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/picsearch.py

36 lines
1.4 KiB
Python
Raw Normal View History

# ex:ts=4:sw=4:sts=4:et
# -*- tab-width: 4; c-basic-offset: 4; indent-tabs-mode: nil -*-
from __future__ import absolute_import
import re
import json
import sys
from svtplay_dl.service import Service, OpenGraphThumbMixin
2014-04-21 21:41:06 +02:00
from svtplay_dl.utils import get_http_data
from svtplay_dl.fetcher.rtmp import RTMP
from svtplay_dl.log import log
class Picsearch(Service, OpenGraphThumbMixin):
supported_domains = ['dn.se']
def get(self, options):
data = self.get_urldata()
ajax_auth = re.search(r"picsearch_ajax_auth = '(\w+)'", data)
if not ajax_auth:
2014-03-31 19:15:40 +02:00
log.error("Cant find token for video")
sys.exit(2)
mediaid = re.search(r"mediaId = '([^']+)';", data)
if not mediaid:
mediaid = re.search(r'media-id="([^"]+)"', data)
if not mediaid:
log.error("Cant find media id")
sys.exit(2)
jsondata = get_http_data("http://csp.picsearch.com/rest?jsonp=&eventParam=1&auth=%s&method=embed&mediaid=%s" % (ajax_auth.group(1), mediaid.group(1)))
jsondata = json.loads(jsondata)
files = jsondata["media"]["playerconfig"]["playlist"][1]["bitrates"]
server = jsondata["media"]["playerconfig"]["plugins"]["bwcheck"]["netConnectionUrl"]
for i in files:
options.other = "-y '%s'" % i["url"]
yield RTMP(options, server, i["height"])