aired episode """ return self._get_end_episode() @property def first_aired_regular_episode(self): # type: (...) -> Optional[TVEpisode] """ returns first aired regular episode """ return self._get_end_episode(exclude_specials=True) @property def latest_aired_episode(self): # type: (...) -> Optional[TVEpisode] """ returns latest aired episode """ return self._get_end_episode(last=True) @property def latest_aired_regular_episode(self): # type: (...) -> Optional[TVEpisode] """ returns latest aired regular episode """ return self._get_end_episode(last=True, exclude_specials=True) @property def cast_list(self): # type: (...) -> List[Character] if None is self._cast_list or (isinstance(self._cast_list, ReferenceType) and None is self._cast_list()): v = self._load_cast_from_db() self._cast_list = weakref.ref(v) return self._cast_list() return self._cast_list() @cast_list.setter def cast_list(self, value): # type: (weakList[Character]) -> None self._cast_list = None if not isinstance(value, weakList) else weakref.ref(value) @property def network_id(self): return self._network_id @network_id.setter def network_id(self, *args): self.dirty_setter('_network_id')(self, *args) @property def network_country(self): return self._network_country @network_country.setter def network_country(self, *args): self.dirty_setter('_network_country')(self, *args) @property def network_country_code(self): return self._network_country_code @network_country_code.setter def network_country_code(self, *args): self.dirty_setter('_network_country_code')(self, *args) @property def network_is_stream(self): return self._network_is_stream @network_is_stream.setter def network_is_stream(self, *args): self.dirty_setter('_network_is_stream')(self, *args) @property def airtime(self): return self._airtime @airtime.setter def airtime(self, *args): self.dirty_setter('_airtime')(self, *args) @property def timezone(self): return self.internal_timezone @timezone.setter def timezone(self, *args): self.dirty_setter('internal_timezone')(self, *args) @staticmethod def create_sid(tvid, prodid): # type: (int, int) -> int return int(prodid) << prodid_bitshift | int(tvid) @property def _tvid(self): return self.tvid @_tvid.setter def _tvid(self, val): self.dirty_setter('tvid')(self, int(val)) self.tvid_prodid = self.create_sid(val, self.prodid) # TODO: remove the following when indexer is gone # in deprecation transition, tvid also sets indexer so that existing uses continue to work normally self.dirty_setter('_indexer')(self, int(val)) @property def _prodid(self): return self.prodid @_prodid.setter def _prodid(self, val): # type: (int) -> None self.dirty_setter('prodid')(self, int(val)) self.tvid_prodid = self.create_sid(self.tvid, val) # TODO: remove the following when indexerid is gone # in deprecation transition, prodid also sets indexerid so that existing usages continue as normal self.dirty_setter('_indexerid')(self, int(val)) @property def tvid_prodid(self): # type: (...) -> AnyStr return TVidProdid({self.tvid: self.prodid})() @tvid_prodid.setter def tvid_prodid(self, val): tvid_prodid_obj = TVidProdid(val) if getattr(self, 'tvid_prodid', None) != tvid_prodid_obj(): self.tvid, self.prodid = tvid_prodid_obj.list self.dirty = True tvid_prodid_int = int(tvid_prodid_obj) if getattr(self, 'sid_int', None) != tvid_prodid_int: self.sid_int = tvid_prodid_int def helper_load_failed_db(self, sql_result=None): # type: (Optional[Row, Dict]) -> None if None is self._not_found_count or -1 == self._last_found_on_indexer: if sql_result and self.prodid == sql_result['indexer_id'] and self.tvid == sql_result['indexer']: sql_result = [sql_result] else: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT fail_count, last_success FROM tv_shows_not_found WHERE indexer = ? AND indexer_id = ? """, [self.tvid, self.prodid]) if sql_result: self._not_found_count = helpers.try_int(sql_result[0]['fail_count']) self._last_found_on_indexer = helpers.try_int(sql_result[0]['last_success']) else: self._not_found_count = 0 self._last_found_on_indexer = 0 @property def not_found_count(self): self.helper_load_failed_db() return self._not_found_count @not_found_count.setter def not_found_count(self, v): if isinstance(v, integer_types) and v != self._not_found_count: self._last_found_on_indexer = self.last_found_on_indexer my_db = db.DBConnection() # noinspection PyUnresolvedReferences last_check = int(timestamp_near(datetime.datetime.now())) # in case of flag change (+/-) don't change last_check date if abs(v) == abs(self._not_found_count): sql_result = my_db.select( """ SELECT last_check FROM tv_shows_not_found WHERE indexer = ? AND indexer_id = ? """, [self.tvid, self.prodid]) if sql_result: last_check = helpers.try_int(sql_result[0]['last_check']) my_db.upsert('tv_shows_not_found', dict(fail_count=v, last_check=last_check, last_success=self._last_found_on_indexer), dict(indexer=self.tvid, indexer_id=self.prodid)) self._not_found_count = v @property def last_found_on_indexer(self): self.helper_load_failed_db() return (self._last_found_on_indexer, self.last_update_indexer)[0 >= self._last_found_on_indexer] def inc_not_found_count(self): my_db = db.DBConnection() sql_result = my_db.select( """ SELECT last_check FROM tv_shows_not_found WHERE indexer = ? AND indexer_id = ? """, [self.tvid, self.prodid]) days = (show_not_found_retry_days - 1, 0)[abs(self.not_found_count) <= concurrent_show_not_found_days] if not sql_result or datetime.datetime.fromtimestamp(helpers.try_int(sql_result[0]['last_check'])) + \ datetime.timedelta(days=days, hours=18) < datetime.datetime.now(): self.not_found_count += (-1, 1)[0 <= self.not_found_count] def reset_not_found_count(self): if 0 != self.not_found_count: self._not_found_count = 0 self._last_found_on_indexer = 0 my_db = db.DBConnection() my_db.action( """ DELETE FROM tv_shows_not_found WHERE indexer = ? AND indexer_id = ? """, [self.tvid, self.prodid]) @property def paused(self): """ :rtype: bool """ return self._paused @paused.setter def paused(self, value): if value != self._paused: if isinstance(value, bool) or (isinstance(value, integer_types) and value in [0, 1]): self._paused = int(value) self.dirty = True else: logger.log('tried to set paused property to invalid value: %s of type: %s' % (value, type(value)), logger.ERROR) @property def ids(self): if not self.internal_ids: acquired_lock = self.lock.acquire(False) if acquired_lock: try: indexermapper.map_indexers_to_show(self) finally: self.lock.release() return self.internal_ids @ids.setter def ids(self, value): if isinstance(value, dict): for k, v in iteritems(value): if k not in indexermapper.indexer_list or \ not isinstance(v, dict) or \ not isinstance(v.get('id'), integer_types) or \ not isinstance(v.get('status'), integer_types) or \ v.get('status') not in indexermapper.MapStatus.allstatus or \ not isinstance(v.get('date'), datetime.date): return self.internal_ids = value @property def is_anime(self): # type: (...) -> bool return 0 < int(self.anime) @property def is_sports(self): # type: (...) -> bool return 0 < int(self.sports) @property def is_scene(self): # type: (...) -> bool return 0 < int(self.scene) def _get_location(self): # type: (...) -> AnyStr # no dir check needed if missing show dirs are created during post-processing if sickgear.CREATE_MISSING_SHOW_DIRS: return self._location if ek.ek(os.path.isdir, self._location): return self._location raise exceptions_helper.ShowDirNotFoundException('Show folder does not exist: \'%s\'' % self._location) def _set_location(self, new_location): # type: (AnyStr) -> None logger.log('Setter sets location to %s' % new_location, logger.DEBUG) # Don't validate dir if user wants to add shows without creating a dir if sickgear.ADD_SHOWS_WO_DIR or ek.ek(os.path.isdir, new_location): self.dirty_setter('_location')(self, new_location) self.path = new_location # self._is_location_good = True else: raise exceptions_helper.NoNFOException('Invalid folder for the show!') location = property(_get_location, _set_location) # delete references to anything that's not in the internal lists def flush_episodes(self): for cur_season_number in self.sxe_ep_obj: for cur_ep_number in self.sxe_ep_obj[cur_season_number]: # noinspection PyUnusedLocal ep_obj = self.sxe_ep_obj[cur_season_number][cur_ep_number] self.sxe_ep_obj[cur_season_number][cur_ep_number] = None del ep_obj def get_all_episodes(self, season=None, has_location=False, check_related_eps=True): # type: (Optional[integer_types], bool, bool) -> List[TVEpisode] """ :param season: None or season number :param has_location: return only with location :param check_related_eps: get related episodes :return: List of TVEpisode objects """ sql_selection = 'SELECT *' if check_related_eps: # subselection to detect multi-episodes early, share_location > 0 sql_selection += """ , (SELECT COUNT (*) FROM tv_episodes WHERE showid = tve.showid AND indexer = tve.indexer AND season = tve.season AND location != '' AND location = tve.location AND episode != tve.episode ) AS share_location """ sql_selection += ' FROM tv_episodes tve WHERE indexer = ? AND showid = ?' sql_parameter = [self.tvid, self.prodid] if None is not season: sql_selection += ' AND season = ?' sql_parameter += [season] if has_location: sql_selection += ' AND location != "" ' # need ORDER episode ASC to rename multi-episodes in order S01E01-02 sql_selection += ' ORDER BY season ASC, episode ASC' my_db = db.DBConnection() sql_result = my_db.select(sql_selection, sql_parameter) ep_obj_list = [] for cur_row in sql_result: ep_obj = self.get_episode(int(cur_row['season']), int(cur_row['episode']), ep_result=[cur_row]) if ep_obj: ep_obj.related_ep_obj = [] if check_related_eps and ep_obj.location: # if there is a location, check if it's a multi-episode (share_location > 0) # and put into related_ep_obj if 0 < cur_row['share_location']: # noinspection SqlRedundantOrderingDirection related_ep_sql_result = my_db.select( """ SELECT * FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND location = ? AND episode != ? ORDER BY episode ASC """, [self.tvid, self.prodid, ep_obj.season, ep_obj.location, ep_obj.episode]) for cur_ep_row in related_ep_sql_result: related_ep_obj = self.get_episode(int(cur_ep_row['season']), int(cur_ep_row['episode']), ep_result=[cur_ep_row]) if related_ep_obj not in ep_obj.related_ep_obj: ep_obj.related_ep_obj.append(related_ep_obj) ep_obj_list.append(ep_obj) return ep_obj_list def get_episode(self, season=None, # type: Optional[int] episode=None, # type: Optional[int] path=None, # type: Optional[AnyStr] no_create=False, # type: bool absolute_number=None, # type: Optional[int] ep_result=None, # type: Optional[List[Row]] existing_only=False # type: bool ): # type: (...) -> Optional[TVEpisode] """ Initialise sxe_ep_obj with db fetched season keys, and then fill the TVShow episode property and return an TVEpisode object if no_create is False :param season: Season number :param episode: Episode number :param path: path to file episode :param no_create: return None instead of an instantiated TVEpisode object :param absolute_number: absolute number :param ep_result: :param existing_only: only return existing episodes :return: TVEpisode object """ # if we get an anime get the real season and episode if self.is_anime and absolute_number and not season and not episode: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT season, episode FROM tv_episodes WHERE indexer = ? AND showid = ? AND absolute_number = ? AND season != 0 LIMIT 2 """, [self.tvid, self.prodid, absolute_number]) if 1 != len(sql_result): msg = 'found for absolute number: %s in show: %s' % (absolute_number, self._name) if not len(sql_result): logger.debug('No entries %s' % msg) else: logger.error('Multiple entries %s' % msg) return season = int(sql_result[0]['season']) episode = int(sql_result[0]['episode']) logger.debug('Found episode by absolute_number: %s which is %sx%s' % (absolute_number, season, episode)) if season not in self.sxe_ep_obj: self.sxe_ep_obj[season] = {} if episode not in self.sxe_ep_obj[season] or None is self.sxe_ep_obj[season][episode]: if no_create: return # logger.log('%s: An object for episode %sx%s did not exist in the cache, trying to create it' % # (self.tvid_prodid, season, episode), logger.DEBUG) if path and not existing_only: ep_obj = TVEpisode(self, season, episode, path, show_result=ep_result) else: ep_obj = TVEpisode(self, season, episode, show_result=ep_result, existing_only=existing_only) if None is not ep_obj: self.sxe_ep_obj[season][episode] = ep_obj return self.sxe_ep_obj[season][episode] def _load_cast_from_db(self): # type: (...) -> List[Character] # cast_list = [] old_cast = [] if None is self._cast_list else self._cast_list() old_list = {c.id for c in old_cast or []} my_db = db.DBConnection() sql_result = my_db.select( """ SELECT castlist.sort_order AS sort_order, characters.name AS name, characters.bio AS c_bio, characters.id AS c_id, characters.image_url AS image_url, characters.thumb_url AS thumb_url, characters.updated AS c_updated, castlist.updated AS cast_updated, persons.name AS p_name, persons.gender AS gender, persons.bio AS p_bio, persons.birthdate AS birthdate, persons.thumb_url AS p_thumb, persons.image_url AS p_image, persons.deathdate AS deathdate, persons.id AS p_id, persons.birthplace AS birthplace, persons.updated AS p_updated, persons.deathplace AS deathplace, persons.height AS height, persons.realname AS realname, persons.nicknames AS nicknames, persons.akas AS akas, (SELECT group_concat(person_ids.src || ':' || person_ids.src_id, ';;;') FROM person_ids WHERE person_ids.person_id = persons.id) AS p_ids, (SELECT group_concat(character_ids.src || ':' || character_ids.src_id, ';;;') FROM character_ids WHERE character_ids.character_id = characters.id) AS c_ids, (SELECT group_concat(character_person_years.person_id || ':' || character_person_years.start_year || '-' || character_person_years.end_year, ';;;') FROM character_person_years WHERE character_person_years.character_id = characters.id) AS p_years FROM castlist LEFT JOIN character_person_map ON castlist.character_id = character_person_map.character_id LEFT JOIN characters ON character_person_map.character_id = characters.id LEFT JOIN persons ON character_person_map.person_id = persons.id WHERE castlist.indexer = ? AND castlist.indexer_id = ? ORDER BY castlist.sort_order """, [self.tvid, self.prodid]) old_cast = old_cast or [] for cur_row in sql_result: existing_character = next( (c for c in old_cast or [] if None is not c.id and c.id == cur_row['c_id']), None) # type: Optional[Character] birthdate = try_int(cur_row['birthdate'], None) birthdate = birthdate and datetime.date.fromordinal(cur_row['birthdate']) deathdate = try_int(cur_row['deathdate'], None) deathdate = deathdate and datetime.date.fromordinal(cur_row['deathdate']) p_years = {} for p in (cur_row['p_years'] and cur_row['p_years'].split(';;;')) or []: p_id, py = p.split(':') start, end = py.split('-') p_years[int(p_id)] = {'start': try_int(start, None), 'end': try_int(end, None)} p_ids, c_ids = {}, {} for i in (cur_row['p_ids'] and cur_row['p_ids'].split(';;;')) or []: k, v = i.split(':') k = try_int(k, None) if v: if k in (TVINFO_INSTAGRAM, TVINFO_TWITTER, TVINFO_FACEBOOK, TVINFO_WIKIPEDIA): p_ids[k] = v else: p_ids[k] = try_int(v, None) for i in (cur_row['c_ids'] and cur_row['c_ids'].split(';;;')) or []: k, v = i.split(':') v = try_int(v, None) if v: c_ids[int(k)] = try_int(v, None) person = Person(cur_row['p_name'], cur_row['gender'], akas=set((cur_row['akas'] and cur_row['akas'].split(';;;')) or []), bio=cur_row['p_bio'], birthday=birthdate, birthplace=cur_row['birthplace'], deathday=deathdate, deathplace=cur_row['deathplace'], height=cur_row['height'], ids=p_ids, image_url=cur_row['p_image'], nicknames=set((cur_row['nicknames'] and cur_row['nicknames'].split(';;;')) or []), real_name=cur_row['realname'], show_obj=self, sid=cur_row['p_id'], thumb_url=cur_row['p_thumb'], updated=cur_row['p_updated']) if existing_character: existing_character.updated = cur_row['cast_updated'] try: old_list.remove(existing_character.id) except (BaseException, Exception): pass existing_person = next((p for p in existing_character.person if (None is not p.id and p.ids.get(self.tvid) == person.ids.get(self.tvid)) or p.name == person.name), None) # type: Optional[Person] existing_character.combine_start_end_years(p_years) if existing_person: existing_person.update_properties( akas=set((cur_row['akas'] and cur_row['akas'].split(';;;')) or []), biography=person.biography, birthday=person.birthday, birthplace=person.birthplace, deathday=person.deathday, deathplace=cur_row['deathplace'], gender=person.gender, height=cur_row['height'], ids=dict(chain.from_iterable(iteritems(d) for d in (existing_person.ids, person.ids or {}))), image_url=person.image_url, name=person.name, nicknames=set((cur_row['nicknames'] and cur_row['nicknames'].split(';;;')) or []), real_name=cur_row['realname'], thumb_url=person.thumb_url ) else: existing_character.person.append(person) else: old_cast.append(Character( cur_row['name'], bio=cur_row['c_bio'], ids=c_ids, image_url=cur_row['image_url'], person=[person], persons_years=p_years, show_obj=self, sid=cur_row['c_id'], thumb_url=cur_row['thumb_url'], updated=cur_row['cast_updated'])) cast_list = weakList(c for c in old_cast or [] if c.id not in old_list) self.cast_list = cast_list return cast_list def cast_list_id(self): # type: (...) -> Set """ return an identifier that represents the current state of the show cast list used to tell if a change has occurred in a cast list after time/process """ return set((c.name, c.image_url or '', c.thumb_url or '', hash(*([', '.join(p.name for p in c.person or [] if p.name)]))) for c in self.cast_list or [] if c.name) @staticmethod def orphaned_cast_sql(): # type: (...) -> List """ returns list of sql that removes all character and person entries that are not linked to any show """ return [ ['DELETE FROM castlist WHERE castlist.id IN (SELECT c.id FROM castlist c' ' LEFT JOIN tv_shows ON tv_shows.indexer = c.indexer AND tv_shows.indexer_id = c.indexer_id' ' WHERE tv_shows.indexer_id is NULL);'], ['DELETE FROM characters WHERE characters.id IN (SELECT c.id FROM characters c' ' LEFT JOIN castlist on castlist.character_id = c.id WHERE castlist.id is NULL);'], ['DELETE FROM character_ids WHERE character_ids.character_id IN' ' (SELECT c.character_id FROM character_ids c' ' LEFT JOIN characters ON c.character_id = characters.id WHERE characters.id is NULL);'], ['DELETE FROM character_person_map WHERE character_id IN (SELECT c.character_id FROM character_person_map c' ' LEFT JOIN characters ON c.character_id = characters.id WHERE characters.id is NULL)'], ['DELETE FROM persons WHERE persons.id IN (SELECT p.id FROM persons p' ' LEFT JOIN character_person_map c on p.id = c.person_id WHERE c.id is NULL);'], ['DELETE FROM person_ids WHERE person_ids.person_id IN (SELECT p.person_id FROM person_ids p' ' LEFT JOIN persons ON p.person_id = persons.id WHERE persons.id is NULL);'], ['DELETE FROM castlist WHERE castlist.id IN (SELECT c.id FROM castlist c' ' LEFT JOIN character_person_map cpm on c.character_id = cpm.character_id WHERE cpm.id IS NULL);'], ['DELETE FROM character_person_years WHERE character_person_years.ROWID IN (SELECT c.ROWID FROM' ' character_person_years c' ' LEFT JOIN characters ch on c.character_id = ch.id WHERE ch.id IS NULL);'], ['DELETE FROM character_person_years WHERE character_person_years.ROWID IN (SELECT c.ROWID FROM' ' character_person_years c' ' LEFT JOIN persons p on c.person_id = p.id WHERE p.id IS NULL);'], ] def _save_cast_list(self, removed_char_ids=None, force=False, stop_event=None, cast_list=None): # type: (Union[List[integer_types], Set[integer_types]], bool, threading.Event, List[Character]) -> None if cast_list: my_db = db.DBConnection() cl = [] for cur_id in removed_char_ids or []: cl.extend([[ """ DELETE FROM castlist WHERE indexer = ? AND indexer_id = ? AND character_id = ?; """, [self.tvid, self.prodid, cur_id]] ]) update_date = datetime.date.today().toordinal() for cur_enum, cur_cast in enumerate(cast_list, 1): if stop_event and stop_event.is_set(): return cur_cast.save_to_db(show_obj=self, force=force, stop_event=stop_event) cl.extend([[ """ UPDATE castlist SET sort_order = ?, updated = ? WHERE indexer = ? AND indexer_id = ? AND character_id = ?; """, [cur_enum, update_date, self.tvid, self.prodid, cur_cast.id] ], [ """ INSERT INTO castlist (indexer, indexer_id, character_id, sort_order, updated) SELECT %s, %s, %s, %s, %s WHERE changes() == 0; """ % (self.tvid, self.prodid, cur_cast.id, cur_enum, update_date)] ]) if removed_char_ids: # remove orphaned entries cl.extend(self.orphaned_cast_sql()) if cl: my_db.mass_action(cl) def should_update(self, update_date=datetime.date.today(), last_indexer_change=None): # type: (datetime.date, integer_types) -> bool if last_indexer_change and self._src_update_time: # update if show was updated or not updated within the last 30 days if self._src_update_time < last_indexer_change or \ (self._last_update_indexer and datetime.date.today().toordinal() - self._last_update_indexer > 30): return True return False # In some situations self.status = None.. need to figure out where that is! if not self._status: self.status = '' logger.log('Status missing for show: [%s] with status: [%s]' % (self.tvid_prodid, self._status), logger.DEBUG) last_update_indexer = datetime.date.fromordinal(self._last_update_indexer) # if show was not found for 1 week, only retry to update once a week if (concurrent_show_not_found_days < abs(self.not_found_count)) \ and (update_date - last_update_indexer) < datetime.timedelta(days=show_not_found_retry_days): return False my_db = db.DBConnection() sql_result = my_db.mass_action([[ """ SELECT airdate FROM [tv_episodes] WHERE indexer = ? AND showid = ? AND season > 0 ORDER BY season DESC, episode DESC LIMIT 1 """, [self.tvid, self.prodid] ], [ """ SELECT airdate FROM [tv_episodes] WHERE indexer = ? AND showid = ? AND season > 0 AND airdate > 1 ORDER BY airdate DESC LIMIT 1 """, [self.tvid, self.prodid]]]) last_airdate_unknown = 1 >= int(sql_result[0][0]['airdate']) if sql_result and sql_result[0] else True last_airdate = datetime.date.fromordinal(sql_result[1][0]['airdate']) \ if sql_result and sql_result[1] else datetime.date.fromordinal(1) # if show is not 'Ended' and last episode aired less then 460 days ago # or don't have an airdate for the last episode always update (status 'Continuing' or '') update_days_limit = 2013 ended_limit = datetime.timedelta(days=update_days_limit) if 'Ended' not in self._status and (last_airdate == datetime.date.fromordinal(1) or last_airdate_unknown or (update_date - last_airdate) <= ended_limit or (update_date - last_update_indexer) > ended_limit): return True # in the first 460 days (last airdate), update regularly airdate_diff = update_date - last_airdate last_update_diff = update_date - last_update_indexer update_step_list = [[60, 1], [120, 3], [180, 7], [1281, 15], [update_days_limit, 30]] for date_diff, interval in update_step_list: if airdate_diff <= datetime.timedelta(days=date_diff) \ and last_update_diff >= datetime.timedelta(days=interval): return True # update shows without an airdate for the last episode for update_days_limit days every 7 days return last_airdate_unknown and airdate_diff <= ended_limit and last_update_diff >= datetime.timedelta(days=7) def write_show_nfo(self, force=False): # type: (bool) -> bool result = False if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, skipping NFO generation' % self.tvid_prodid) return False logger.log('%s: Writing NFOs for show' % self.tvid_prodid) for cur_provider in itervalues(sickgear.metadata_provider_dict): result = cur_provider.create_show_metadata(self, force) or result return result def write_metadata(self, show_only=False, force=False): # type:(bool, bool) -> None """ :param show_only: only for show :param force: """ if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, skipping NFO generation' % self.tvid_prodid) return self.get_images() force_nfo = force or not db.DBConnection().has_flag('kodi_nfo_uid') self.write_show_nfo(force_nfo) if not show_only: self.write_episode_nfo(force_nfo) def write_episode_nfo(self, force=False): # type: (bool) -> None if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, skipping NFO generation' % self.tvid_prodid) return logger.log('%s: Writing NFOs for all episodes' % self.tvid_prodid) my_db = db.DBConnection() # noinspection SqlResolve sql_result = my_db.select( """ SELECT * FROM tv_episodes WHERE indexer = ? AND showid = ? AND location != '' """, [self.tvid, self.prodid]) processed = [] for cur_row in sql_result: if (cur_row['season'], cur_row['episode']) in processed: continue logger.log('%s: Retrieving/creating episode %sx%s' % (self.tvid_prodid, cur_row['season'], cur_row['episode']), logger.DEBUG) ep_obj = self.get_episode(cur_row['season'], cur_row['episode'], ep_result=[cur_row]) if not ep_obj.related_ep_obj: processed += [(cur_row['season'], cur_row['episode'])] else: logger.log('%s: Found related to %sx%s episode(s)... %s' % (self.tvid_prodid, cur_row['season'], cur_row['episode'], ', '.join(['%sx%s' % (x.season, x.episode) for x in ep_obj.related_ep_obj])), logger.DEBUG) processed += list(set([(cur_row['season'], cur_row['episode'])] + [(x.season, x.episode) for x in ep_obj.related_ep_obj])) ep_obj.create_meta_files(force) def update_metadata(self): if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, skipping NFO generation' % self.tvid_prodid) return self.update_show_nfo() def update_show_nfo(self): result = False if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, skipping NFO generation' % self.tvid_prodid) return False logger.log('%s: Updating NFOs for show with new TV info' % self.tvid_prodid) for cur_provider in itervalues(sickgear.metadata_provider_dict): result = cur_provider.update_show_indexer_metadata(self) or result return result # find all media files in the show folder and create episodes for as many as possible def load_episodes_from_dir(self): if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, not loading episodes from disk' % self.tvid_prodid) return logger.log('%s: Loading all episodes from the show directory %s' % (self.tvid_prodid, self._location)) # get file list file_list = helpers.list_media_files(self._location) # create TVEpisodes from each media file (if possible) sql_l = [] for cur_media_file in file_list: parse_result = None ep_obj = None logger.log('%s: Creating episode from %s' % (self.tvid_prodid, cur_media_file), logger.DEBUG) try: ep_obj = self.ep_obj_from_file(ek.ek(os.path.join, self._location, cur_media_file)) except (exceptions_helper.ShowNotFoundException, exceptions_helper.EpisodeNotFoundException) as e: logger.log('Episode %s returned an exception: %s' % (cur_media_file, ex(e)), logger.ERROR) continue except exceptions_helper.EpisodeDeletedException: logger.log('The episode deleted itself when I tried making an object for it', logger.DEBUG) if None is ep_obj: continue # see if we should save the release name in the db ep_file_name = ek.ek(os.path.basename, ep_obj.location) ep_file_name = ek.ek(os.path.splitext, ep_file_name)[0] try: parse_result = None np = NameParser(False, show_obj=self) parse_result = np.parse(ep_file_name) except (InvalidNameException, InvalidShowException): pass if ep_file_name and parse_result and None is not parse_result.release_group and not ep_obj.release_name: logger.log( 'Name %s gave release group of %s, seems valid' % (ep_file_name, parse_result.release_group), logger.DEBUG) ep_obj.release_name = ep_file_name # store the reference in the show if None is not ep_obj: if sickgear.USE_SUBTITLES and self.subtitles: try: ep_obj.refresh_subtitles() except (BaseException, Exception): logger.log('%s: Could not refresh subtitles' % self.tvid_prodid, logger.ERROR) logger.log(traceback.format_exc(), logger.ERROR) result = ep_obj.get_sql() if None is not result: sql_l.append(result) if 0 < len(sql_l): my_db = db.DBConnection() my_db.mass_action(sql_l) def load_episodes_from_db(self, update=False): # type: (bool) -> Dict[int, Dict[int, TVEpisode]] """ :param update: :return: """ logger.log('Loading all episodes for [%s] from the DB' % self._name) my_db = db.DBConnection() sql_result = my_db.select( """ SELECT * FROM tv_episodes WHERE indexer = ? AND showid = ? """, [self.tvid, self.prodid]) scanned_eps = {} tvinfo_config = sickgear.TVInfoAPI(self.tvid).api_params.copy() if self._lang: tvinfo_config['language'] = self._lang if 0 != self._dvdorder: tvinfo_config['dvdorder'] = True t = sickgear.TVInfoAPI(self.tvid).setup(**tvinfo_config) cached_show = None try: cached_show = t.get_show(self.prodid, language=self._lang) except BaseTVinfoError as e: logger.log('Unable to find cached seasons from %s: %s' % ( sickgear.TVInfoAPI(self.tvid).name, ex(e)), logger.WARNING) if None is cached_show: return scanned_eps scene_sql_result = my_db.select( """ SELECT * FROM scene_numbering WHERE indexer == ? AND indexer_id = ? """, [self.tvid, self.prodid]) cached_seasons = {} cl = [] for cur_row in sql_result: delete_ep = False season = int(cur_row['season']) episode = int(cur_row['episode']) if season not in cached_seasons: try: cached_seasons[season] = cached_show[season] except BaseTVinfoSeasonnotfound as e: logger.log('Error when trying to load the episode for [%s] from %s: %s' % (self._name, sickgear.TVInfoAPI(self.tvid).name, ex(e)), logger.WARNING) delete_ep = True if season not in scanned_eps: scanned_eps[season] = {} logger.log('Loading episode %sx%s for [%s] from the DB' % (season, episode, self.name), logger.DEBUG) try: ep_obj = self.get_episode(season, episode, ep_result=[cur_row]) # type: TVEpisode # if we found out that the ep is no longer on TVDB then delete it from our database too if delete_ep and helpers.should_delete_episode(ep_obj.status): cl.extend(ep_obj.delete_episode(return_sql=True)) else: ep_obj.load_from_db(season, episode, show_result=[cur_row], scene_result=scene_sql_result) ep_obj.load_from_tvinfo(tvapi=t, update=update, cached_show=cached_show) scanned_eps[season][episode] = True except exceptions_helper.EpisodeDeletedException: logger.log('Tried loading an episode that should have been deleted from the DB [%s], skipping it' % self._name, logger.DEBUG) continue if cl: my_db.mass_action(cl) return scanned_eps def switch_ep_change_sql(self, old_tvid, old_prodid, season, episode, reason): # type: (int, integer_types, int, int, int) -> List[AnyStr] return [ """ REPLACE INTO switch_ep_result (old_indexer, old_indexer_id, new_indexer, new_indexer_id, season, episode, reason) VALUES (?,?,?,?,?,?,?) """, [old_tvid, old_prodid, self.tvid, self.prodid, season, episode, reason]] def load_episodes_from_tvinfo(self, cache=True, update=False, tvinfo_data=None, switch=False, old_tvid=None, old_prodid=None): # type: (bool, bool, TVInfoShow, bool, int, integer_types) -> Optional[Dict[int, Dict[int, TVEpisode]]] """ :param cache: :param update: :param tvinfo_data: :param switch: :param old_tvid: :param old_prodid: :return: """ tvinfo_config = sickgear.TVInfoAPI(self.tvid).api_params.copy() if not cache: tvinfo_config['cache'] = False if self._lang: tvinfo_config['language'] = self._lang if 0 != self._dvdorder: tvinfo_config['dvdorder'] = True logger.log('%s: Loading all episodes for [%s] from %s..' % (self.tvid_prodid, self._name, sickgear.TVInfoAPI(self.tvid).name)) if getattr(tvinfo_data, 'id', None) == self.prodid: show_obj = tvinfo_data t = None else: try: t = sickgear.TVInfoAPI(self.tvid).setup(**tvinfo_config) show_obj = t.get_show(self.prodid, language=self._lang) except BaseTVinfoError: logger.log('%s timed out, unable to update episodes for [%s] from %s' % (sickgear.TVInfoAPI(self.tvid).name, self._name, sickgear.TVInfoAPI(self.tvid).name), logger.ERROR) return None scanned_eps = {} my_db = db.DBConnection() sql_result = my_db.select( """ SELECT * FROM tv_episodes WHERE indexer = ? AND showid = ? """, [self.tvid, self.prodid]) sql_l = [] for season in show_obj: scanned_eps[season] = {} for episode in show_obj[season]: # need some examples of wtf episode 0 means to decide if we want it or not if 0 == episode: continue try: ep_obj = self.get_episode(season, episode, ep_result=sql_result) # type: TVEpisode except exceptions_helper.EpisodeNotFoundException: logger.log('%s: %s object for %sx%s from [%s] is incomplete, skipping this episode' % (self.tvid_prodid, sickgear.TVInfoAPI(self.tvid).name, season, episode, self._name)) continue else: try: ep_obj.load_from_tvinfo(tvapi=t, update=update, cached_show=show_obj, switch=switch, old_tvid=old_tvid, old_prodid=old_prodid, switch_list=sql_l) except exceptions_helper.EpisodeDeletedException: logger.log('The episode from [%s] was deleted, skipping the rest of the load' % self._name) continue with ep_obj.lock: logger.log('%s: Loading info from %s for episode %sx%s from [%s]' % (self.tvid_prodid, sickgear.TVInfoAPI(self.tvid).name, season, episode, self._name), logger.DEBUG) ep_obj.load_from_tvinfo(season, episode, tvapi=t, update=update, cached_show=show_obj, switch=switch, old_tvid=old_tvid, old_prodid=old_prodid, switch_list=sql_l) result = ep_obj.get_sql() if None is not result: sql_l.append(result) scanned_eps[season][episode] = True if 0 < len(sql_l): my_db = db.DBConnection() my_db.mass_action(sql_l) # Done updating save last update date self.last_update_indexer = datetime.date.today().toordinal() self.save_to_db() return scanned_eps def get_images(self): fanart_result = poster_result = banner_result = False season_posters_result = season_banners_result = season_all_poster_result = season_all_banner_result = False for cur_provider in itervalues(sickgear.metadata_provider_dict): # FIXME: Needs to not show this message if the option is not enabled? logger.log('Running metadata routines for %s' % cur_provider.name, logger.DEBUG) fanart_result = cur_provider.create_fanart(self) or fanart_result poster_result = cur_provider.create_poster(self) or poster_result banner_result = cur_provider.create_banner(self) or banner_result season_posters_result = cur_provider.create_season_posters(self) or season_posters_result season_banners_result = cur_provider.create_season_banners(self) or season_banners_result season_all_poster_result = cur_provider.create_season_all_poster(self) or season_all_poster_result season_all_banner_result = cur_provider.create_season_all_banner(self) or season_all_banner_result return fanart_result or poster_result or banner_result or season_posters_result or season_banners_result \ or season_all_poster_result or season_all_banner_result # make a TVEpisode object from a media file def ep_obj_from_file(self, path): # type: (AnyStr) -> Optional[TVEpisode] """ :param path: :return: """ if not ek.ek(os.path.isfile, path): logger.log('%s: Not a real file... %s' % (self.tvid_prodid, path)) return None logger.log('%s: Creating episode object from %s' % (self.tvid_prodid, path), logger.DEBUG) try: my_parser = NameParser(show_obj=self) parse_result = my_parser.parse(path) except InvalidNameException: logger.log('Unable to parse the filename %s into a valid episode' % path, logger.DEBUG) return None except InvalidShowException: logger.log('Unable to parse the filename %s into a valid show' % path, logger.DEBUG) return None if not len(parse_result.episode_numbers): logger.log('parse_result: %s' % parse_result) logger.log('No episode number found in %s, ignoring it' % path, logger.ERROR) return None # for now lets assume that any episode in the show dir belongs to that show season_number = parse_result.season_number if None is not parse_result.season_number else 1 episode_numbers = parse_result.episode_numbers root_ep_obj = None sql_l = [] for cur_ep_num in episode_numbers: cur_ep_num = int(cur_ep_num) logger.log('%s: %s parsed to %s %sx%s' % (self.tvid_prodid, path, self._name, season_number, cur_ep_num), logger.DEBUG) check_quality_again = False same_file = False ep_obj = self.get_episode(season_number, cur_ep_num) if None is ep_obj: try: ep_obj = self.get_episode(season_number, cur_ep_num, path) except exceptions_helper.EpisodeNotFoundException: logger.log('%s: Unable to figure out what this file is, skipping' % self.tvid_prodid, logger.ERROR) continue else: # if there is a new file associated with this ep then re-check the quality status, quality = sickgear.common.Quality.splitCompositeStatus(ep_obj.status) if IGNORED == status: continue if (ep_obj.location and ek.ek(os.path.normpath, ep_obj.location) != ek.ek(os.path.normpath, path)) or \ (not ep_obj.location and path) or \ (SKIPPED == status): logger.log('The old episode had a different file associated with it, re-checking the quality ' + 'based on the new filename %s' % path, logger.DEBUG) check_quality_again = True with ep_obj.lock: old_size = ep_obj.file_size if ep_obj.location and status != SKIPPED else 0 ep_obj.location = path # if the sizes are the same then it's probably the same file if old_size and ep_obj.file_size == old_size: same_file = True else: same_file = False ep_obj.check_for_meta_files() if None is root_ep_obj: root_ep_obj = ep_obj else: if ep_obj not in root_ep_obj.related_ep_obj: root_ep_obj.related_ep_obj.append(ep_obj) # if it's a new file then if not same_file: ep_obj.release_name = '' # if user replaces a file, attempt to recheck the quality unless it's know to be the same file if check_quality_again and not same_file: new_quality = Quality.nameQuality(path, self.is_anime) if Quality.UNKNOWN == new_quality: new_quality = Quality.fileQuality(path) logger.log('Since this file was renamed, file %s was checked and quality "%s" found' % (path, Quality.qualityStrings[new_quality]), logger.DEBUG) status, quality = sickgear.common.Quality.splitCompositeStatus(ep_obj.status) if Quality.UNKNOWN != new_quality or status in (SKIPPED, UNAIRED): ep_obj.status = Quality.compositeStatus(DOWNLOADED, new_quality) # check for status/quality changes as long as it's a new file elif not same_file and sickgear.helpers.has_media_ext(path)\ and ep_obj.status not in Quality.DOWNLOADED + Quality.ARCHIVED + [IGNORED]: old_status, old_quality = Quality.splitCompositeStatus(ep_obj.status) new_quality = Quality.nameQuality(path, self.is_anime) if Quality.UNKNOWN == new_quality: new_quality = Quality.fileQuality(path) if Quality.UNKNOWN == new_quality: new_quality = Quality.assumeQuality(path) new_status = None # if it was snatched and now exists then set the status correctly if SNATCHED == old_status and old_quality <= new_quality: logger.log('STATUS: this episode used to be snatched with quality %s but' ' a file exists with quality %s so setting the status to DOWNLOADED' % (Quality.qualityStrings[old_quality], Quality.qualityStrings[new_quality]), logger.DEBUG) new_status = DOWNLOADED # if it was snatched proper and we found a higher quality one then allow the status change elif SNATCHED_PROPER == old_status and old_quality < new_quality: logger.log('STATUS: this episode used to be snatched proper with quality %s but' ' a file exists with quality %s so setting the status to DOWNLOADED' % (Quality.qualityStrings[old_quality], Quality.qualityStrings[new_quality]), logger.DEBUG) new_status = DOWNLOADED elif old_status not in SNATCHED_ANY: new_status = DOWNLOADED if None is not new_status: with ep_obj.lock: logger.log('STATUS: we have an associated file, so setting the status from %s to DOWNLOADED/%s' % (ep_obj.status, Quality.compositeStatus(new_status, new_quality)), logger.DEBUG) ep_obj.status = Quality.compositeStatus(new_status, new_quality) elif same_file: status, quality = Quality.splitCompositeStatus(ep_obj.status) if status in (SKIPPED, UNAIRED): new_quality = Quality.nameQuality(path, self.is_anime) if Quality.UNKNOWN == new_quality: new_quality = Quality.fileQuality(path) logger.log('Since this file has status: "%s", file %s was checked and quality "%s" found' % (statusStrings[status], path, Quality.qualityStrings[new_quality]), logger.DEBUG) ep_obj.status = Quality.compositeStatus(DOWNLOADED, new_quality) with ep_obj.lock: result = ep_obj.get_sql() if None is not result: sql_l.append(result) if 0 < len(sql_l): my_db = db.DBConnection() my_db.mass_action(sql_l) # creating metafiles on the root should be good enough if sickgear.USE_FAILED_DOWNLOADS and None is not root_ep_obj: with root_ep_obj.lock: root_ep_obj.create_meta_files() return root_ep_obj def load_from_db(self, show_result=None, imdb_info_result=None): # type: (Optional[Row], Optional[Union[Row, Dict]]) -> Optional[bool] """ :return: """ if not show_result or self.tvid != show_result['indexer'] or self.prodid != show_result['indexer_id']: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT * FROM tv_shows WHERE indexer = ? AND indexer_id = ? """, [self.tvid, self.prodid]) if 1 != len(sql_result): if len(sql_result): tvinfo_config = sickgear.TVInfoAPI(self.tvid).api_params.copy() if self._lang: tvinfo_config['language'] = self._lang t = sickgear.TVInfoAPI(self.tvid).setup(**tvinfo_config) cached_show = t.get_show(self.prodid, load_episodes=False, language=self._lang) vals = (self.prodid, '' if not cached_show else ' [%s]' % cached_show['seriesname'].strip()) if len(sql_result): logger.log('%s: Loading show info%s from database' % vals) raise exceptions_helper.MultipleDBShowsException() logger.log('%s-%s: Unable to find the show%s in the database' % (self.tvid, self.prodid, self.name)) return show_result = next(iter(sql_result)) if not self.tvid: self.tvid, self.prodid = int(show_result['indexer']), int(show_result['indexer_id']) self._air_by_date = show_result['air_by_date'] or 0 self._airs = show_result['airs'] or '' self._airtime = self._make_airtime(show_result['airtime']) self._anime = show_result['anime'] or 0 self._classification = None is self._classification and show_result['classification'] or '' self._dvdorder = show_result['dvdorder'] or 0 self._flatten_folders = int(show_result['flatten_folders']) self._genre = '|'.join([_g for _g in (self._genre or show_result['genre'] or '').split('|') if _g]) self._lang = self._lang or show_result['lang'] or '' self._last_update_indexer = show_result['last_update_indexer'] self._name = self._name or show_result['show_name'] or '' self._network_country = show_result['network_country'] self._network_country_code = show_result['network_country_code'] self._network_id = show_result['network_id'] self._network_is_stream = bool(show_result['network_is_stream']) self._overview = self._overview or show_result['overview'] or '' self._paused = int(show_result['paused']) self._prune = show_result['prune'] or 0 self._quality = int(show_result['quality']) self._runtime = show_result['runtime'] self._scene = show_result['scene'] or 0 self._sports = show_result['sports'] or 0 self._src_update_time = try_int(show_result['src_update_timestamp'], None) self._startyear = show_result['startyear'] or 0 self._status = show_result['status'] or '' self._subtitles = show_result['subtitles'] and 1 or 0 self._tag = show_result['tag'] or 'Show List' self._upgrade_once = show_result['archive_firstmatch'] or 0 if not self._imdbid: imdbid = show_result['imdb_id'] or '' self._imdbid = ('', imdbid)[2 < len(imdbid)] self._rls_global_exclude_ignore = helpers.split_word_str( show_result['rls_global_exclude_ignore'])[0] self._rls_global_exclude_require = helpers.split_word_str( show_result['rls_global_exclude_require'])[0] self._rls_ignore_words, self._rls_ignore_words_regex = helpers.split_word_str( show_result['rls_ignore_words']) self._rls_require_words, self._rls_require_words_regex = helpers.split_word_str( show_result['rls_require_words']) self.internal_network = self.internal_network or show_result['network'] or '' self.internal_timezone = show_result['timezone'] if not self.internal_timezone and self.internal_network: _, self.internal_timezone = network_timezones.get_network_timezone(self.internal_network, return_name=True) try: self.location = show_result['location'] except (BaseException, Exception): self.dirty_setter('_location')(self, show_result['location']) # self._is_location_good = False self.release_groups = self._anime and AniGroupList(self.tvid, self.prodid, self.tvid_prodid) or None logger.log(u'Loaded.. {: <9} {: <8} {}'.format( sickgear.TVInfoAPI(self.tvid).config.get('name') + ',', '%s,' % self.prodid, self.name)) # Get IMDb_info from database if not imdb_info_result or \ self.tvid != imdb_info_result['indexer'] or self.prodid != imdb_info_result['indexer_id']: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT * FROM imdb_info WHERE indexer = ? AND indexer_id = ? """, [self.tvid, self.prodid]) else: sql_result = [imdb_info_result] if 0 < len(sql_result): # this keys() is not a dict if isinstance(sql_result[0], dict): self._imdb_info = sql_result[0] else: self._imdb_info = dict(zip(sql_result[0].keys(), [(r, '')[None is r] for r in sql_result[0]])) if 'is_mini_series' in self._imdb_info: self._imdb_info['is_mini_series'] = bool(self._imdb_info['is_mini_series']) elif sickgear.USE_IMDB_INFO: logger.log('%s: The next show update will attempt to find IMDb info for [%s]' % (self.tvid_prodid, self.name), logger.DEBUG) return self.dirty = False return True def _get_tz_info(self): if self.internal_timezone: return tz.gettz(self.internal_timezone, zoneinfo_priority=True) elif self.internal_network: return network_timezones.get_network_timezone(self.internal_network) def _make_airtime(self, airtime=None): # type: (Optional[integer_types]) -> Optional[datetime.time] if isinstance(airtime, integer_types): return int_to_time(airtime) if self._airs: hour, minute = network_timezones.parse_time(self._airs) return datetime.time(hour=hour, minute=minute) return None def _get_airtime(self): # type: (...) -> Tuple[Optional[datetime.time], Optional[string_types], Optional[string_types]] """ try to get airtime from trakt """ try: tvinfo_config = sickgear.TVInfoAPI(TVINFO_TRAKT).api_params.copy() t = sickgear.TVInfoAPI(TVINFO_TRAKT).setup(**tvinfo_config) results = t.search_show(ids={self._tvid: self._prodid}) for _r in results: if 'show' != _r['type']: continue if _r['ids'][TVINFO_TMDB] == self._prodid: try: h, m = _r['airs']['time'].split(':') h, m = try_int(h, None), try_int(m, None) if None is not h and None is not m: return datetime.time(hour=h, minute=m), _r['airs'].get('time'), _r['airs'].get('day') except (BaseException, Exception): continue except (BaseException, Exception): return None, None, None return None, None, None def _should_cast_update(self, show_info_cast): # type: (CastList) -> bool """ return if cast list has changed and should be updated """ cast_list = self.cast_list existing_cast = set(hash(*([', '.join(p.name for p in c.person or [] if p.name)])) for c in cast_list or [] if c.name) new_cast = set(hash(*([', '.join(p.name for p in c.person or [] if p.name)])) for c_t, c_l in iteritems(show_info_cast or {}) for c in c_l or [] if c.name and c_t in (RoleTypes.ActorMain, RoleTypes.Host, RoleTypes.Interviewer, RoleTypes.Presenter)) now = datetime.date.today().toordinal() max_age = random.randint(30, 60) no_image_cast = any(not (c.image_url or c.thumb_url) and (random.randint(4, 14) < now - c.updated) for c in cast_list or [] if c.name) too_old_char = any(max_age < now - c.updated for c in cast_list or [] if c.name) # too_old_person = any(any(max_age < now - p.updated for p in c.person or [] if p.name) # for c in cast_list or [] if c.name) return existing_cast != new_cast or no_image_cast or too_old_char # or too_old_person def load_from_tvinfo(self, cache=True, tvapi=None, tvinfo_data=None, scheduled_update=False, switch=False): # type: (bool, bool, TVInfoShow, bool, bool) -> Optional[Union[bool, TVInfoShow]] """ :param cache: :param tvapi: :param tvinfo_data: :param scheduled_update: :param switch: """ # There's gotta be a better way of doing this but we don't wanna # change the cache value elsewhere if None is tvapi: tvinfo_config = sickgear.TVInfoAPI(self.tvid).api_params.copy() if not cache: tvinfo_config['cache'] = False if self._lang: tvinfo_config['language'] = self._lang if 0 != self._dvdorder: tvinfo_config['dvdorder'] = True t = sickgear.TVInfoAPI(self.tvid).setup(**tvinfo_config) else: t = tvapi if getattr(tvinfo_data, 'id', None) == self.prodid: show_info = tvinfo_data else: show_info = t.get_show(self.prodid, actors=True, language=self._lang) # type: Optional[TVInfoShow] if None is show_info or getattr(t, 'show_not_found', False): if getattr(t, 'show_not_found', False): self.inc_not_found_count() logger.log('Show [%s] not found (maybe even removed?)' % self._name, logger.WARNING) else: logger.log('Show data [%s] not found' % self._name, logger.WARNING) return False self.reset_not_found_count() try: self.name = show_info['seriesname'].strip() except AttributeError: raise BaseTVinfoAttributenotfound( "Found %s, but attribute 'seriesname' was empty." % self.tvid_prodid) if show_info: logger.log('%s: Loading show info [%s] from %s' % ( self.tvid_prodid, self._name, sickgear.TVInfoAPI(self.tvid).name)) self.classification = self.dict_prevent_nonetype(show_info, 'classification', 'Scripted') self.genre = '|'.join([_g for _g in (self.dict_prevent_nonetype(show_info, 'genre') or '').split('|') if _g]) self.network = self.dict_prevent_nonetype(show_info, 'network') self.runtime = self.dict_prevent_nonetype(show_info, 'runtime') self.dirty_setter('_src_update_time')(self, show_info.updated_timestamp) old_imdb = self.imdbid if show_info.ids.imdb: self.imdbid = 'tt%07d' % show_info.ids.imdb else: self.imdbid = self.dict_prevent_nonetype(show_info, 'imdb_id') if old_imdb != self.imdbid: try: imdb_id = try_int(self.imdbid.replace('tt', ''), None) mapped_imdb = self.ids.get(TVINFO_IMDB, {'id': 0})['id'] if imdb_id != mapped_imdb: indexermapper.map_indexers_to_show(self, recheck=True) except (BaseException, Exception): pass # tmdb doesn't provide air time, try to get it from other source if TVINFO_TMDB == self._tvid and None is show_info.time: show_info.time, show_info.airs_time, show_info.airs_dayofweek = self._get_airtime() if None is not getattr(show_info, 'airs_dayofweek', None) and None is not getattr(show_info, 'airs_time', None): self.airs = ('%s %s' % (show_info['airs_dayofweek'], show_info['airs_time'])).strip() if None is not getattr(show_info, 'firstaired', None): self.startyear = int(str(show_info["firstaired"]).split('-')[0]) self.status = self.dict_prevent_nonetype(show_info, 'status') self.overview = self.dict_prevent_nonetype(show_info, 'overview') self.network_id = show_info.network_id self.network_country = show_info.network_country self.network_country_code = show_info.network_country_code self.network_is_stream = show_info.network_is_stream self.timezone = show_info.network_timezone if not self.internal_timezone and self.internal_network: _, self.timezone = network_timezones.get_network_timezone(self.internal_network, return_name=True) self.airtime = self._make_airtime(show_info.time) if show_info.cast and self._should_cast_update(show_info.cast): sickgear.people_queue_scheduler.action.add_cast_update(show_obj=self, show_info_cast=show_info.cast, scheduled_update=scheduled_update, switch=switch) else: logger.log('Not updating cast for show because data is unchanged.') return show_info @staticmethod def _update_person_properties_helper(person_obj, src_person, p_ids): # type: (Person, TVINFO_Person, Dict) -> None person_obj.update_properties( name=src_person.name, gender=src_person.gender, birthday=src_person.birthdate, deathday=src_person.deathdate, biography=src_person.bio, ids=dict(chain.from_iterable(iteritems(d) for d in (person_obj.ids, p_ids))), deathplace=src_person.deathplace, akas=src_person.akas, nicknames=src_person.nicknames, real_name=src_person.real_name, height=src_person.height) def load_cast_from_tvinfo(self, show_info_cast, force=False, stop_event=None): # type: (CastList, bool, threading.Event) -> None """ add and fetch cast list info for show Note: currently not multithread safe :param show_info_cast: tvinfo castlist :param force: force reload of cast info :param stop_event: """ if stop_event and stop_event.is_set(): return if not show_info_cast: tvinfo_config = sickgear.TVInfoAPI(self.tvid).api_params.copy() t = sickgear.TVInfoAPI(self.tvid).setup(**tvinfo_config) show_info = t.get_show(self.prodid, load_episodes=False, actors=True, language=self._lang) # type: Optional[TVInfoShow] if None is show_info: return show_info_cast = show_info.cast cast_list = self._load_cast_from_db() remove_char_ids = {c.id for c in cast_list or []} cast_ordered = weakList() for ct, c_l in iteritems(show_info_cast): # type: (integer_types, List[TVINFO_Character]) if ct not in (RoleTypes.ActorMain, RoleTypes.Host, RoleTypes.Interviewer, RoleTypes.Presenter): continue for c in c_l: if stop_event and stop_event.is_set(): return unique_name = 1 == len([cu for cu in c_l if (None is not c.id and cu.id == c.id) or cu.name == c.name]) mc = next((cl for cl in cast_list or [] if (None is not c.id and cl.ids.get(self.tvid) == c.id) or (unique_name and c.name and cl.name == c.name) or any(cl.ids.get(src) == c.ids.get(src) for src in c.ids or {})), None) # type: Optional[Character] if not mc: unique_person = not any(1 for _cp in Counter([_p.name for _cha in c_l for _p in _cha.person if any(_p.name == _pp.name for _pp in c.person)]).values() if 1 != _cp) if unique_person: pc = [cl for cl in cast_list or [] if cl.person and any(1 for p in cl.person if c.person and ((None is not p.ids.get(self.tvid) and any(p.ids.get(self.tvid) == cp.id for cp in c.person)) or any(p.name == cp.name for cp in c.person)))] if 1 == len(pc): mc = pc[0] if mc: try: remove_char_ids.remove(mc.id) except KeyError: # can happen with duplicate characters on source logger.debug('%s - error character: %s (%s)' % (self.name, mc.id, mc.name)) old_person_ids = {op.id for op in mc.person if None is not op.id} for src_person in c.person: if src_person.id: person_ids = {self.tvid: src_person.id} else: person_ids = {} existing_person = next( (cur_person for cur_person in mc.person if (None is not src_person.id and cur_person.ids.get(self.tvid) == src_person.id) or (cur_person.name and cur_person.name == src_person.name)), None) # type: Optional[Person] new_person = None if not existing_person: new_person = Person(src_person.name, src_person.gender, src_person.birthdate, src_person.deathdate, src_person.bio, akas=src_person.akas, character_obj=mc, deathplace=src_person.deathplace, height=src_person.height, ids=person_ids, image_url=src_person.image, nicknames=src_person.nicknames, real_name=src_person.real_name, show_obj=self, thumb_url=src_person.thumb_url) if new_person and new_person.id and any(1 for pi in mc.person if pi.id == new_person.id): existing_person = next((pi for pi in mc.person if pi.id == new_person.id), None) if existing_person: try: old_person_ids.remove(existing_person.id) except KeyError: logger.log('%s - Person error: %s (%s)' % (self.name, existing_person.name, existing_person.id), logger.ERROR) pass if force: existing_person.reset(src_person) self._update_person_properties_helper(existing_person, src_person, person_ids) elif None is not new_person: mc.person.append(new_person) if old_person_ids: mc.person = [cp for cp in mc.person if cp.id not in old_person_ids] mc.update_properties( name=c.name, image_url=c.image, thumb_url=c.thumb_url, ids=dict(chain.from_iterable( iteritems(d) for d in (mc.ids, ({}, {self.tvid: c.id})[None is not c.id])))) else: persons = [] for s_pers in c.person: existing_person = next((_p for _c in cast_list for _p in _c.person if (s_pers.id and _p.ids.get(self.tvid) == s_pers.id) or (not s_pers.id and s_pers.name == _p.name)), None) # type: Optional[Person] if s_pers.id: person_ids = {self.tvid: s_pers.id} else: person_ids = {} if existing_person: if force: existing_person.reset(s_pers) self._update_person_properties_helper(existing_person, s_pers, person_ids) persons.append(existing_person) else: tmp_char = Character(c.name, ids=({}, {self.tvid: c.id})[None is not c.id], image_url=c.image, show_obj=self, thumb_url=c.thumb_url, tmp=True) new_person = Person( s_pers.name, s_pers.gender, s_pers.birthdate, s_pers.deathdate, s_pers.bio, akas=s_pers.akas, deathplace=s_pers.deathplace, height=s_pers.height, ids=({}, {self.tvid: s_pers.id})[None is not s_pers.id], image_url=s_pers.image, nicknames=s_pers.nicknames, real_name=s_pers.real_name, show_obj=self, thumb_url=s_pers.thumb_url, tmp_character_obj=tmp_char ) if force: new_person.reset(s_pers) self._update_person_properties_helper(new_person, s_pers, person_ids) persons.append(new_person) mc = Character(c.name, ids=({}, {self.tvid: c.id})[None is not c.id], image_url=c.image, person=persons, show_obj=self, thumb_url=c.thumb_url) cast_list.append(mc) cast_ordered.append(mc) if stop_event and stop_event.is_set(): return if remove_char_ids: [c.remove_all_img(include_person=True) for c in cast_list or [] if c.id in remove_char_ids] self.cast_list = cast_ordered self._save_cast_list(force=force, removed_char_ids=remove_char_ids, stop_event=stop_event, cast_list=cast_ordered) def load_imdb_info(self): if not sickgear.USE_IMDB_INFO: return logger.log('Retrieving show info [%s] from IMDb' % self._name, logger.DEBUG) try: self._get_imdb_info() except (BaseException, Exception) as e: logger.log('Error loading IMDb info: %s' % ex(e), logger.ERROR) logger.log('%s' % traceback.format_exc(), logger.ERROR) @staticmethod def check_imdb_redirect(imdb_id): # type: (Union[AnyStr, integer_types]) -> Optional[AnyStr] """ :param imdb_id: imdb id """ page_url = 'https://www.imdb.com/title/{0}/'.format(imdb_id) try: response = requests.head(page_url, allow_redirects=True) if response.history and any([h for h in response.history if 301 == h.status_code]): return helpers.parse_imdb_id(response.url) except (BaseException, Exception): pass def _get_imdb_info(self, retry=False): # type: (bool) -> None if not self._imdbid and 0 >= self.ids.get(indexermapper.TVINFO_IMDB, {'id': 0}).get('id', 0): return imdb_info = {'imdb_id': self._imdbid or 'tt%07d' % self.ids[indexermapper.TVINFO_IMDB]['id'], 'title': '', 'year': '', 'akas': '', 'runtimes': try_int(self._runtime, None), 'is_mini_series': False, 'episode_count': None, 'genres': '', 'countries': '', 'country_codes': '', 'certificates': '', 'rating': '', 'votes': '', 'last_update': ''} imdb_id = None imdb_certificates = None try: imdb_id = str(self._imdbid or 'tt%07d' % self.ids[indexermapper.TVINFO_IMDB]['id']) redirect_check = self.check_imdb_redirect(imdb_id) if redirect_check: self._imdbid = redirect_check imdb_id = redirect_check imdb_info['imdb_id'] = self.imdbid i = imdbpie.Imdb(exclude_episodes=True, cachedir=ek.ek(os.path.join, sickgear.CACHE_DIR, 'imdb-pie')) if not helpers.parse_imdb_id(imdb_id): logger.log('Not a valid imdbid: %s for show: %s' % (imdb_id, self._name), logger.WARNING) return imdb_ratings = i.get_title_ratings(imdb_id=imdb_id) imdb_akas = i.get_title_versions(imdb_id=imdb_id) imdb_tv = i.get_title_auxiliary(imdb_id=imdb_id) ipie = getattr(imdbpie.__dict__.get('imdbpie'), '_SIMPLE_GET_ENDPOINTS', None) if ipie: ipie.update({ u'get_title_certificates': u'/title/{imdb_id}/certificates', u'get_title_parentalguide': u'/title/{imdb_id}/parentalguide', }) imdb_certificates = i.get_title_certificates(imdb_id=imdb_id) except LookupError as e: if 'Title was an episode' in ex(e) and imdb_id == 'tt%07d' % self.ids[indexermapper.TVINFO_IMDB]['id']: self.ids[indexermapper.TVINFO_IMDB]['id'] = 0 self.ids[indexermapper.TVINFO_IMDB]['status'] = MapStatus.NOT_FOUND if datetime.date.today() != self.ids[indexermapper.TVINFO_IMDB]['date']: indexermapper.map_indexers_to_show(self, force=True) if not retry and imdb_id != 'tt%07d' % self.ids[indexermapper.TVINFO_IMDB]['id']: # add retry arg to prevent endless loops logger.log('imdbid: %s not found. retrying with newly found id: %s' % (imdb_id, 'tt%07d' % self.ids[indexermapper.TVINFO_IMDB]['id']), logger.DEBUG) self._get_imdb_info(retry=True) return logger.log('imdbid: %s not found. Error: %s' % (imdb_id, ex(e)), logger.WARNING) return except ImdbAPIError as e: logger.log('Imdb API Error: %s' % ex(e), logger.WARNING) return except (BaseException, Exception) as e: logger.log('Error: %s retrieving imdb id: %s' % (ex(e), imdb_id), logger.WARNING) return # ratings if isinstance(imdb_ratings.get('rating'), (int, float)): imdb_info['rating'] = try_float(imdb_ratings.get('rating'), '') if isinstance(imdb_ratings.get('ratingCount'), int): imdb_info['votes'] = try_int(imdb_ratings.get('ratingCount'), '') en_cc = ['GB', 'US', 'CA', 'AU'] # akas if isinstance(imdb_akas.get('alternateTitles'), (list, tuple)): akas_head = OrderedDict([(k, None) for k in en_cc]) akas_tail = [] for t in imdb_akas.get('alternateTitles'): if isinstance(t, dict) and t.get('title') and t.get('region'): cc = t.get('region').upper() cc_aka = '%s::%s' % (cc, t.get('title')) if cc in akas_head: akas_head[cc] = cc_aka else: akas_tail += [cc_aka] imdb_info['akas'] = '|'.join([aka for aka in itervalues(akas_head) if aka] + sorted(akas_tail)) # tv if isinstance(imdb_tv.get('title'), string_types): imdb_info['title'] = imdb_tv.get('title') if isinstance(imdb_tv.get('year'), (int, string_types)): imdb_info['year'] = try_int(imdb_tv.get('year'), '') if isinstance(imdb_tv.get('runningTimes'), list): try: for _t in imdb_tv.get('runningTimes'): try: if isinstance(_t.get('attributes'), list) and \ any(1 for _a in _t.get('attributes') if 'entire' in _a): continue except (BaseException, Exception): continue imdb_info['runtimes'] = try_int(_t.get('timeMinutes'), '') break except (BaseException, Exception): pass if isinstance(imdb_tv.get('titleType'), string_types): imdb_info['is_mini_series'] = 'mini' in imdb_tv.get('titleType').lower() if isinstance(imdb_tv.get('numberOfEpisodes'), (int, string_types)): imdb_info['episode_count'] = try_int(imdb_tv.get('numberOfEpisodes'), 1) if isinstance(imdb_tv.get('genres'), (list, tuple)): imdb_info['genres'] = '|'.join(filter_iter(lambda _v: _v, imdb_tv.get('genres'))) if isinstance(imdb_tv.get('origins'), list): imdb_info['country_codes'] = '|'.join(filter_iter(lambda _v: _v, imdb_tv.get('origins'))) # certificate if isinstance(imdb_certificates.get('certificates'), dict): certs_head = OrderedDict([(k, None) for k in en_cc]) certs_tail = [] for cc, values in iteritems(imdb_certificates.get('certificates')): if cc and isinstance(values, (list, tuple)): for cert in values: if isinstance(cert, dict) and cert.get('certificate'): extra_info = '' if isinstance(cert.get('attributes'), list): extra_info = ' (%s)' % ', '.join(cert.get('attributes')) cc = cc.upper() cc_cert = '%s:%s%s' % (cc, cert.get('certificate'), extra_info) if cc in certs_head: certs_head[cc] = cc_cert else: certs_tail += [cc_cert] imdb_info['certificates'] = '|'.join([cert for cert in itervalues(certs_head) if cert] + sorted(certs_tail)) if (not imdb_info['certificates'] and isinstance(imdb_tv.get('certificate'), dict) and isinstance(imdb_tv.get('certificate').get('certificate'), string_types)): imdb_info['certificates'] = '%s:%s' % (u'US', imdb_tv.get('certificate').get('certificate')) imdb_info['last_update'] = datetime.date.today().toordinal() # Rename dict keys without spaces for DB upsert self.imdb_info = dict( [(k.replace(' ', '_'), k(v) if hasattr(v, 'keys') else v) for k, v in iteritems(imdb_info)]) logger.log('%s: Obtained info from IMDb -> %s' % (self.tvid_prodid, self._imdb_info), logger.DEBUG) logger.log('%s: Parsed latest IMDb show info for [%s]' % (self.tvid_prodid, self._name)) def next_episode(self): logger.log('%s: Finding the episode which airs next for: %s' % (self.tvid_prodid, self._name), logger.DEBUG) cur_date = datetime.date.today().toordinal() if not self.nextaired or self.nextaired and cur_date > self.nextaired: my_db = db.DBConnection() # noinspection SqlRedundantOrderingDirection sql_result = my_db.select( """ SELECT airdate, season, episode FROM tv_episodes WHERE indexer = ? AND showid = ? AND airdate >= ? AND status in (?,?,?) ORDER BY airdate ASC LIMIT 1 """, [self.tvid, self.prodid, datetime.date.today().toordinal(), UNAIRED, WANTED, FAILED]) if None is sql_result or 0 == len(sql_result): logger.log('%s: No episode found... need to implement a show status' % self.tvid_prodid, logger.DEBUG) self.nextaired = '' else: logger.log('%s: Found episode %sx%s' % ( self.tvid_prodid, sql_result[0]['season'], sql_result[0]['episode']), logger.DEBUG) self.nextaired = sql_result[0]['airdate'] return self.nextaired def delete_show(self, full=False): # type: (bool) -> None """ :param full: """ try: sickgear.people_queue_scheduler.action.abort_cast_update(show_obj=self) except (BaseException, Exception): pass try: sickgear.show_queue_scheduler.action.abort_show(show_obj=self) except (BaseException, Exception): pass try: sickgear.search_queue_scheduler.action.abort_show(show_obj=self) except (BaseException, Exception): pass sql_l = [['DELETE FROM tv_episodes WHERE indexer = ? AND showid = ?', [self.tvid, self.prodid]], ['DELETE FROM tv_shows WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]], ['DELETE FROM imdb_info WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]], ['DELETE FROM xem_refresh WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]], ['DELETE FROM scene_numbering WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]], ['DELETE FROM allowlist WHERE indexer = ? AND show_id = ?', [self.tvid, self.prodid]], ['DELETE FROM blocklist WHERE indexer = ? AND show_id = ?', [self.tvid, self.prodid]], ['DELETE FROM indexer_mapping WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]], ['DELETE FROM tv_shows_not_found WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]], ['DELETE FROM castlist WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid]] ] + self.orphaned_cast_sql() my_db = db.DBConnection() my_db.mass_action(sql_l) self.remove_character_images() name_cache.remove_from_namecache(self.tvid, self.prodid) try: sickgear.name_parser.parser.name_parser_cache.flush(self) except (BaseException, Exception): pass action = ('delete', 'trash')[sickgear.TRASH_REMOVE_SHOW] # remove self from show list sickgear.showList = filter_list(lambda so: so.tvid_prodid != self.tvid_prodid, sickgear.showList) try: del sickgear.showDict[self.sid_int] except (BaseException, Exception): pass sickgear.webserve.Home.make_showlist_unique_names() sickgear.MEMCACHE['history_tab'] = sickgear.webserve.History.menu_tab(sickgear.MEMCACHE['history_tab_limit']) try: tvid_prodid = self.tvid_prodid if tvid_prodid in sickgear.switched_shows: sickgear.switched_shows.pop(tvid_prodid) elif tvid_prodid in itervalues(sickgear.switched_shows): sickgear.switched_shows = {k: v for k, v in iteritems(sickgear.switched_shows) if tvid_prodid != v} except (BaseException, Exception): pass # clear the cache ic = image_cache.ImageCache() for cache_obj in ek.ek(glob.glob, ic.fanart_path(self.tvid, self.prodid).replace('fanart.jpg', '*')) \ + ek.ek(glob.glob, ic.poster_thumb_path(self.tvid, self.prodid).replace('poster.jpg', '*')) \ + ek.ek(glob.glob, ic.poster_path(self.tvid, self.prodid).replace('poster.jpg', '*')): cache_dir = ek.ek(os.path.isdir, cache_obj) result = helpers.remove_file(cache_obj, tree=cache_dir, log_level=logger.WARNING) if result: logger.log('%s cache %s %s' % (result, cache_dir and 'dir' or 'file', cache_obj)) if self.tvid_prodid in sickgear.FANART_RATINGS: del sickgear.FANART_RATINGS[self.tvid_prodid] # remove entire show folder if full: try: logger.log('Attempt to %s show folder %s' % (action, self._location)) # check first the read-only attribute file_attribute = ek.ek(os.stat, self.location)[0] if not file_attribute & stat.S_IWRITE: # File is read-only, so make it writeable logger.log('Attempting to make writeable the read only folder %s' % self._location, logger.DEBUG) try: ek.ek(os.chmod, self.location, stat.S_IWRITE) except (BaseException, Exception): logger.log('Unable to change permissions of %s' % self._location, logger.WARNING) result = helpers.remove_file(self.location, tree=True) if result: logger.log('%s show folder %s' % (result, self._location)) except exceptions_helper.ShowDirNotFoundException: logger.log('Show folder does not exist, no need to %s %s' % (action, self._location), logger.WARNING) except OSError as e: logger.log('Unable to %s %s: %s / %s' % (action, self._location, repr(e), ex(e)), logger.WARNING) def populate_cache(self, force=False): # type: (bool) -> None """ :param force: """ cache_inst = image_cache.ImageCache() logger.log('Checking & filling cache for show %s' % self._name) cache_inst.fill_cache(self, force) def refresh_dir(self): # make sure the show dir is where we think it is unless dirs are created on the fly if not ek.ek(os.path.isdir, self._location) and not sickgear.CREATE_MISSING_SHOW_DIRS: return False # load from dir self.load_episodes_from_dir() # run through all locations from DB, check that they exist logger.log('%s: Loading all episodes for [%s] with a location from the database' % (self.tvid_prodid, self._name)) my_db = db.DBConnection() # noinspection SqlResolve sql_result = my_db.select( """ SELECT * FROM tv_episodes WHERE indexer = ? AND showid = ? AND location != '' ORDER BY season DESC, episode DESC """, [self.tvid, self.prodid]) kept = 0 deleted = 0 attempted = [] sql_l = [] for cur_row in sql_result: season = int(cur_row['season']) episode = int(cur_row['episode']) location = ek.ek(os.path.normpath, cur_row['location']) try: ep_obj = self.get_episode(season, episode, ep_result=[cur_row]) except exceptions_helper.EpisodeDeletedException: logger.log('The episode from [%s] was deleted while we were refreshing it, moving on to the next one' % self._name, logger.DEBUG) continue # if the path exist and if it's in our show dir if (self.prune and season and ep_obj.location not in attempted and 0 < helpers.get_size(ep_obj.location) and ek.ek(os.path.normpath, location).startswith(ek.ek(os.path.normpath, self.location))): with ep_obj.lock: if ep_obj.status in Quality.DOWNLOADED: # locations repeat but attempt to delete once attempted += ep_obj.location if kept >= self.prune: result = helpers.remove_file(ep_obj.location, prefix_failure=u'%s: ' % self.tvid_prodid) if result: logger.log(u'%s: %s file %s' % (self.tvid_prodid, result, ep_obj.location), logger.DEBUG) deleted += 1 else: kept += 1 # if the path doesn't exist or if it's not in our show dir if not ek.ek(os.path.isfile, location) or not ek.ek(os.path.normpath, location).startswith( ek.ek(os.path.normpath, self.location)): # check if downloaded files still exist, update our data if this has changed if 1 != sickgear.SKIP_REMOVED_FILES: with ep_obj.lock: # if it used to have a file associated with it and it doesn't anymore then set it to IGNORED if ep_obj.location and ep_obj.status in Quality.DOWNLOADED: if ARCHIVED == sickgear.SKIP_REMOVED_FILES: ep_obj.status = Quality.compositeStatus( ARCHIVED, Quality.qualityDownloaded(ep_obj.status)) else: ep_obj.status = (sickgear.SKIP_REMOVED_FILES, IGNORED)[ not sickgear.SKIP_REMOVED_FILES] logger.log( '%s: File no longer at location for s%02de%02d,' % (self.tvid_prodid, season, episode) + ' episode removed and status changed to %s' % statusStrings[ep_obj.status], logger.DEBUG) ep_obj.subtitles = list() ep_obj.subtitles_searchcount = 0 ep_obj.subtitles_lastsearch = str(datetime.datetime.min) ep_obj.location = '' ep_obj.hasnfo = False ep_obj.hastbn = False ep_obj.release_name = '' result = ep_obj.get_sql() if None is not result: sql_l.append(result) else: # the file exists, set its modify file stamp if sickgear.AIRDATE_EPISODES: ep_obj.airdate_modify_stamp() if deleted: logger.log('%s: %s %s media file%s and kept %s most recent downloads' % ( self.tvid_prodid, ('Permanently deleted', 'Trashed')[sickgear.TRASH_REMOVE_SHOW], deleted, helpers.maybe_plural(deleted), kept)) if 0 < len(sql_l): my_db = db.DBConnection() my_db.mass_action(sql_l) def download_subtitles(self, force=False): # type: (bool) -> None """ :param force: """ # TODO: Add support for force option if not ek.ek(os.path.isdir, self._location): logger.log('%s: Show directory doesn\'t exist, can\'t download subtitles' % self.tvid_prodid, logger.DEBUG) return logger.log('%s: Downloading subtitles' % self.tvid_prodid, logger.DEBUG) try: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT location FROM tv_episodes WHERE indexer = ? AND showid = ? AND LENGTH(location) != 0 ORDER BY season DESC, episode DESC """, [self.tvid, self.prodid]) for cur_row in sql_result: ep_obj = self.ep_obj_from_file(cur_row['location']) _ = ep_obj.download_subtitles(force=force) except (BaseException, Exception): logger.log('Error occurred when downloading subtitles: %s' % traceback.format_exc(), logger.ERROR) return def remove_character_images(self): try: img_obj = image_cache.ImageCache() delete_list = [] for character in self.cast_list: for p in character.person: person_img = img_obj.person_both_paths(p) for i in person_img: delete_list.append(i) character_img = img_obj.character_both_path(character, self) for i in character_img: delete_list.append(i) for d in delete_list: try: remove_file_perm(d) except (BaseException, Exception): pass except (BaseException, Exception): pass def switch_infosrc(self, old_tvid, old_prodid, pausestatus_after=None, update_show=True): # type: (integer_types, integer_types, Optional[bool], bool) -> None """ :param old_tvid: old tvid :param old_prodid: old prodid :param pausestatus_after: pause after switch :param update_show: """ with self.lock: my_db = db.DBConnection() my_db.mass_action([ ['UPDATE tv_shows SET indexer = ?, indexer_id = ? WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE tv_episodes SET indexer = ?, showid = ?, indexerid = 0 WHERE indexer = ? AND showid = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE blocklist SET indexer = ?, show_id = ? WHERE indexer = ? AND show_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE history SET indexer = ?, showid = ? WHERE indexer = ? AND showid = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE imdb_info SET indexer = ?, indexer_id = ? WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE scene_exceptions SET indexer = ?, indexer_id = ? WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE scene_numbering SET indexer = ?, indexer_id = ? WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE allowlist SET indexer = ?, show_id = ? WHERE indexer = ? AND show_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['UPDATE xem_refresh SET indexer = ?, indexer_id = ? WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]], ['DELETE FROM tv_shows_not_found WHERE indexer = ? AND indexer_id = ?', [old_tvid, old_prodid]], ['UPDATE castlist SET indexer = ?, indexer_id = ? WHERE indexer = ? AND indexer_id = ?', [self.tvid, self.prodid, old_tvid, old_prodid]] ]) my_failed_db = db.DBConnection('failed.db') my_failed_db.action('UPDATE history SET indexer = ?, showid = ? WHERE indexer = ? AND showid = ?', [self.tvid, self.prodid, old_tvid, old_prodid]) del_mapping(old_tvid, old_prodid) try: for c in self.cast_list: # type: Character c.remove_all_img(tvid=old_tvid, proid=old_prodid) except (BaseException, Exception): pass self._cast_list = None self.sxe_ep_obj = {} self.ids[old_tvid]['status'] = MapStatus.NONE self.ids[self.tvid]['status'] = MapStatus.SOURCE self.ids[self.tvid]['id'] = self.prodid if isinstance(self.imdb_info, dict): self.imdb_info['indexer'] = self.tvid self.imdb_info['indexer_id'] = self.prodid save_mapping(self) name_cache.remove_from_namecache(old_tvid, old_prodid) image_cache_dir = ek.ek(os.path.join, sickgear.CACHE_DIR, 'images', 'shows') old_dir = ek.ek(os.path.join, image_cache_dir, '%s-%s' % (old_tvid, old_prodid)) new_dir = ek.ek(os.path.join, image_cache_dir, '%s-%s' % (self.tvid, self.prodid)) try: ek.ek(os.rename, old_dir, new_dir) except (BaseException, Exception) as e: logger.log('Unable to rename %s to %s: %s / %s' % (old_dir, new_dir, repr(e), ex(e)), logger.WARNING) old_id = TVidProdid({old_tvid: old_prodid})() rating = sickgear.FANART_RATINGS.get(old_id) if rating: del sickgear.FANART_RATINGS[old_id] sickgear.FANART_RATINGS[self.tvid_prodid] = rating sickgear.save_config() name_cache.buildNameCache(self) self.reset_not_found_count() old_sid_int = self.create_sid(old_tvid, old_prodid) if old_sid_int != self.sid_int: try: del sickgear.showDict[old_sid_int] except (BaseException, Exception): pass sickgear.showDict[self.sid_int] = self self.save_to_db() if update_show: # force the update try: sickgear.show_queue_scheduler.action.updateShow( self, force=True, web=True, priority=QueuePriorities.VERYHIGH, pausestatus_after=pausestatus_after, switch_src=True) except exceptions_helper.CantUpdateException as e: logger.log('Unable to update this show. %s' % ex(e), logger.ERROR) def save_to_db(self, force_save=False): # type: (bool) -> None """ :param force_save: """ if not self.dirty and not force_save: logger.log('%s: Not saving show to db - record is not dirty' % self.tvid_prodid, logger.DEBUG) return logger.log('%s: Saving show info to database' % self.tvid_prodid, logger.DEBUG) new_value_dict = dict( air_by_date=self._air_by_date, airs=self._airs, airtime=time_to_int(self._airtime), anime=self._anime, archive_firstmatch=self._upgrade_once, classification=self._classification, dvdorder=self._dvdorder, flatten_folders=self._flatten_folders, genre=self._genre, indexer=self.tvid, lang=self._lang, imdb_id=self._imdbid, last_update_indexer=self._last_update_indexer, location=self._location, network=self.internal_network, network_country=self._network_country, network_country_code=self._network_country_code, network_id=self._network_id, network_is_stream=self._network_is_stream, overview=self._overview, paused=self._paused, prune=self._prune, quality=self._quality, rls_global_exclude_ignore=','.join(self._rls_global_exclude_ignore), rls_global_exclude_require=','.join(self._rls_global_exclude_require), rls_ignore_words=helpers.generate_word_str(self._rls_ignore_words, self._rls_ignore_words_regex), rls_require_words=helpers.generate_word_str(self._rls_require_words, self._rls_require_words_regex), runtime=self._runtime, scene=self._scene, show_name=self._name, sports=self._sports, startyear=self.startyear, status=self._status, subtitles=self._subtitles, tag=self._tag, timezone=self.internal_timezone, ) control_value_dict = dict(indexer=self.tvid, indexer_id=self.prodid) my_db = db.DBConnection() my_db.upsert('tv_shows', new_value_dict, control_value_dict) self.dirty = False if sickgear.USE_IMDB_INFO and len(self._imdb_info): new_value_dict = self._imdb_info my_db = db.DBConnection() my_db.upsert('imdb_info', new_value_dict, control_value_dict) def __ne__(self, o): # type: (TVShow) -> bool return not self.__eq__(o) def __eq__(self, o): # type: (TVShow) -> bool if not isinstance(o, TVShow): return False return o.tvid == self.tvid and o.prodid == self.prodid def __hash__(self): return hash((self.tvid, self.prodid)) def __repr__(self): return 'TVShow(%s)' % self.__str__() def __str__(self): return 'prodid: %s\n' % self.prodid \ + 'tvid: %s\n' % self.tvid \ + 'name: %s\n' % self.name \ + 'location: %s\n' % self._location \ + ('', 'network: %s\n' % self.network)[self.network not in (None, '')] \ + ('', 'airs: %s\n' % self.airs)[self.airs not in (None, '')] \ + ('', 'status: %s\n' % self.status)[self.status not in (None, '')] \ + 'startyear: %s\n' % self.startyear \ + ('', 'genre: %s\n' % self.genre)[self.genre not in (None, '')] \ + 'classification: %s\n' % self.classification \ + 'runtime: %s\n' % self.runtime \ + 'quality: %s\n' % self.quality \ + 'scene: %s\n' % self.is_scene \ + 'sports: %s\n' % self.is_sports \ + 'anime: %s\n' % self.is_anime \ + 'prune: %s\n' % self.prune def want_episode(self, season, episode, quality, manual_search=False, multi_ep=False): # type: (integer_types, integer_types, integer_types, bool, bool) -> bool """ :param season: season number :param episode: episode number :param quality: quality :param manual_search: manual search :param multi_ep: multiple episodes :return: """ logger.log('Checking if found %sepisode %sx%s is wanted at quality %s' % (('', 'multi-part ')[multi_ep], season, episode, Quality.qualityStrings[quality]), logger.DEBUG) if not multi_ep: try: wq = getattr(self.sxe_ep_obj.get(season, {}).get(episode, {}), 'wanted_quality', None) if None is not wq: if quality in wq: cur_status, cur_quality = Quality.splitCompositeStatus(self.sxe_ep_obj[season][episode].status) if cur_status in (WANTED, UNAIRED, SKIPPED, FAILED): logger.log('Existing episode status is wanted/unaired/skipped/failed,' ' getting found episode', logger.DEBUG) return True elif manual_search: logger.log('Usually ignoring found episode, but forced search allows the quality,' ' getting found episode', logger.DEBUG) return True elif quality > cur_quality: logger.log( 'Episode already exists but the found episode has better quality,' ' getting found episode', logger.DEBUG) return True logger.log('None of the conditions were met, ignoring found episode', logger.DEBUG) return False except (BaseException, Exception): pass # if the quality isn't one we want under any circumstances then just say no initial_qualities, archive_qualities = Quality.splitQuality(self._quality) all_qualities = list(set(initial_qualities + archive_qualities)) initial = '= (%s)' % ','.join([Quality.qualityStrings[qual] for qual in initial_qualities]) if 0 < len(archive_qualities): initial = '+ upgrade to %s + (%s)'\ % (initial, ','.join([Quality.qualityStrings[qual] for qual in archive_qualities])) logger.log('Want initial %s and found %s' % (initial, Quality.qualityStrings[quality]), logger.DEBUG) if quality not in all_qualities: logger.log('Don\'t want this quality, ignoring found episode', logger.DEBUG) return False my_db = db.DBConnection() sql_result = my_db.select( """ SELECT status FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ? """, [self.tvid, self.prodid, season, episode]) if not sql_result or not len(sql_result): logger.log('Unable to find a matching episode in database, ignoring found episode', logger.DEBUG) return False cur_status, cur_quality = Quality.splitCompositeStatus(int(sql_result[0]['status'])) ep_status_text = statusStrings[cur_status] logger.log('Existing episode status: %s (%s)' % (statusStrings[cur_status], ep_status_text), logger.DEBUG) # if we know we don't want it then just say no if cur_status in [IGNORED, ARCHIVED] + ([SKIPPED], [])[multi_ep] and not manual_search: logger.log('Existing episode status is %signored/archived, ignoring found episode' % ('skipped/', '')[multi_ep], logger.DEBUG) return False # if it's one of these then we want it as long as it's in our allowed initial qualities if quality in all_qualities: if cur_status in [WANTED, UNAIRED, SKIPPED, FAILED] + ([], SNATCHED_ANY)[multi_ep]: logger.log('Existing episode status is wanted/unaired/skipped/failed, getting found episode', logger.DEBUG) return True elif manual_search: logger.log( 'Usually ignoring found episode, but forced search allows the quality, getting found episode', logger.DEBUG) return True else: logger.log('Quality is on wanted list, need to check if it\'s better than existing quality', logger.DEBUG) downloaded_status_list = SNATCHED_ANY + [DOWNLOADED] # special case: already downloaded quality is not in any of the wanted Qualities if cur_status in downloaded_status_list and cur_quality not in all_qualities: wanted_qualities = all_qualities else: wanted_qualities = archive_qualities # if re-downloading then only keep items in the archiveQualities list and better than what we have if cur_status in downloaded_status_list and quality in wanted_qualities and quality > cur_quality: logger.log('Episode already exists but the found episode has better quality, getting found episode', logger.DEBUG) return True else: logger.log('Episode already exists and the found episode has same/lower quality, ignoring found episode', logger.DEBUG) logger.log('None of the conditions were met, ignoring found episode', logger.DEBUG) return False def get_overview(self, ep_status, split_snatch=False): # type: (integer_types, bool) -> integer_types """ :param ep_status: episode status :param split_snatch: :return: :rtype: int """ return helpers.get_overview(ep_status, self.quality, self.upgrade_once, split_snatch=split_snatch) def __getstate__(self): d = dict(self.__dict__) del d['lock'] return d def __setstate__(self, d): d['lock'] = threading.RLock() self.__dict__.update(d) def __bool__(self): return bool(self.tvid) and bool(self.prodid) __nonzero__ = __bool__ class TVEpisode(TVEpisodeBase): def __init__(self, show_obj, season, episode, path='', existing_only=False, show_result=None): # type: (TVShow, integer_types, integer_types, AnyStr, bool, List) -> None super(TVEpisode, self).__init__(season, episode, int(show_obj.tvid)) self._airtime = None # type: Optional[datetime.time] self._epid = 0 # type: int self._location = path # type: AnyStr self._network = None # type: Optional[AnyStr] self._network_country = None # type: Optional[AnyStr] self._network_country_code = None # type: Optional[AnyStr] self._network_id = None # type: Optional[int] self._network_is_stream = None # type: Optional[bool] self._runtime = 0 # type: integer_types self._show_obj = show_obj # type: TVShow self._timestamp = None # type: Optional[int] self._timezone = None # type: Optional[AnyStr] self._tvid = int(show_obj.tvid) # type: int self.check_for_meta_files() self.lock = threading.RLock() self.related_ep_obj = [] # type: List self.scene_absolute_number = 0 # type: int self.scene_episode = 0 # type: int self.scene_season = 0 # type: int self.specify_episode(self._season, self._episode, existing_only=existing_only, show_result=show_result) self.wanted_quality = [] # type: List @property def network(self): return self._network @network.setter def network(self, *args): self.dirty_setter('_network')(self, *args) @property def network_id(self): return self._network_id @network_id.setter def network_id(self, *args): self.dirty_setter('_network_id')(self, *args) @property def network_country(self): return self._network_country @network_country.setter def network_country(self, *args): self.dirty_setter('_network_country')(self, *args) @property def network_country_code(self): return self._network_country_code @network_country_code.setter def network_country_code(self, *args): self.dirty_setter('_network_country_code')(self, *args) @property def network_is_stream(self): return self._network_is_stream @network_is_stream.setter def network_is_stream(self, *args): self.dirty_setter('_network_is_stream')(self, *args) @property def airtime(self): # type: (...) -> Optional[datetime.time] return self._airtime @airtime.setter def airtime(self, *args): # type: (Optional[datetime.time, AnyStr]) -> None self.dirty_setter('_airtime')(self, *args) @property def runtime(self): return self._runtime @runtime.setter def runtime(self, *args): self.dirty_setter('_runtime')(self, *args) @property def timezone(self): return self._timezone @timezone.setter def timezone(self, *args): self.dirty_setter('_timezone')(self, *args) @property def timestamp(self): return self._timestamp @timestamp.setter def timestamp(self, *args): self.dirty_setter('_timestamp')(self, *args) @property def show_obj(self): """ :return: TVShow object :rtype: TVShow """ return self._show_obj @show_obj.setter def show_obj(self, val): self._show_obj = val @property def tvid(self): """ :rtype: int """ return self._tvid @tvid.setter def tvid(self, val): self.dirty_setter('_tvid')(self, int(val)) # TODO: remove the following when indexerid is gone # in deprecation transition, tvid will also set indexer so that existing uses continue to work normally self.dirty_setter('_indexer')(self, int(val)) @property def epid(self): """ :rtype: int or long """ return self._epid @epid.setter def epid(self, val): self.dirty_setter('_epid')(self, int(val)) # TODO: remove the following when indexerid is gone # in deprecation transition, epid will also set indexerid so that existing uses continue as normal self.dirty_setter('_indexerid')(self, int(val)) def _set_location(self, val): log_vals = (('clears', ''), ('sets', ' to ' + val))[any(val)] # noinspection PyStringFormat logger.log(u'Setter %s location%s' % log_vals, logger.DEBUG) # self._location = newLocation self.dirty_setter('_location')(self, val) if val and ek.ek(os.path.isfile, val): self.file_size = ek.ek(os.path.getsize, val) else: self.file_size = 0 location = property(lambda self: self._location, _set_location) def refresh_subtitles(self): """Look for subtitles files and refresh the subtitles property""" if sickgear.USE_SUBTITLES: self.subtitles = subtitles.subtitles_languages(self.location) def download_subtitles(self, force=False): """ :param force: :type force: bool :return: :rtype: """ if not sickgear.USE_SUBTITLES: return # TODO: Add support for force option if not ek.ek(os.path.isfile, self.location): logger.log('%s: Episode file doesn\'t exist, can\'t download subtitles for episode %sx%s' % (self.show_obj.tvid_prodid, self.season, self.episode), logger.DEBUG) return logger.log('%s: Downloading subtitles for episode %sx%s' % (self.show_obj.tvid_prodid, self.season, self.episode), logger.DEBUG) previous_subtitles = self.subtitles try: need_languages = list(set(sickgear.SUBTITLES_LANGUAGES) - set(self.subtitles)) subs = subliminal.download_subtitles([self.location], languages=need_languages, services=sickgear.subtitles.get_enabled_service_list(), force=force, multi=True, cache_dir=sickgear.CACHE_DIR, os_auth=sickgear.SUBTITLES_SERVICES_AUTH[0], os_hash=sickgear.SUBTITLES_OS_HASH) if sickgear.SUBTITLES_DIR: for video in subs: subs_new_path = ek.ek(os.path.join, ek.ek(os.path.dirname, video.path), sickgear.SUBTITLES_DIR) dir_exists = helpers.make_dir(subs_new_path) if not dir_exists: logger.log('Unable to create subtitles folder %s' % subs_new_path, logger.ERROR) else: helpers.chmod_as_parent(subs_new_path) for subtitle in subs.get(video): new_file_path = ek.ek(os.path.join, subs_new_path, ek.ek(os.path.basename, subtitle.path)) helpers.move_file(subtitle.path, new_file_path) helpers.chmod_as_parent(new_file_path) else: for video in subs: for subtitle in subs.get(video): helpers.chmod_as_parent(subtitle.path) except (BaseException, Exception): logger.log('Error occurred when downloading subtitles: %s' % traceback.format_exc(), logger.ERROR) return self.refresh_subtitles() # added the if because sometime it raises an error self.subtitles_searchcount = self.subtitles_searchcount + 1 if self.subtitles_searchcount else 1 self.subtitles_lastsearch = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') self.save_to_db() newsubtitles = set(self.subtitles).difference(set(previous_subtitles)) if newsubtitles: try: subtitle_list = ", ".join([subliminal.language.Language(x).name for x in newsubtitles]) except (BaseException, Exception): logger.log('Could not parse a language to use to fetch subtitles for episode %sx%s' % (self.season, self.episode), logger.DEBUG) return logger.log('%s: Downloaded %s subtitles for episode %sx%s' % (self.show_obj.tvid_prodid, subtitle_list, self.season, self.episode), logger.DEBUG) notifiers.notify_subtitle_download(self, subtitle_list) else: logger.log('%s: No subtitles downloaded for episode %sx%s' % (self.show_obj.tvid_prodid, self.season, self.episode), logger.DEBUG) if sickgear.SUBTITLES_HISTORY: for video in subs: for subtitle in subs.get(video): history.log_subtitle(self.show_obj.tvid, self.show_obj.prodid, self.season, self.episode, self.status, subtitle) return subs def check_for_meta_files(self): """ :return: :rtype: bool """ oldhasnfo = self._hasnfo oldhastbn = self._hastbn hasnfo = False hastbn = False # check for nfo and tbn if ek.ek(os.path.isfile, self.location): for cur_provider in itervalues(sickgear.metadata_provider_dict): if cur_provider.episode_metadata: new_result = cur_provider.has_episode_metadata(self) else: new_result = False hasnfo = new_result or hasnfo if cur_provider.episode_thumbnails: new_result = cur_provider.has_episode_thumb(self) else: new_result = False hastbn = new_result or hastbn self.hasnfo = hasnfo self.hastbn = hastbn # if either setting has changed return true, if not return false return oldhasnfo != self._hasnfo or oldhastbn != self._hastbn def specify_episode(self, season, episode, existing_only=False, **kwargs): """ kwargs['show_result']: type: Optional[List[Row]] passed thru :param season: season number :type season: int :param episode: episode number :type episode: int :param existing_only: :type existing_only: bool """ if not self.load_from_db(season, episode, **kwargs): # only load from NFO if we didn't load from DB if ek.ek(os.path.isfile, self.location): try: self.load_from_nfo(self.location) except exceptions_helper.NoNFOException: logger.log('%s: There was an error loading the NFO for episode %sx%s' % (self.show_obj.tvid_prodid, season, episode), logger.ERROR) pass # if we tried loading it from NFO and didn't find the NFO, try the Indexers if not self._hasnfo: try: self.load_from_tvinfo(season, episode) except exceptions_helper.EpisodeDeletedException: # if we failed SQL *and* NFO, Indexers then fail raise exceptions_helper.EpisodeNotFoundException( 'Couldn\'t find episode %sx%s' % (season, episode)) elif existing_only: raise exceptions_helper.EpisodeNotFoundException( 'Couldn\'t find episode %sx%s' % (season, episode)) def load_from_db(self, season, episode, show_result=None, **kwargs): # type: (int, int, Optional[List[Row]], Any) -> bool """ kwargs['scene_result']: type: Optional[List[Row]] passed thru :param season: season number :param episode: episode number :param show_result: """ logger.log('%s: Loading episode details from DB for episode %sx%s' % (self._show_obj.tvid_prodid, season, episode), logger.DEBUG) show_result = show_result and next(iter(show_result), None) if not show_result or episode != show_result['episode'] or season != show_result['season']: my_db = db.DBConnection() sql_result = my_db.select( """ SELECT * FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ? LIMIT 2 """, [self._show_obj.tvid, self._show_obj.prodid, season, episode]) if 1 != len(sql_result): if len(sql_result): raise exceptions_helper.MultipleDBEpisodesException('DB has multiple records for the same show') logger.log('%s: Episode %sx%s not found in the database' % (self._show_obj.tvid_prodid, self._season, self._episode), logger.DEBUG) return False show_result = next(iter(sql_result)) # todo: change to _tvid , _epid after removing indexer, indexerid self._tvid = int(show_result['indexer']) self._epid = int(show_result['indexerid']) self._absolute_number = show_result['absolute_number'] self._airdate = datetime.date.fromordinal(int(show_result['airdate'])) self._airtime = int_to_time(try_int(show_result['airtime'], None)) self._description = self._description if not show_result['description'] else show_result['description'] self._episode = episode self._file_size = try_int(show_result['file_size']) self._is_proper = self._is_proper if not show_result['is_proper'] else try_int(show_result['is_proper']) self._name = self._name if not show_result['name'] else show_result['name'] self._network = show_result['network'] or self._show_obj.internal_network self._network_country = show_result['network_country'] self._network_country_code = show_result['network_country_code'] self._network_id = show_result['network_id'] self._network_is_stream = bool(show_result['network_is_stream']) self._runtime = show_result['runtime'] self._season = season self._status = self._status if None is show_result['status'] else int(show_result['status']) subt_value = show_result['subtitles'] if ',,' in subt_value: # a defensive fix for a data issue subt_value = re.sub(r',+', '', subt_value) subt_value = ('', subt_value)[2 == len(subt_value)] self._subtitles = subt_value and subt_value.split(',') or [] self._subtitles_lastsearch = show_result['subtitles_lastsearch'] self._subtitles_searchcount = show_result['subtitles_searchcount'] self._timestamp = show_result['timestamp'] or self._make_timestamp() self._version = self._version if not show_result['version'] else int(show_result['version']) self.location = show_result['location'] and ek.ek(os.path.normpath, show_result['location']) or self.location if None is not show_result['release_group']: self._release_group = show_result['release_group'] if None is not show_result['release_name']: self._release_name = show_result['release_name'] sickgear.scene_numbering.xem_refresh(self._show_obj.tvid, self._show_obj.prodid) self.scene_absolute_number = try_int(show_result['scene_absolute_number']) if 0 == self.scene_absolute_number: self.scene_absolute_number = sickgear.scene_numbering.get_scene_absolute_numbering( self._show_obj.tvid, self._show_obj.prodid, absolute_number=self._absolute_number, season=self._season, episode=episode, show_result=show_result, show_obj=self._show_obj, **kwargs) self.scene_season = try_int(show_result['scene_season']) self.scene_episode = try_int(show_result['scene_episode']) if 0 == self.scene_season or 0 == self.scene_episode: self.scene_season, self.scene_episode = sickgear.scene_numbering.get_scene_numbering( self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode, show_obj=self._show_obj, show_result=show_result, **kwargs) self._timezone = show_result['timezone'] if not self._timezone and (self._network or self._show_obj.internal_timezone): if self._show_obj.internal_timezone: self._timezone = self._show_obj.internal_timezone else: _, self.timezone = network_timezones.get_network_timezone(self._network, return_name=True) self.dirty = False return True # noinspection PyProtectedMember def _make_timestamp(self): # type: (...) -> Optional[integer_types] if self.airdate: if isinstance(self._airtime, datetime.time): ep_time = self._airtime elif isinstance(self._show_obj._airtime, datetime.time): ep_time = self._show_obj._airtime else: ep_time = datetime.time(hour=0, minute=0) tzinfo = None if isinstance(self._timezone, string_types) and self._timezone: tzinfo = tz.gettz(self._timezone, zoneinfo_priority=True) elif isinstance(self._network, string_types) and self._network: tzinfo = tz.gettz(self._network, zoneinfo_priority=True) elif isinstance(self._show_obj.timezone, string_types) and self._show_obj.timezone: tzinfo = self._show_obj.timezone elif isinstance(self._show_obj.network, string_types) and self._show_obj.network: tzinfo = network_timezones.get_network_timezone(self._show_obj.network) if PY2: return SGDatetime.combine(self.airdate, ep_time).replace(tzinfo=tzinfo).timestamp_far() return SGDatetime.combine(self.airdate, ep_time, tzinfo=tzinfo).timestamp_far() return None def load_from_tvinfo( self, season=None, # type: integer_types episode=None, # type: integer_types cache=True, # type: bool tvapi=None, # type: Any cached_season=None, # type: Dict update=False, # type: bool cached_show=None, # type: Dict switch=False, # type: bool old_tvid=None, # type: int old_prodid=None, # type: integer_types switch_list=None # type: List ): # type: (...) -> Optional[bool] """ :param season: season number :param episode: episode number :param cache: :param tvapi: :param cached_season: :param update: :param cached_show: :param switch: :param old_tvid: :param old_prodid: :param switch_list: """ if None is season: season = self._season if None is episode: episode = self._episode logger.log('%s: Loading episode details from %s for episode %sx%s' % (self._show_obj.tvid_prodid, sickgear.TVInfoAPI(self._show_obj.tvid).name, season, episode), logger.DEBUG) try: if cached_show: ep_info = cached_show[season][episode] elif None is cached_season: if None is tvapi: tvinfo_config = sickgear.TVInfoAPI(self.tvid).api_params.copy() if not cache: tvinfo_config['cache'] = False show_lang = self._show_obj.lang if show_lang: tvinfo_config['language'] = show_lang if 0 != self._show_obj.dvdorder: tvinfo_config['dvdorder'] = True t = sickgear.TVInfoAPI(self.tvid).setup(**tvinfo_config) else: t = tvapi ep_info = t.get_show(self._show_obj.prodid, language=self.show_obj.lang)[season][episode] # type: TVInfoEpisode else: ep_info = cached_season[episode] # type: TVInfoEpisode except (BaseTVinfoEpisodenotfound, BaseTVinfoSeasonnotfound): logger.log('Unable to find the episode on %s... has it been removed? Should I delete from db?' % sickgear.TVInfoAPI(self.tvid).name, logger.DEBUG) # if I'm no longer on the Indexers but I once was then delete myself from the DB if -1 != self._epid and helpers.should_delete_episode(self._status): self.delete_episode() elif UNKNOWN == self._status: self.status = SKIPPED return except (BaseTVinfoError, IOError) as e: logger.log('%s threw up an error: %s' % (sickgear.TVInfoAPI(self.tvid).name, ex(e)), logger.DEBUG) # if the episode is already valid just log it, if not throw it up if UNKNOWN == self._status: self.status = SKIPPED if self._name: logger.log('%s timed out but there is enough info from other sources, allowing the error' % sickgear.TVInfoAPI(self.tvid).name, logger.DEBUG) return logger.log('%s timed out, unable to create the episode' % sickgear.TVInfoAPI(self.tvid).name, logger.ERROR) return False if getattr(ep_info, 'absolute_number', None) in (None, ''): logger.debug('This episode (%s - %sx%s) has no absolute number on %s' % (self.show_obj.unique_name, season, episode, sickgear.TVInfoAPI(self.tvid).name)) else: logger.log('%s: The absolute_number for %sx%s is : %s' % (self._show_obj.tvid_prodid, season, episode, ep_info['absolute_number']), logger.DEBUG) self.absolute_number = int(ep_info['absolute_number']) if switch and None is not switch_list: if self._name != self.dict_prevent_nonetype(ep_info, 'episodename'): switch_list.append(self.show_obj.switch_ep_change_sql(old_tvid, old_prodid, episode, season, TVSWITCH_EP_RENAMED)) self.name = self.dict_prevent_nonetype(ep_info, 'episodename') self.season = season self.episode = episode sickgear.scene_numbering.xem_refresh(self._show_obj.tvid, self._show_obj.prodid) self.scene_absolute_number = sickgear.scene_numbering.get_scene_absolute_numbering( self._show_obj.tvid, self._show_obj.prodid, absolute_number=self._absolute_number, season=self._season, episode=self._episode, show_obj=self._show_obj) self.scene_season, self.scene_episode = sickgear.scene_numbering.get_scene_numbering( self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode, show_obj=self._show_obj) self.description = self.dict_prevent_nonetype(ep_info, 'overview') firstaired = getattr(ep_info, 'firstaired', None) if None is firstaired or firstaired in '0000-00-00': firstaired = str(datetime.date.fromordinal(1)) raw_airdate = [int(x) for x in firstaired.split('-')] old_airdate_future = self._airdate == datetime.date.fromordinal(1) or self._airdate >= datetime.date.today() try: self.airdate = datetime.date(raw_airdate[0], raw_airdate[1], raw_airdate[2]) except (ValueError, IndexError): logger.error('Malformed air date retrieved from %s (%s - %sx%s)' % (sickgear.TVInfoAPI(self.tvid).name, self.show_obj.unique_name, season, episode)) # if I'm incomplete on TVDB but I once was complete then just delete myself from the DB for now if -1 != self._epid and helpers.should_delete_episode(self._status): self.delete_episode() elif UNKNOWN == self._status: self.status = SKIPPED return False self.network = ep_info.network if not self._network: self.network = self._show_obj.internal_network self.network_id = ep_info.network_id self.network_country = ep_info.network_country self.network_country_code = ep_info.network_country_code self.timezone = ep_info.network_timezone if not self._timezone and (self._network or self._show_obj.internal_timezone): if self._show_obj.internal_timezone: self._timezone = self._show_obj.internal_timezone else: _, self.timezone = network_timezones.get_network_timezone(self._network, return_name=True) self.network_is_stream = ep_info.network_is_stream self.airtime = ep_info.airtime self.runtime = ep_info.runtime invalid_date = False if not (self._airdate and self._airdate < invalid_date_limit): self.timestamp = ep_info.timestamp or self._make_timestamp() else: self.timestamp = None invalid_date = True today = datetime.date.today() delta = datetime.timedelta(days=1) if invalid_date: show_time = None elif self._timestamp and self._timezone: try: show_time = SGDatetime.from_timestamp(self._timestamp, tz_aware=True, local_time=False, tzinfo=tz.gettz(self._timezone)) except OverflowError: logger.debug('Invalid timestamp: %s, using fallback' % self._timestamp) show_time = network_timezones.parse_date_time(self._airdate.toordinal(), self._airtime or self._show_obj.airs, self._network or self._show_obj.network) else: show_time = network_timezones.parse_date_time(self._airdate.toordinal(), self._airtime or self._show_obj.airs, self._network or self._show_obj.network) tz_now = datetime.datetime.now(network_timezones.SG_TIMEZONE) show_length = datetime.timedelta(minutes=helpers.try_int(self._runtime or self._show_obj.runtime, 60)) future_airtime = not invalid_date and ( (self._airdate > (today + delta) or (not self._airdate < (today - delta) and ((show_time + show_length) > tz_now)))) # early conversion to int so that episode doesn't get marked dirty self.epid = getattr(ep_info, 'id', None) if None is self._epid: logger.log('Failed to retrieve ID from %s' % sickgear.TVInfoAPI(self.tvid).name, logger.ERROR) if helpers.should_delete_episode(self._status): self.delete_episode() elif UNKNOWN == self._status: self.status = (SKIPPED, UNAIRED)[future_airtime] return False # don't update show status if show dir is missing, unless it's missing on purpose # noinspection PyProtectedMember if not ek.ek(os.path.isdir, self._show_obj._location) \ and not sickgear.CREATE_MISSING_SHOW_DIRS and not sickgear.ADD_SHOWS_WO_DIR: if UNKNOWN == self._status: self.status = (SKIPPED, UNAIRED)[future_airtime] logger.log('The show directory is missing but an episode status found at Unknown is set Skipped') else: logger.log('The show directory is missing,' ' not bothering to change the episode statuses since it\'d probably be invalid') return if self._location: logger.log('%s: Setting status for %sx%s based on status %s and existence of %s' % (self._show_obj.tvid_prodid, season, episode, statusStrings[self._status], self._location), logger.DEBUG) # if we don't have the file if not ek.ek(os.path.isfile, self._location): if self._status in [SKIPPED, UNAIRED, UNKNOWN, WANTED]: very_old_delta = datetime.timedelta(days=90) very_old_airdate = datetime.date.fromordinal(1) < self._airdate < (today - very_old_delta) # if this episode hasn't aired yet set the status to UNAIRED if future_airtime: msg = 'Episode airs in the future, marking it %s' self.status = UNAIRED # if there's no airdate then set it to unaired (and respect ignored) elif self._airdate == datetime.date.fromordinal(1): if IGNORED == self._status: msg = 'Episode has no air date and marked %s, no change' else: msg = 'Episode has no air date, marking it %s' self.status = UNAIRED # if the airdate is in the past elif UNAIRED == self._status: msg = ('Episode status %s%s, with air date in the past, marking it ' % ( statusStrings[self._status], ','.join([(' is a special', '')[0 < self._season], ('', ' is paused')[self._show_obj.paused]])) + '%s') self.status = (SKIPPED, WANTED)[0 < self._season and not self._show_obj.paused and not very_old_airdate] # if still UNKNOWN or SKIPPED with the deprecated future airdate method elif UNKNOWN == self._status or (SKIPPED == self._status and old_airdate_future): msg = ('Episode status %s%s, with air date in the past, marking it ' % ( statusStrings[self._status], ','.join([ ('', ' has old future date format')[SKIPPED == self._status and old_airdate_future], ('', ' is being updated')[bool(update)], (' is a special', '')[0 < self._season]])) + '%s') self.status = (SKIPPED, WANTED)[update and not self._show_obj.paused and 0 < self._season and not very_old_airdate] else: msg = 'Not touching episode status %s, with air date in the past, because there is no file' else: msg = 'Not touching episode status %s, because there is no file' logger.log(msg % statusStrings[self._status], logger.DEBUG) # if we have a media file then it's downloaded elif sickgear.helpers.has_media_ext(self._location): if IGNORED == self._status: logger.log('File exists for %sx%s, ignoring because of status %s' % (self._season, self._episode, statusStrings[self._status]), logger.DEBUG) # leave propers alone, you have to either post-process them or manually change them back elif self._status not in Quality.SNATCHED_ANY + Quality.DOWNLOADED + Quality.ARCHIVED: msg = '(1) Status changes from %s to ' % statusStrings[self._status] self.status = Quality.statusFromNameOrFile(self._location, anime=self._show_obj.is_anime) logger.log('%s%s' % (msg, statusStrings[self._status]), logger.DEBUG) # shouldn't get here probably else: msg = '(2) Status changes from %s to ' % statusStrings[self._status] self.status = UNKNOWN logger.log('%s%s' % (msg, statusStrings[self._status]), logger.DEBUG) def load_from_nfo(self, location): """ :param location: :type location: AnyStr """ # noinspection PyProtectedMember if not ek.ek(os.path.isdir, self._show_obj._location): logger.log('%s: The show directory is missing, not bothering to try loading the episode NFO' % self._show_obj.tvid_prodid) return logger.log('%s: Loading episode details from the NFO file associated with %s' % (self.show_obj.tvid_prodid, location), logger.DEBUG) self.location = location if '' != self.location: if UNKNOWN == self._status and sickgear.helpers.has_media_ext(self.location): status_quality = Quality.statusFromNameOrFile(self.location, anime=self._show_obj.is_anime) logger.log('(3) Status changes from %s to %s' % (self._status, status_quality), logger.DEBUG) self.status = status_quality nfo_file = sickgear.helpers.replace_extension(self.location, 'nfo') logger.log('%s: Using NFO name %s' % (self._show_obj.tvid_prodid, nfo_file), logger.DEBUG) if ek.ek(os.path.isfile, nfo_file): try: show_xml = etree.ElementTree(file=nfo_file) except (SyntaxError, ValueError) as e: logger.log('Error loading the NFO, backing up the NFO and skipping for now: %s' % ex(e), logger.ERROR) # TODO: figure out what's wrong and fix it try: ek.ek(os.rename, nfo_file, '%s.old' % nfo_file) except (BaseException, Exception) as e: logger.log( 'Failed to rename your episode\'s NFO file - you need to delete it or fix it: %s' % ex(e), logger.ERROR) raise exceptions_helper.NoNFOException('Error in NFO format') # TODO: deprecated function getiterator needs to be replaced # for epDetails in showXML.getiterator('episodedetails'): for epDetails in list(show_xml.iter('episodedetails')): if None is epDetails.findtext('season') or int(epDetails.findtext('season')) != self._season or \ None is epDetails.findtext('episode') or int( epDetails.findtext('episode')) != self._episode: logger.log('%s: NFO has an block for a different episode - wanted %sx%s' ' but got %sx%s' % (self._show_obj.tvid_prodid, self._season, self._episode, epDetails.findtext('season'), epDetails.findtext('episode')), logger.DEBUG) continue if None is epDetails.findtext('title') or None is epDetails.findtext('aired'): raise exceptions_helper.NoNFOException('Error in NFO format (missing episode title or airdate)') self.name = epDetails.findtext('title') self.episode = int(epDetails.findtext('episode')) self.season = int(epDetails.findtext('season')) sickgear.scene_numbering.xem_refresh(self._show_obj.tvid, self._show_obj.prodid) self.scene_absolute_number = sickgear.scene_numbering.get_scene_absolute_numbering( self._show_obj.tvid, self._show_obj.prodid, absolute_number=self._absolute_number, season=self._season, episode=self._episode, show_obj=self._show_obj) self.scene_season, self.scene_episode = sickgear.scene_numbering.get_scene_numbering( self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode, show_obj=self._show_obj) self.description = epDetails.findtext('plot') if None is self._description: self.description = '' if epDetails.findtext('aired'): raw_airdate = [int(x) for x in epDetails.findtext('aired').split("-")] self.airdate = datetime.date(raw_airdate[0], raw_airdate[1], raw_airdate[2]) else: self.airdate = datetime.date.fromordinal(1) self.hasnfo = True else: self.hasnfo = False if ek.ek(os.path.isfile, sickgear.helpers.replace_extension(nfo_file, 'tbn')): self.hastbn = True else: self.hastbn = False def __ne__(self, o): # type: (TVEpisode) -> bool return not self.__eq__(o) def __eq__(self, o): # type: (TVEpisode) -> bool if not isinstance(o, TVEpisode): return False return o._show_obj == self._show_obj and o._epid == self._epid def __hash__(self): return hash((self._show_obj, self._epid)) def __repr__(self): return 'TVEpisode(%s)' % self.__str__() def __str__(self): return '%s - %sx%s - %s\n' % (self.show_obj.unique_name, self.season, self.episode, self.name) \ + 'location: %s\n' % self.location \ + 'description: %s\n' % self.description \ + 'subtitles: %s\n' % ','.join(self.subtitles) \ + 'subtitles_searchcount: %s\n' % self.subtitles_searchcount \ + 'subtitles_lastsearch: %s\n' % self.subtitles_lastsearch \ + 'airdate: %s (%s)\n' % (self.airdate.toordinal(), self.airdate) \ + 'hasnfo: %s\n' % self.hasnfo \ + 'hastbn: %s\n' % self.hastbn \ + 'status: %s\n' % self.status def create_meta_files(self, force=False): # noinspection PyProtectedMember if not ek.ek(os.path.isdir, self.show_obj._location): logger.log('%s: The show directory is missing, not bothering to try to create metadata' % self.show_obj.tvid_prodid) return self.create_nfo(force) self.create_thumbnail() if self.check_for_meta_files(): self.save_to_db() def create_nfo(self, force=False): """ :return: :rtype: bool """ result = False for cur_provider in itervalues(sickgear.metadata_provider_dict): result = cur_provider.create_episode_metadata(self, force) or result return result def create_thumbnail(self): """ :return: :rtype: bool """ result = False for cur_provider in itervalues(sickgear.metadata_provider_dict): result = cur_provider.create_episode_thumb(self) or result return result def delete_episode(self, return_sql=False): # type: (bool) -> Optional[List[List]] """ deletes episode from db, alternatively returns sql to remove episode from db :param return_sql: only return sql to delete episode """ logger.debug('Deleting %s %sx%s from the DB' % (self._show_obj.unique_name, self._season, self._episode)) # remove myself from the show dictionary if self.show_obj.get_episode(self._season, self._episode, no_create=True) == self: logger.log('Removing myself from my show\'s list', logger.DEBUG) del self.show_obj.sxe_ep_obj[self._season][self._episode] # delete myself from the DB logger.log('Deleting myself from the database', logger.DEBUG) sql = [['DELETE FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ?', [self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode]]] if return_sql: return sql my_db = db.DBConnection() my_db.mass_action(sql) raise exceptions_helper.EpisodeDeletedException() def get_sql(self, force_save=False): # type: (bool) -> Optional[List[AnyStr, List]] """ Creates SQL queue for this episode if any of its data has been changed since the last save. :param force_save: If True it will create SQL queue even if no data has been changed since the last save (aka if the record is not dirty). """ if not self.dirty and not force_save: logger.log('%s: Not creating SQL queue - record is not dirty' % self._show_obj.tvid_prodid, logger.DEBUG) return self.dirty = False return [ """ INSERT OR REPLACE INTO tv_episodes (episode_id, indexerid, indexer, name, description, subtitles, subtitles_searchcount, subtitles_lastsearch, airdate, hasnfo, hastbn, status, location, file_size, release_name, is_proper, showid, season, episode, absolute_number, version, release_group, network, network_id, network_country, network_country_code, network_is_stream, airtime, runtime, timestamp, timezone, scene_absolute_number, scene_season, scene_episode) VALUES ((SELECT episode_id FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ?) ,?,? ,?,? ,?,?,? ,?,?,?,?,?,? ,?,? ,?,?,?,? ,?,? ,?,?,?,?,? ,?,?,?,?, (SELECT scene_absolute_number FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ?), (SELECT scene_season FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ?), (SELECT scene_episode FROM tv_episodes WHERE indexer = ? AND showid = ? AND season = ? AND episode = ?)); """, [self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode, self._epid, self._tvid, self._name, self._description, ','.join([sub for sub in self._subtitles]), self._subtitles_searchcount, self._subtitles_lastsearch, self._airdate.toordinal(), self._hasnfo, self._hastbn, self._status, self._location, self._file_size, self._release_name, self._is_proper, self._show_obj.prodid, self._season, self._episode, self._absolute_number, self._version, self._release_group, self._network, self._network_id, self._network_country, self._network_country_code, self._network_is_stream, time_to_int(self._airtime), self._runtime, self._timestamp, self._timezone, self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode, self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode, self._show_obj.tvid, self._show_obj.prodid, self._season, self._episode]] def save_to_db(self, force_save=False): """ Saves this episode to the database if any of its data has been changed since the last save. :param force_save: If True it will save to the database even if no data has been changed since the last save (aka if the record is not dirty). """ if not self.dirty and not force_save: logger.log('%s: Not saving episode to db - record is not dirty' % self._show_obj.tvid_prodid, logger.DEBUG) return logger.log('%s: Saving episode details to database' % self._show_obj.tvid_prodid, logger.DEBUG) logger.log('STATUS IS %s' % statusStrings[self._status], logger.DEBUG) new_value_dict = dict( absolute_number=self._absolute_number, airdate=self._airdate.toordinal(), airtime=time_to_int(self._airtime), description=self._description, file_size=self._file_size, hasnfo=self._hasnfo, hastbn=self._hastbn, indexer=self._tvid, indexerid=self._epid, is_proper=self._is_proper, location=self._location, name=self._name, network=self._network, network_country=self._network_country, network_country_code=self._network_country_code, network_id=self._network_id, network_is_stream=self._network_is_stream, release_group=self._release_group, release_name=self._release_name, runtime=self._runtime, status=self._status, subtitles=','.join([sub for sub in self._subtitles]), subtitles_lastsearch=self._subtitles_lastsearch, subtitles_searchcount=self._subtitles_searchcount, timestamp=self._timestamp, timezone=self._timezone, version=self._version, ) control_value_dict = dict( indexer=self.show_obj.tvid, showid=self.show_obj.prodid, season=self.season, episode=self.episode) # use a custom update/insert method to get the data into the DB my_db = db.DBConnection() my_db.upsert('tv_episodes', new_value_dict, control_value_dict) self.dirty = False # # TODO: remove if unused # def full_location(self): # if self.location in (None, ''): # return None # return ek.ek(os.path.join, self.show_obj.location, self.location) # # # TODO: remove if unused # def create_strings(self, pattern=None): # patterns = [ # '%S.N.S%SE%0E', # '%S.N.S%0SE%E', # '%S.N.S%SE%E', # '%S.N.S%0SE%0E', # '%SN S%SE%0E', # '%SN S%0SE%E', # '%SN S%SE%E', # '%SN S%0SE%0E' # ] # # strings = [] # if not pattern: # for p in patterns: # strings += [self._format_pattern(p)] # return strings # return self._format_pattern(pattern) def pretty_name(self): """ Returns the name of this episode in a "pretty" human-readable format. Used for logging and notifications and such. :return: A string representing the episode's name and season/ep numbers :rtype: AnyStr """ if self._show_obj.anime and not self._show_obj.scene: return self._format_pattern('%SN - %AB - %EN') if self._show_obj.air_by_date: return self._format_pattern('%SN - %AD - %EN') return self._format_pattern('%SN - %Sx%0E - %EN') def _ep_name(self): """ :return: the name of the episode to use during renaming. Combines the names of related episodes. Eg. "Ep Name (1)" and "Ep Name (2)" becomes "Ep Name" "Ep Name" and "Other Ep Name" becomes "Ep Name & Other Ep Name" :rtype: AnyStr """ multi_name_regex = r'(.*) \(\d{1,2}\)' self.related_ep_obj = sorted(self.related_ep_obj, key=lambda se: se.episode) if 0 == len(self.related_ep_obj): good_name = self._name else: single_name = True known_good_name = None for cur_name in [self._name] + [x.name for x in self.related_ep_obj]: match = re.match(multi_name_regex, cur_name) if not match: single_name = False break if None is known_good_name: known_good_name = match.group(1) elif known_good_name != match.group(1): single_name = False break if single_name: good_name = known_good_name else: good_name = self._name for ep_obj in self.related_ep_obj: good_name += ' & ' + ep_obj.name return good_name or 'tba' def _replace_map(self): """ Generates a replacement map for this episode which maps all possible custom naming patterns to the correct value for this episode. Returns: A dict with patterns as the keys and their replacement values as the values. """ ep_name = self._ep_name() def dot(name): return helpers.sanitize_scene_name(name) def us(name): return re.sub('[ -]', '_', name) def release_name(name, is_anime=False): if name: name = helpers.remove_non_release_groups(name, is_anime) return name def release_group(show_obj, name): if name: name = helpers.remove_non_release_groups(name, show_obj.is_anime) else: return '' try: np = NameParser(name, show_obj=show_obj, naming_pattern=True) parse_result = np.parse(name) except (InvalidNameException, InvalidShowException) as e: logger.log('Unable to get parse release_group: %s' % ex(e), logger.DEBUG) return '' if not parse_result.release_group: return '' return parse_result.release_group ep_status, ep_qual = Quality.splitCompositeStatus(self._status) if sickgear.NAMING_STRIP_YEAR: show_name = re.sub(r'\(\d+\)$', '', self._show_obj.name).rstrip() else: show_name = self._show_obj.name return { '%SN': show_name, '%S.N': dot(show_name), '%S_N': us(show_name), '%EN': ep_name, '%E.N': dot(ep_name), '%E_N': us(ep_name), '%QN': Quality.qualityStrings[ep_qual], '%Q.N': dot(Quality.qualityStrings[ep_qual]), '%Q_N': us(Quality.qualityStrings[ep_qual]), '%S': str(self._season), '%0S': '%02d' % self._season, '%E': str(self._episode), '%0E': '%02d' % self._episode, '%XS': str(self.scene_season), '%0XS': '%02d' % self.scene_season, '%XE': str(self.scene_episode), '%0XE': '%02d' % self.scene_episode, '%AB': '%(#)03d' % {'#': self._absolute_number}, '%XAB': '%(#)03d' % {'#': self.scene_absolute_number}, '%RN': release_name(self._release_name, self._show_obj.is_anime), '%RG': release_group(self._show_obj, self._release_name), '%AD': str(self._airdate).replace('-', ' '), '%A.D': str(self._airdate).replace('-', '.'), '%A_D': us(str(self._airdate)), '%A-D': str(self._airdate), '%Y': str(self._airdate.year), '%M': str(self._airdate.month), '%D': str(self._airdate.day), '%0M': '%02d' % self._airdate.month, '%0D': '%02d' % self._airdate.day, '%RT': "PROPER" if self.is_proper else "", '%V': 'v%s' % self._version if self._show_obj.is_anime and 1 < self._version else '', } @staticmethod def _format_string(pattern, replace_map): """ Replaces all template strings with the correct value """ result_name = pattern # do the replacements for cur_replacement in sorted(list_keys(replace_map), reverse=True): result_name = result_name.replace(cur_replacement, helpers.sanitize_filename(replace_map[cur_replacement])) result_name = result_name.replace(cur_replacement.lower(), helpers.sanitize_filename(replace_map[cur_replacement].lower())) return result_name def _format_pattern(self, pattern=None, multi=None, anime_type=None): """ Manipulates an episode naming pattern and then fills the template in """ if None is pattern: pattern = sickgear.NAMING_PATTERN if None is multi: multi = sickgear.NAMING_MULTI_EP if None is anime_type: anime_type = sickgear.NAMING_ANIME replace_map = self._replace_map() result_name = pattern # if there's no release group then replace it with a reasonable facsimile if not replace_map['%RN']: if self._show_obj.air_by_date or self._show_obj.sports: result_name = result_name.replace('%RN', '%S.N.%A.D.%E.N-SickGear') result_name = result_name.replace('%rn', '%s.n.%A.D.%e.n-SickGear') elif 3 != anime_type: result_name = result_name.replace('%RN', '%S.N.%AB.%E.N-SickGear') result_name = result_name.replace('%rn', '%s.n.%ab.%e.n-SickGear') else: result_name = result_name.replace('%RN', '%S.N.S%0SE%0E.%E.N-SickGear') result_name = result_name.replace('%rn', '%s.n.s%0se%0e.%e.n-SickGear') result_name = result_name.replace('%RG', 'SickGear') result_name = result_name.replace('%rg', 'SickGear') logger.log('Episode has no release name, replacing it with a generic one: %s' % result_name, logger.DEBUG) if not replace_map['%RT']: result_name = re.sub('([ _.-]*)%RT([ _.-]*)', r'\2', result_name) # split off ep name part only name_groups = re.split(r'[\\/]', result_name) # figure out the double-ep numbering style for each group, if applicable for cur_name_group in name_groups: season_ep_regex = r''' (?P[ _.-]*) ((?:s(?:eason|eries)?\s*)?%0?S(?![._]?N)) (.*?) (%0?E(?![._]?N)) (?P[ _.-]*) ''' ep_only_regex = '(E?%0?E(?![._]?N))' # try the normal way season_ep_match = re.search(season_ep_regex, cur_name_group, re.I | re.X) ep_only_match = re.search(ep_only_regex, cur_name_group, re.I | re.X) # if we have a season and episode then collect the necessary data if season_ep_match: season_format = season_ep_match.group(2) ep_sep = season_ep_match.group(3) ep_format = season_ep_match.group(4) sep = season_ep_match.group('pre_sep') if not sep: sep = season_ep_match.group('post_sep') if not sep: sep = ' ' # force 2-3-4 format if they chose to extend if multi in (NAMING_EXTEND, NAMING_LIMITED_EXTEND, NAMING_LIMITED_EXTEND_E_PREFIXED): ep_sep = '-' regex_used = season_ep_regex # if there's no season then there's not much choice so we'll just force them to use 03-04-05 style elif ep_only_match: season_format = '' ep_sep = '-' ep_format = ep_only_match.group(1) sep = '' regex_used = ep_only_regex else: continue # we need at least this much info to continue if not ep_sep or not ep_format: continue # start with the ep string, eg. E03 ep_string = self._format_string(ep_format.upper(), replace_map) for cur_ep_obj in self.related_ep_obj: # for limited extend we only append the last ep if multi in (NAMING_LIMITED_EXTEND, NAMING_LIMITED_EXTEND_E_PREFIXED) \ and cur_ep_obj != self.related_ep_obj[-1]: continue elif multi == NAMING_DUPLICATE: # add " - S01" ep_string += sep + season_format elif multi == NAMING_SEPARATED_REPEAT: ep_string += sep # add "E04" ep_string += ep_sep if multi == NAMING_LIMITED_EXTEND_E_PREFIXED: ep_string += 'E' # noinspection PyProtectedMember ep_string += cur_ep_obj._format_string(ep_format.upper(), cur_ep_obj._replace_map()) if 3 != anime_type: absolute_number = (self._absolute_number, self._episode)[0 == self._absolute_number] if 0 != self._season: # dont set absolute numbers if we are on specials ! if 1 == anime_type: # this crazy person wants both ! (note: +=) ep_string += sep + '%(#)03d' % {'#': absolute_number} elif 2 == anime_type: # total anime freak only need the absolute number ! (note: =) ep_string = '%(#)03d' % {'#': absolute_number} for cur_ep_obj in self.related_ep_obj: if 0 != cur_ep_obj.absolute_number: ep_string += '-' + '%(#)03d' % {'#': cur_ep_obj.absolute_number} else: ep_string += '-' + '%(#)03d' % {'#': cur_ep_obj.episode} regex_replacement = None if 2 == anime_type: regex_replacement = r'\g' + ep_string + r'\g' elif season_ep_match: regex_replacement = r'\g\g<2>\g<3>' + ep_string + r'\g' elif ep_only_match: regex_replacement = ep_string if regex_replacement: # fill out the template for this piece and then insert this piece into the actual pattern cur_name_group_result = re.sub('(?i)(?x)' + regex_used, regex_replacement, cur_name_group) # cur_name_group_result = cur_name_group.replace(ep_format, ep_string) # logger.log(u"found "+ep_format+" as the ep pattern using "+regex_used+" # and replaced it with "+regex_replacement+" to result in "+cur_name_group_result+" # from "+cur_name_group, logger.DEBUG) result_name = result_name.replace(cur_name_group, cur_name_group_result) result_name = self._format_string(result_name, replace_map) logger.log('formatting pattern: %s -> %s' % (pattern, result_name), logger.DEBUG) return result_name def proper_path(self): """ Figures out the path where this episode SHOULD live according to the renaming rules, relative from the show dir :return: :rtype: AnyStr """ anime_type = sickgear.NAMING_ANIME if not self._show_obj.is_anime: anime_type = 3 result = self.formatted_filename(anime_type=anime_type) # if they want us to flatten it and we're allowed to flatten it then we will if self._show_obj.flatten_folders and not sickgear.NAMING_FORCE_FOLDERS: return result # if not we append the folder on and use that return ek.ek(os.path.join, self.formatted_dir(), result) def formatted_dir(self, pattern=None, multi=None): """ Just the folder name of the episode """ if None is pattern: # we only use ABD if it's enabled, this is an ABD show, AND this is not a multi-ep if self._show_obj.air_by_date and sickgear.NAMING_CUSTOM_ABD and not self.related_ep_obj: pattern = sickgear.NAMING_ABD_PATTERN elif self._show_obj.sports and sickgear.NAMING_CUSTOM_SPORTS and not self.related_ep_obj: pattern = sickgear.NAMING_SPORTS_PATTERN elif self._show_obj.anime and sickgear.NAMING_CUSTOM_ANIME: pattern = sickgear.NAMING_ANIME_PATTERN else: pattern = sickgear.NAMING_PATTERN # split off the dirs only, if they exist name_groups = re.split(r'[\\/]', pattern) if 1 == len(name_groups): logger.debug('No Season Folder set in Naming pattern: %s' % pattern) return '' return self._format_pattern(ek.ek(os.sep.join, name_groups[:-1]), multi) def formatted_filename(self, pattern=None, multi=None, anime_type=None): """ Just the filename of the episode, formatted based on the naming settings """ if None is pattern: # we only use ABD if it's enabled, this is an ABD show, AND this is not a multi-ep if self._show_obj.air_by_date and sickgear.NAMING_CUSTOM_ABD and not self.related_ep_obj: pattern = sickgear.NAMING_ABD_PATTERN elif self._show_obj.sports and sickgear.NAMING_CUSTOM_SPORTS and not self.related_ep_obj: pattern = sickgear.NAMING_SPORTS_PATTERN elif self._show_obj.anime and sickgear.NAMING_CUSTOM_ANIME: pattern = sickgear.NAMING_ANIME_PATTERN else: pattern = sickgear.NAMING_PATTERN # split off the dirs only, if they exist name_groups = re.split(r'[\\/]', pattern) return self._format_pattern(name_groups[-1], multi, anime_type) def rename(self): """ Renames an episode file and all related files to the location and filename as specified in the naming settings. """ if not ek.ek(os.path.isfile, self.location): logger.log('Can\'t perform rename on %s when it doesn\'t exist, skipping' % self.location, logger.WARNING) return proper_path = self.proper_path() absolute_proper_path = ek.ek(os.path.join, self._show_obj.location, proper_path) absolute_current_path_no_ext, file_ext = ek.ek(os.path.splitext, self.location) absolute_current_path_no_ext_length = len(absolute_current_path_no_ext) related_subs = [] current_path = absolute_current_path_no_ext if absolute_current_path_no_ext.startswith(self._show_obj.location): current_path = absolute_current_path_no_ext[len(self._show_obj.location):] logger.log('Renaming/moving episode from the base path %s to %s' % (self.location, absolute_proper_path), logger.DEBUG) # if it's already named correctly then don't do anything if proper_path == current_path: logger.log('%s: File %s is already named correctly, skipping' % (self._epid, self.location), logger.DEBUG) return related_files = postProcessor.PostProcessor(self.location).list_associated_files( self.location, base_name_only=True) if self.show_obj.subtitles and '' != sickgear.SUBTITLES_DIR: related_subs = postProcessor.PostProcessor(self.location).list_associated_files(sickgear.SUBTITLES_DIR, subtitles_only=True) # absolute_proper_subs_path = ek.ek(os.path.join, sickgear.SUBTITLES_DIR, self.formatted_filename()) logger.log('Files associated to %s: %s' % (self.location, related_files), logger.DEBUG) # move the ep file result = helpers.rename_ep_file(self.location, absolute_proper_path, absolute_current_path_no_ext_length) # move related files for cur_related_file in related_files: renamed = helpers.rename_ep_file(cur_related_file, absolute_proper_path, absolute_current_path_no_ext_length) if not renamed: logger.log('%s: Unable to rename file %s' % (self._epid, cur_related_file), logger.ERROR) for cur_related_sub in related_subs: absolute_proper_subs_path = ek.ek(os.path.join, sickgear.SUBTITLES_DIR, self.formatted_filename()) renamed = helpers.rename_ep_file(cur_related_sub, absolute_proper_subs_path, absolute_current_path_no_ext_length) if not renamed: logger.log('%s: Unable to rename file %s' % (self._epid, cur_related_sub), logger.ERROR) # save the ep with self.lock: if result: self.location = absolute_proper_path + file_ext for ep_obj in self.related_ep_obj: ep_obj.location = absolute_proper_path + file_ext # in case something changed with the metadata just do a quick check for cur_ep_obj in [self] + self.related_ep_obj: cur_ep_obj.check_for_meta_files() # save any changes to the database sql_l = [] with self.lock: for ep_obj in [self] + self.related_ep_obj: result = ep_obj.get_sql() if None is not result: sql_l.append(result) if 0 < len(sql_l): my_db = db.DBConnection() my_db.mass_action(sql_l) def airdate_modify_stamp(self): """ Make the modify date and time of a file reflect the show air date and time. Note: Also called from postProcessor """ has_timestamp = isinstance(self._timestamp, int) and 1 < self._timestamp if not has_timestamp and (not isinstance(self._airdate, datetime.date) or 1 == self._airdate.year): logger.log('%s: Did not change modify date of %s because episode date is never aired or invalid' % (self._show_obj.tvid_prodid, ek.ek(os.path.basename, self.location)), logger.DEBUG) return aired_dt = None if not has_timestamp: if isinstance(self._airtime, datetime.time): airtime = self._airtime else: hr, m = network_timezones.parse_time(self._show_obj.airs) airtime = datetime.time(hr, m) aired_dt = SGDatetime.combine(self.airdate, airtime) try: aired_epoch = SGDatetime.to_file_timestamp(aired_dt) filemtime = int(ek.ek(os.path.getmtime, self.location)) except (BaseException, Exception): return if filemtime != aired_epoch: result, loglevel = 'Changed', logger.MESSAGE if not helpers.touch_file(self.location, aired_epoch): result, loglevel = 'Error changing', logger.WARNING logger.log('%s: %s modify date of %s to show air date %s' % (self._show_obj.tvid_prodid, result, ek.ek(os.path.basename, self.location), 'n/a' if not aired_dt else aired_dt.strftime('%b %d,%Y (%H:%M)')), loglevel) def __getstate__(self): d = dict(self.__dict__) del d['lock'] return d def __setstate__(self, d): d['lock'] = threading.RLock() self.__dict__.update(d) def __bool__(self): return bool(self._tvid) and bool(self._epid) __nonzero__ = __bool__