From 0aadc7a73e6c780519e03257dbe92152d8fd7de3 Mon Sep 17 00:00:00 2001 From: JackDandy Date: Mon, 22 Jun 2015 01:02:07 +0100 Subject: [PATCH] Change refactor BTS to use torrent provider simplification and PEP8. Change BTS login process to use General Config/Advanced/Proxy host setting. --- CHANGES.md | 4 +- sickbeard/providers/bitsoup.py | 260 +++++++++------------------------ 2 files changed, 74 insertions(+), 190 deletions(-) diff --git a/CHANGES.md b/CHANGES.md index c773c87e..8d120c61 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -22,11 +22,12 @@ * Change provider IPT login process to use General Config/Advanced/Proxy host setting * Change provider TB PEP8 and code convention cleanup * Change provider TB login process to use General Config/Advanced/Proxy host setting -* Change provider TD login process to use General Config/Advanced/Proxy host setting * Remove useless webproxies from provider TPB as they fail for one reason or another * Change provider TPB to use mediaExtensions from common instead of hard-coded private list * Add new tld variants to provider TPB * Add test for authenticity to provider TPB to notify of 3rd party block +* Change provider TD login process to use General Config/Advanced/Proxy host setting +* Change provider BTS login process to use General Config/Advanced/Proxy host setting * Change provider Womble's use SSL * Change provider KAT remove dead url * Change provider KAT to use mediaExtensions from common instead of private list @@ -38,6 +39,7 @@ * Change refactor TBP to use torrent provider simplification and PEP8 * Change refactor TD to use torrent provider simplification and PEP8 * Change refactor TL to use torrent provider simplification and PEP8 +* Change refactor BTS to use torrent provider simplification and PEP8 * Remove HDTorrents torrent provider * Remove NextGen torrent provider * Add Rarbg torrent provider diff --git a/sickbeard/providers/bitsoup.py b/sickbeard/providers/bitsoup.py index 1e3971a5..6cd8d1f1 100644 --- a/sickbeard/providers/bitsoup.py +++ b/sickbeard/providers/bitsoup.py @@ -1,5 +1,4 @@ -# Author: Idan Gutman -# URL: http://code.google.com/p/sickbeard/ +# coding=utf-8 # # This file is part of SickGear. # @@ -17,247 +16,130 @@ # along with SickGear. If not, see . import re -import traceback import datetime +import traceback -import sickbeard -import generic -import requests -import requests.exceptions -from sickbeard.common import Quality -from sickbeard import logger, tvcache, db, classes, helpers, show_name_helpers -from sickbeard.exceptions import ex, AuthException -from sickbeard.helpers import sanitizeSceneName +from . import generic +from sickbeard import logger, tvcache, helpers from sickbeard.bs4_parser import BS4Parser -from unidecode import unidecode +from lib.unidecode import unidecode class BitSoupProvider(generic.TorrentProvider): - urls = {'base_url': 'https://www.bitsoup.me', - 'login': 'https://www.bitsoup.me/takelogin.php', - 'detail': 'https://www.bitsoup.me/details.php?id=%s', - 'search': 'https://www.bitsoup.me/browse.php?search=%s%s', - 'download': 'https://bitsoup.me/%s', - } def __init__(self): - generic.TorrentProvider.__init__(self, 'BitSoup', True, False) - self.username = None - self.password = None - self.ratio = None - self.minseed = None - self.minleech = None + generic.TorrentProvider.__init__(self, 'BitSoup') + + self.url_base = 'https://www.bitsoup.me/' + self.urls = {'config_provider_home_uri': self.url_base, + 'login': self.url_base + 'takelogin.php', + 'search': self.url_base + 'browse.php?search=%s%s', + 'get': self.url_base + '%s'} + + self.categories = '&c42=1&c45=1&c49=1&c7=1&incldead=0&blah=0' + + self.url = self.urls['config_provider_home_uri'] + + self.username, self.password, self.minseed, self.minleech = 4 * [None] self.cache = BitSoupCache(self) - self.url = self.urls['base_url'] - self.categories = '&c42=1&c45=1&c49=1&c7=1' - - def getQuality(self, item, anime=False): - - quality = Quality.sceneQuality(item[0], anime) - return quality - - def _checkAuth(self): - if not self.username or not self.password: - raise AuthException("Your authentication credentials for " + self.name + " are missing, check your config.") - - return True def _doLogin(self): - login_params = {'username': self.username, - 'password': self.password, - 'ssl': 'yes' - } + logged_in = lambda: 'uid' in self.session.cookies and 'pass' in self.session.cookies + if logged_in(): + return True - if not self.session: - self.session = requests.session() + if self._checkAuth(): + login_params = {'username': self.username, 'password': self.password, 'ssl': 'yes'} + response = helpers.getURL(self.urls['login'], post_data=login_params, session=self.session) + if response and logged_in(): + return True - try: - response = self.session.post(self.urls['login'], data=login_params, timeout=30, verify=False) - except (requests.exceptions.ConnectionError, requests.exceptions.HTTPError) as e: - logger.log(u'Unable to connect to ' + self.name + ' provider: ' + ex(e), logger.ERROR) - return False + msg = u'Failed to authenticate with %s, abort provider' + if response and 'Username or password incorrect' in response: + msg = u'Invalid username or password for %s. Check settings' + logger.log(msg % self.name, logger.ERROR) - if re.search('Username or password incorrect', response.text): - logger.log(u'Your authentication credentials for ' + self.name + ' are incorrect, check your config.', logger.ERROR) - return False - - return True - - def _get_season_search_strings(self, ep_obj): - - search_string = {'Season': []} - for show_name in set(show_name_helpers.allPossibleShowNames(self.show)): - if ep_obj.show.air_by_date or ep_obj.show.sports: - ep_string = show_name + ' ' + str(ep_obj.airdate).split('-')[0] - elif ep_obj.show.anime: - ep_string = show_name + ' ' + "%d" % ep_obj.scene_absolute_number - else: - ep_string = show_name + ' S%02d' % int(ep_obj.scene_season) #1) showName SXX - - search_string['Season'].append(ep_string) - - return [search_string] - - def _get_episode_search_strings(self, ep_obj, add_string=''): - - search_string = {'Episode': []} - - if not ep_obj: - return [] - - if self.show.air_by_date: - for show_name in set(show_name_helpers.allPossibleShowNames(self.show)): - ep_string = sanitizeSceneName(show_name) + ' ' + \ - str(ep_obj.airdate).replace('-', '|') - search_string['Episode'].append(ep_string) - elif self.show.sports: - for show_name in set(show_name_helpers.allPossibleShowNames(self.show)): - ep_string = sanitizeSceneName(show_name) + ' ' + \ - str(ep_obj.airdate).replace('-', '|') + '|' + \ - ep_obj.airdate.strftime('%b') - search_string['Episode'].append(ep_string) - elif self.show.anime: - for show_name in set(show_name_helpers.allPossibleShowNames(self.show)): - ep_string = sanitizeSceneName(show_name) + ' ' + \ - "%i" % int(ep_obj.scene_absolute_number) - search_string['Episode'].append(ep_string) - else: - for show_name in set(show_name_helpers.allPossibleShowNames(self.show)): - ep_string = show_name_helpers.sanitizeSceneName(show_name) + ' ' + \ - sickbeard.config.naming_ep_type[2] % {'seasonnumber': ep_obj.scene_season, - 'episodenumber': ep_obj.scene_episode} - - search_string['Episode'].append(re.sub('\s+', ' ', ep_string)) - - return [search_string] + return False def _doSearch(self, search_params, search_mode='eponly', epcount=0, age=0): results = [] - items = {'Season': [], 'Episode': [], 'RSS': []} - if not self._doLogin(): - return [] + return results + items = {'Season': [], 'Episode': [], 'Cache': []} + + rc = dict((k, re.compile('(?i)' + v)) for (k, v) in {'info': 'detail', 'get': 'download'}.items()) for mode in search_params.keys(): for search_string in search_params[mode]: if isinstance(search_string, unicode): search_string = unidecode(search_string) - searchURL = self.urls['search'] % (search_string, self.categories) - - logger.log(u"Search string: " + searchURL, logger.DEBUG) - - data = self.getURL(searchURL) - if not data: - continue + search_url = self.urls['search'] % (search_string, self.categories) + html = self.getURL(search_url) + cnt = len(items[mode]) try: - with BS4Parser(data, "html.parser") as html: - torrent_table = html.find('table', attrs={'class': 'koptekst'}) - torrent_rows = torrent_table.find_all('tr') if torrent_table else [] + if not html or self._has_no_results(html): + raise generic.HaltParseException - #Continue only if one Release is found - if len(torrent_rows) < 2: - logger.log(u"The data returned from " + self.name + " does not contain any torrents", - logger.DEBUG) - continue + with BS4Parser(html, 'html.parser') as soup: + torrent_table = soup.find('table', attrs={'class': 'koptekst'}) + torrent_rows = [] if not torrent_table else torrent_table.find_all('tr') - for result in torrent_rows[1:]: - cells = result.find_all('td') - - link = cells[1].find('a') - download_url = self.urls['download'] % cells[2].find('a')['href'] - - id = link['href'] - id = id.replace('details.php?id=','') - id = id.replace('&hit=1', '') + if 2 > len(torrent_rows): + raise generic.HaltParseException + for tr in torrent_rows[1:]: try: - title = link.getText() - id = int(id) - seeders = int(cells[10].getText()) - leechers = int(cells[11].getText()) + seeders, leechers = [int(tr.find_all('td')[x].get_text().strip()) for x in (-3, -2)] + if 'Cache' != mode and (seeders < self.minseed or leechers < self.minleech): + continue + + info = tr.find('a', href=rc['info']) + title = info.get_text().strip() + + download_url = self.urls['get'] % str(tr.find('a', href=rc['get'])['href']).lstrip('/') except (AttributeError, TypeError): continue - #Filter unseeded torrent - if mode != 'RSS' and (seeders < self.minseed or leechers < self.minleech): - continue + if title and download_url: + items[mode].append((title, download_url, seeders)) - if not title or not download_url: - continue + except generic.HaltParseException: + pass + except Exception: + logger.log(u'Failed to parse. Traceback: %s' % traceback.format_exc(), logger.ERROR) + self._log_result(mode, len(items[mode]) - cnt, search_url) - item = title, download_url, id, seeders, leechers - logger.log(u"Found result: " + title + "(" + searchURL + ")", logger.DEBUG) - - items[mode].append(item) - - except Exception as e: - logger.log(u"Failed parsing " + self.name + " Traceback: " + traceback.format_exc(), logger.ERROR) - - #For each search mode sort all the items by seeders - items[mode].sort(key=lambda tup: tup[3], reverse=True) + # for each search mode sort all the items by seeders + items[mode].sort(key=lambda tup: tup[2], reverse=True) results += items[mode] return results - def _get_title_and_url(self, item): - - title, url, id, seeders, leechers = item - - if url: - url = str(url).replace('&', '&') - - return (title, url) - def findPropers(self, search_date=datetime.datetime.today()): - results = [] + return self._find_propers(search_date) - myDB = db.DBConnection() - sqlResults = myDB.select( - 'SELECT s.show_name, e.showid, e.season, e.episode, e.status, e.airdate FROM tv_episodes AS e' + - ' INNER JOIN tv_shows AS s ON (e.showid = s.indexer_id)' + - ' WHERE e.airdate >= ' + str(search_date.toordinal()) + - ' AND (e.status IN (' + ','.join([str(x) for x in Quality.DOWNLOADED]) + ')' + - ' OR (e.status IN (' + ','.join([str(x) for x in Quality.SNATCHED]) + ')))' - ) + def _get_episode_search_strings(self, ep_obj, add_string='', **kwargs): - if not sqlResults: - return [] - - for sqlshow in sqlResults: - self.show = helpers.findCertainShow(sickbeard.showList, int(sqlshow["showid"])) - if self.show: - curEp = self.show.getEpisode(int(sqlshow["season"]), int(sqlshow["episode"])) - - searchString = self._get_episode_search_strings(curEp, add_string='PROPER|REPACK') - - for item in self._doSearch(searchString[0]): - title, url = self._get_title_and_url(item) - results.append(classes.Proper(title, url, datetime.datetime.today(), self.show)) - - return results - - def seedRatio(self): - return self.ratio + return generic.TorrentProvider._get_episode_search_strings(self, ep_obj, add_string, sep_date='|', use_or=False) class BitSoupCache(tvcache.TVCache): - def __init__(self, provider): - tvcache.TVCache.__init__(self, provider) + def __init__(self, this_provider): + tvcache.TVCache.__init__(self, this_provider) - # only poll TorrentBytes every 20 minutes max - self.minTime = 20 + self.minTime = 20 # cache update frequency def _getRSSData(self): - search_params = {'RSS': ['']} - return self.provider._doSearch(search_params) + + return self.provider.get_cache_data() provider = BitSoupProvider()