diff options
Diffstat (limited to 'module/plugins/crypter/EmbeduploadCom.py')
-rw-r--r-- | module/plugins/crypter/EmbeduploadCom.py | 60 |
1 files changed, 60 insertions, 0 deletions
diff --git a/module/plugins/crypter/EmbeduploadCom.py b/module/plugins/crypter/EmbeduploadCom.py new file mode 100644 index 000000000..3265a939e --- /dev/null +++ b/module/plugins/crypter/EmbeduploadCom.py @@ -0,0 +1,60 @@ +# -*- coding: utf-8 -*- + +import re +from pyload.plugin.Crypter import Crypter +from pyload.network.HTTPRequest import BadHeader + + +class EmbeduploadCom(Crypter): + __name__ = "EmbeduploadCom" + __type__ = "crypter" + __version__ = "0.02" + + __pattern__ = r'http://(?:www\.)?embedupload\.com/\?d=.*' + __config__ = [("use_subfolder", "bool", "Save package to subfolder", True), + ("subfolder_per_package", "bool", "Create a subfolder for each package", True), + ("preferedHoster", "str", "Prefered hoster list (bar-separated)", "embedupload"), + ("ignoredHoster", "str", "Ignored hoster list (bar-separated)", "")] + + __description__ = """EmbedUpload.com decrypter plugin""" + __license__ = "GPLv3" + __authors__ = [("zoidberg", "zoidberg@mujmail.cz")] + + + LINK_PATTERN = r'<div id="([^"]+)"[^>]*>\s*<a href="([^"]+)" target="_blank" (?:class="DownloadNow"|style="color:red")>' + + + def decrypt(self, pyfile): + self.html = self.load(pyfile.url, decode=True) + tmp_links = [] + + m = re.findall(self.LINK_PATTERN, self.html) + if m: + prefered_set = set(self.getConfig("preferedHoster").split('|')) + prefered_set = map(lambda s: s.lower().split('.')[0], prefered_set) + + self.logDebug("PF: %s" % prefered_set) + + tmp_links.extend([x[1] for x in m if x[0] in prefered_set]) + self.urls = self.getLocation(tmp_links) + + if not self.urls: + ignored_set = set(self.getConfig("ignoredHoster").split('|')) + ignored_set = map(lambda s: s.lower().split('.')[0], ignored_set) + + self.logDebug("IG: %s" % ignored_set) + + tmp_links.extend([x[1] for x in m if x[0] not in ignored_set]) + self.urls = self.getLocation(tmp_links) + + + def getLocation(self, tmp_links): + new_links = [] + for link in tmp_links: + try: + header = self.load(link, just_header=True) + if 'location' in header: + new_links.append(header['location']) + except BadHeader: + pass + return new_links |