1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-22 18:22:21 +00:00
youtube-dl/youtube_dl/extractor/gameinformer.py
2015-12-04 08:24:04 +01:00

43 lines
1.5 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..compat import compat_str
from ..utils import int_or_none
class GameInformerIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?gameinformer\.com/(?:[^/]+/)*(?P<id>.+)\.aspx'
_TEST = {
'url': 'http://www.gameinformer.com/b/features/archive/2015/09/26/replay-animal-crossing.aspx',
'info_dict': {
'id': '4515472681001',
'ext': 'm3u8',
'title': 'Replay - Animal Crossing',
'description': 'md5:2e211891b215c85d061adc7a4dd2d930',
'timestamp': 1443457610706,
},
'params': {
# m3u8 download
'skip_download': True,
},
}
def _real_extract(self, url):
display_id = self._match_id(url)
webpage = self._download_webpage(url, display_id)
bc_api_url = self._search_regex(r"getVideo\('([^']+)'", webpage, 'brightcove api url')
json_data = self._download_json(
bc_api_url + '&video_fields=id,name,shortDescription,publishedDate,videoStillURL,length,IOSRenditions',
display_id)
return {
'id': compat_str(json_data['id']),
'display_id': display_id,
'url': json_data['IOSRenditions'][0]['url'],
'title': json_data['name'],
'description': json_data.get('shortDescription'),
'timestamp': int_or_none(json_data.get('publishedDate')),
'duration': int_or_none(json_data.get('length')),
}