1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-25 03:32:05 +00:00
youtube-dl/youtube_dl/extractor/jwplatform.py

39 lines
1.3 KiB
Python
Raw Normal View History

# coding: utf-8
from __future__ import unicode_literals
2015-12-21 03:24:58 +00:00
import re
from .common import InfoExtractor
class JWPlatformIE(InfoExtractor):
_VALID_URL = r'(?:https?://content\.jwplatform\.com/(?:feeds|players|jw6)/|jwplatform:)(?P<id>[a-zA-Z0-9]{8})'
_TEST = {
'url': 'http://content.jwplatform.com/players/nPripu9l-ALJ3XQCI.js',
'md5': 'fa8899fa601eb7c83a64e9d568bdf325',
'info_dict': {
'id': 'nPripu9l',
'ext': 'mov',
'title': 'Big Buck Bunny Trailer',
'description': 'Big Buck Bunny is a short animated film by the Blender Institute. It is made using free and open source software.',
'upload_date': '20081127',
'timestamp': 1227796140,
}
}
@staticmethod
def _extract_url(webpage):
urls = JWPlatformIE._extract_urls(webpage)
return urls[0] if urls else None
@staticmethod
def _extract_urls(webpage):
return re.findall(
r'<(?:script|iframe)[^>]+?src=["\']((?:https?:)?//content\.jwplatform\.com/players/[a-zA-Z0-9]{8})',
webpage)
def _real_extract(self, url):
video_id = self._match_id(url)
json_data = self._download_json('http://content.jwplatform.com/feeds/%s.json' % video_id, video_id)
return self._parse_jwplayer_data(json_data, video_id)