2015-01-09 14:11:21 +03:00
|
|
|
# -*- coding: utf-8 -*-
|
|
|
|
'''
|
|
|
|
Torrenter plugin for XBMC
|
|
|
|
Copyright (C) 2012 Vadim Skorba
|
|
|
|
vadim.skorba@gmail.com
|
|
|
|
|
|
|
|
This program 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.
|
|
|
|
|
|
|
|
This program 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 this program. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
'''
|
|
|
|
|
|
|
|
import urllib
|
|
|
|
import re
|
|
|
|
import sys
|
|
|
|
|
|
|
|
import SearcherABC
|
|
|
|
|
|
|
|
|
|
|
|
class RuTorOrg(SearcherABC.SearcherABC):
|
|
|
|
'''
|
|
|
|
Weight of source with this searcher provided.
|
|
|
|
Will be multiplied on default weight.
|
|
|
|
Default weight is seeds number
|
|
|
|
'''
|
|
|
|
sourceWeight = 1
|
|
|
|
|
|
|
|
'''
|
|
|
|
Relative (from root directory of plugin) path to image
|
|
|
|
will shown as source image at result listing
|
|
|
|
'''
|
|
|
|
searchIcon = '/resources/searchers/icons/rutor.org.png'
|
|
|
|
|
|
|
|
'''
|
|
|
|
Flag indicates is this source - magnet links source or not.
|
|
|
|
Used for filtration of sources in case of old library (setting selected).
|
|
|
|
Old libraries won't to convert magnet as torrent file to the storage
|
|
|
|
'''
|
|
|
|
|
|
|
|
@property
|
|
|
|
def isMagnetLinkSource(self):
|
|
|
|
return False
|
|
|
|
|
|
|
|
'''
|
|
|
|
Main method should be implemented for search process.
|
|
|
|
Receives keyword and have to return dictionary of proper tuples:
|
|
|
|
filesList.append((
|
|
|
|
int(weight),# Calculated global weight of sources
|
|
|
|
int(seeds),# Seeds count
|
|
|
|
str(title),# Title will be shown
|
|
|
|
str(link),# Link to the torrent/magnet
|
|
|
|
str(image),# Path/URL to image shown at the list
|
|
|
|
))
|
|
|
|
'''
|
|
|
|
|
|
|
|
def search(self, keyword):
|
|
|
|
filesList = []
|
|
|
|
self.timeout(5)
|
|
|
|
#print 'strating Rutor'
|
2015-07-01 23:15:29 +03:00
|
|
|
url = "http://anti-tor.org/search/0/0/100/2/%s" % urllib.quote_plus(keyword)
|
2015-01-09 14:11:21 +03:00
|
|
|
headers = [('User-Agent',
|
|
|
|
'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/37.0.2062.124 YaBrowser/14.10.2062.12061 Safari/537.36'),
|
2015-07-01 23:15:29 +03:00
|
|
|
('Referer', 'http://anti-tor.org'), ('Accept-encoding', 'gzip'),
|
2015-01-09 14:11:21 +03:00
|
|
|
('Cookie', str(sys.modules["__main__"].__settings__.getSetting("rutor-auth")))]
|
|
|
|
response = self.makeRequest(url, headers=headers)
|
|
|
|
|
|
|
|
#print response
|
|
|
|
|
|
|
|
if None != response and 0 < len(response):
|
|
|
|
cookie = re.compile("document.cookie='(.+?)';").findall(response)
|
|
|
|
if cookie:
|
|
|
|
#print 'ok found cookie'
|
|
|
|
cookie = cookie[0]
|
|
|
|
#print str(cookie)
|
|
|
|
sys.modules["__main__"].__settings__.setSetting("rutor-auth", cookie)
|
|
|
|
headers = [('User-Agent',
|
|
|
|
'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/37.0.2062.124 YaBrowser/14.10.2062.12061 Safari/537.36'),
|
2015-07-01 23:15:29 +03:00
|
|
|
('Referer', 'http://anti-tor.org'), ('Accept-encoding', 'gzip'),
|
2015-01-09 14:11:21 +03:00
|
|
|
('Cookie', str(sys.modules["__main__"].__settings__.getSetting("rutor-auth")))]
|
|
|
|
response = self.makeRequest(url, headers=headers)
|
|
|
|
|
2015-05-14 19:20:49 +03:00
|
|
|
#print response
|
|
|
|
|
2015-01-09 14:11:21 +03:00
|
|
|
if None != response and 0 < len(response):
|
|
|
|
dat = re.compile(
|
2015-05-14 19:20:49 +03:00
|
|
|
'<tr class="[tg].+?<a.+?href="(.+?download/\d+)">.+?<a href=".+?">.+?<a href=".+?">(.+?)</a></td>.+?<td align="right">(\d*?\..+? .+?)</td>.+?<img .+?alt="S".+?> (\d+)</span>.+?alt="L".+?> (\d+)',
|
2015-01-09 14:11:21 +03:00
|
|
|
re.DOTALL).findall(response)
|
|
|
|
if dat and len(dat) > 0:
|
|
|
|
for (link, title, size, seeds, leechers) in dat:
|
|
|
|
torrentTitle = title.strip() #"%s [S\L: %s\%s]" % (title, seeds, leechers)
|
|
|
|
size = size.replace(' ', ' ')
|
2015-07-01 23:15:29 +03:00
|
|
|
if link[0]=='/': link='http://anti-tor.org'+link
|
2015-01-09 14:11:21 +03:00
|
|
|
image = sys.modules["__main__"].__root__ + self.searchIcon
|
|
|
|
filesList.append((
|
|
|
|
int(int(self.sourceWeight) * int(seeds)),
|
|
|
|
int(seeds), int(leechers), size,
|
|
|
|
self.unescape(self.stripHtml(torrentTitle)),
|
|
|
|
self.__class__.__name__ + '::' + link,
|
|
|
|
image,
|
|
|
|
))
|
|
|
|
return filesList
|
|
|
|
|