1
0
mirror of https://github.com/spaam/svtplay-dl.git synced 2024-11-24 20:25:41 +01:00
svtplay-dl/lib/svtplay_dl/service/mtvnn.py
2015-08-31 22:25:37 +02:00

78 lines
2.8 KiB
Python

from __future__ import absolute_import
import re
import os
import xml.etree.ElementTree as ET
from svtplay_dl.service import Service, OpenGraphThumbMixin
from svtplay_dl.utils import is_py2_old
from svtplay_dl.utils.urllib import urlparse
from svtplay_dl.log import log
from svtplay_dl.fetcher.rtmp import RTMP
# This is _very_ similar to mtvservices..
class Mtvnn(Service, OpenGraphThumbMixin):
supported_domains = ['nickelodeon.se', "nickelodeon.nl", "nickelodeon.no"]
def get(self, options):
data = self.get_urldata()
match = re.search(r'"(http://api.mtvnn.com/v2/mrss.xml[^"]+)"', data)
if not match:
log.error("Can't find id for the video")
return
data = self.http.request("get", match.group(1)).content
xml = ET.XML(data)
mediagen = xml.find("channel").find("item").find("{http://search.yahoo.com/mrss/}group")
title = xml.find("channel").find("item").find("title").text
if options.output_auto:
directory = os.path.dirname(options.output)
if len(directory):
options.output = os.path.join(directory, title)
else:
options.output = title
if self.exclude(options):
return
swfurl = mediagen.find("{http://search.yahoo.com/mrss/}player").attrib["url"]
parse = urlparse(swfurl)
options.other = "-W %s://%s%s" % (parse.scheme, parse.hostname, self.http.check_redirect(swfurl))
contenturl = mediagen.find("{http://search.yahoo.com/mrss/}content").attrib["url"]
content = self.http.request("get", contenturl).content
xml = ET.XML(content)
ss = xml.find("video").find("item")
if is_py2_old:
sa = list(ss.getiterator("rendition"))
else:
sa = list(ss.iter("rendition"))
for i in sa:
yield RTMP(options, i.find("src").text, i.attrib["bitrate"])
def find_all_episodes(self, options):
match = re.search(r"data-franchise='([^']+)'", self.get_urldata())
if match is None:
log.error("Couldn't program id")
return
programid = match.group(1)
match = re.findall(r"<li class='(divider playlist-item|playlist-item)'( data-item-id='([^']+)')?", self.get_urldata())
if not match:
log.error("Couldn't retrieve episode list")
return
episodNr = []
for i in match:
if i[0] == "playlist-item":
episodNr.append(i[2])
else:
break
episodes = []
n = 0
for i in sorted(episodNr):
if n == options.all_last:
break
episodes.append("http://www.nickelodeon.se/serier/%s-something/videos/%s-something" % (programid, i))
n += 1
return episodes