# coding=utf-8 # Author: Nic Wolfe # URL: http://code.google.com/p/sickbeard/ # # This file is part of Sick Beard. # # Sick Beard 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. # # Sick Beard 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 Sick Beard. If not, see . from __future__ import with_statement import datetime import os import re import urllib import urlparse import sickbeard from lib import requests from lib.feedparser import feedparser from sickbeard import helpers, classes, logger, db from sickbeard.common import MULTI_EP_RESULT, SEASON_RESULT #, SEED_POLICY_TIME, SEED_POLICY_RATIO from sickbeard import tvcache from sickbeard import encodingKludge as ek from sickbeard.exceptions import ex from lib.hachoir_parser import createParser from sickbeard.name_parser.parser import NameParser, InvalidNameException from sickbeard.common import Quality class GenericProvider: NZB = "nzb" TORRENT = "torrent" def __init__(self, name): # these need to be set in the subclass self.providerType = None self.name = name self.url = '' self.show = None self.supportsBacklog = False self.cache = tvcache.TVCache(self) self.session = requests.session() self.session.verify = False self.session.headers.update({ 'user-agent': 'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/32.0.1700.107 Safari/537.36'}) def getID(self): return GenericProvider.makeID(self.name) @staticmethod def makeID(name): return re.sub("[^\w\d_]", "_", name.strip().lower()) def imageName(self): return self.getID() + '.png' def _checkAuth(self): return def isActive(self): if self.providerType == GenericProvider.NZB and sickbeard.USE_NZBS: return self.isEnabled() elif self.providerType == GenericProvider.TORRENT and sickbeard.USE_TORRENTS: return self.isEnabled() else: return False def isEnabled(self): """ This should be overridden and should return the config setting eg. sickbeard.MYPROVIDER """ return False def getResult(self, episodes): """ Returns a result of the correct type for this provider """ if self.providerType == GenericProvider.NZB: result = classes.NZBSearchResult(episodes) elif self.providerType == GenericProvider.TORRENT: result = classes.TorrentSearchResult(episodes) else: result = classes.SearchResult(episodes) result.provider = self return result def getURL(self, url, post_data=None, headers=None, json=False): """ By default this is just a simple urlopen call but this method should be overridden for providers with special URL requirements (like cookies) """ if not headers: headers = [] data = helpers.getURL(url, post_data, headers, json=json) if not data: logger.log(u"Error loading " + self.name + " URL: " + url, logger.ERROR) return None return data def downloadResult(self, result): """ Save the result to disk. """ logger.log(u"Downloading a result from " + self.name + " at " + result.url) data = self.getURL(result.url) if data is None: return False # use the appropriate watch folder if self.providerType == GenericProvider.NZB: saveDir = sickbeard.NZB_DIR writeMode = 'w' elif self.providerType == GenericProvider.TORRENT: saveDir = sickbeard.TORRENT_DIR writeMode = 'wb' else: return False # use the result name as the filename file_name = ek.ek(os.path.join, saveDir, helpers.sanitizeFileName(result.name) + '.' + self.providerType) logger.log(u"Saving to " + file_name, logger.DEBUG) try: with open(file_name, writeMode) as fileOut: fileOut.write(data) helpers.chmodAsParent(file_name) except EnvironmentError, e: logger.log("Unable to save the file: " + ex(e), logger.ERROR) return False # as long as it's a valid download then consider it a successful snatch return self._verify_download(file_name) def _verify_download(self, file_name=None): """ Checks the saved file to see if it was actually valid, if not then consider the download a failure. """ # primitive verification of torrents, just make sure we didn't get a text file or something if self.providerType == GenericProvider.TORRENT: parser = createParser(file_name) if parser: mime_type = parser._getMimeType() try: parser.stream._input.close() except: pass if mime_type != 'application/x-bittorrent': logger.log(u"Result is not a valid torrent file", logger.WARNING) return False return True def searchRSS(self): self._checkAuth() self.cache.updateCache() return self.cache.findNeededEpisodes() def getQuality(self, item): """ Figures out the quality of the given RSS item node item: An elementtree.ElementTree element representing the tag of the RSS feed Returns a Quality value obtained from the node's data """ (title, url) = self._get_title_and_url(item) # @UnusedVariable quality = Quality.sceneQuality(title) return quality def _doSearch(self, search_params, show=None, age=None): return [] def _get_season_search_strings(self, episode): return [] def _get_episode_search_strings(self, eb_obj, add_string=''): return [] def _get_title_and_url(self, item): """ Retrieves the title and URL data from the item XML node item: An elementtree.ElementTree element representing the tag of the RSS feed Returns: A tuple containing two strings representing title and URL respectively """ title = item.title if title: title = title.replace(' ', '.') url = item.link if url: url = url.replace('&', '&') return (title, url) def findSearchResults(self, show, season, episodes, seasonSearch=False, manualSearch=False): self._checkAuth() self.show = show results = {} searchStrings = [] itemList = [] for epObj in episodes: cacheResult = self.cache.searchCache(epObj, manualSearch) if len(cacheResult): return cacheResult if epObj.show.air_by_date: logger.log(u'Searching "%s" for "%s"' % (self.name, epObj.prettyABDName())) else: logger.log(u'Searching "%s" for "%s" as "%s"' % (self.name, epObj.prettyName(), epObj.prettySceneName())) # get our search strings if seasonSearch: searchStrings += self._get_season_search_strings(epObj) searchStrings += self._get_episode_search_strings(epObj) # remove duplicate search strings if len(searchStrings): searchStrings = [i for n, i in enumerate(searchStrings) if i not in searchStrings[n + 1:]] for curString in sorted(searchStrings): itemList += self._doSearch(curString) # remove duplicate items if len(itemLisT): itemList = list(set(itemList)) for item in itemList: (title, url) = self._get_title_and_url(item) quality = self.getQuality(item) # parse the file name try: myParser = NameParser(False) parse_result = myParser.parse(title).convert() except InvalidNameException: logger.log(u"Unable to parse the filename " + title + " into a valid episode", logger.WARNING) continue if not (self.show.air_by_date or self.show.sports): if not parse_result.episode_numbers and (parse_result.season_number != None and parse_result.season_number != season) or ( parse_result.season_number == None and season != 1): logger.log(u"The result " + title + " doesn't seem to be a valid season for season " + str( season) + ", ignoring", logger.DEBUG) continue elif len(parse_result.episode_numbers) and (parse_result.season_number != season or parse_result.episode_numbers[0] not in parse_result.episode_numbers): logger.log(u"Episode " + title + " isn't " + str(season) + "x" + str( parse_result.episode_numbers[0]) + ", skipping it", logger.DEBUG) continue # we just use the existing info for normal searches actual_season = season actual_episodes = parse_result.episode_numbers else: if not (parse_result.air_by_date or parse_result.sports): logger.log( u"This is supposed to be a date search but the result " + title + " didn't parse as one, skipping it", logger.DEBUG) continue myDB = db.DBConnection() sql_results = myDB.select("SELECT season, episode FROM tv_episodes WHERE showid = ? AND airdate = ?", [show.indexerid, parse_result.air_date.toordinal() or parse_result.sports_event_date.toordinal()]) if len(sql_results) != 1: logger.log( u"Tried to look up the date for the episode " + title + " but the database didn't give proper results, skipping it", logger.WARNING) continue actual_season = int(sql_results[0]["season"]) actual_episodes = [int(sql_results[0]["episode"])] # make sure we want the episode wantEp = True for epNo in actual_episodes: if not show.wantEpisode(actual_season, epNo, quality): wantEp = False break if not wantEp: logger.log(u"Ignoring result " + title + " because we don't want an episode that is " + Quality.qualityStrings[quality], logger.DEBUG) continue logger.log(u"Found result " + title + " at " + url, logger.DEBUG) # make a result object epObj = [] for curEp in actual_episodes: epObj.append(show.getEpisode(actual_season, curEp)) result = self.getResult(epObj) result.url = url result.name = title result.quality = quality result.provider = self result.content = None if len(epObj) == 1: epNum = epObj[0].episode logger.log(u"Single episode result.", logger.DEBUG) elif len(epObj) > 1: epNum = MULTI_EP_RESULT logger.log(u"Separating multi-episode result to check for later - result contains episodes: " + str( parse_result.episode_numbers), logger.DEBUG) elif len(epObj) == 0: epNum = SEASON_RESULT result.extraInfo = [show] logger.log(u"Separating full season result to check for later", logger.DEBUG) if epNum in results: results[epNum].append(result) else: results[epNum] = [result] # remove duplicate results results[epNum] = list(set(results[epNum])) return results def findPropers(self, search_date=None): results = self.cache.listPropers(search_date) return [classes.Proper(x['name'], x['url'], datetime.datetime.fromtimestamp(x['time'])) for x in results] class NZBProvider(GenericProvider): def __init__(self, name): GenericProvider.__init__(self, name) self.providerType = GenericProvider.NZB class TorrentProvider(GenericProvider): def __init__(self, name): GenericProvider.__init__(self, name) self.providerType = GenericProvider.TORRENT