1
1
mirror of https://github.com/ytdl-org/youtube-dl synced 2024-12-05 00:52:55 +01:00
youtube-dl/youtube_dl/extractor/tonline.py
2016-10-04 08:00:25 +01:00

60 lines
2.0 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..utils import int_or_none
class TOnlineIE(InfoExtractor):
IE_NAME = 't-online.de'
_VALID_URL = r'https?://(?:www\.)?t-online\.de/tv/(?:[^/]+/)*id_(?P<id>\d+)'
_TEST = {
'url': 'http://www.t-online.de/tv/sport/fussball/id_79166266/drittes-remis-zidane-es-muss-etwas-passieren-.html',
'md5': '7d94dbdde5f9d77c5accc73c39632c29',
'info_dict': {
'id': '79166266',
'ext': 'mp4',
'title': 'Drittes Remis! Zidane: "Es muss etwas passieren"',
'description': 'Es läuft nicht rund bei Real Madrid. Das 1:1 gegen den SD Eibar war das dritte Unentschieden in Folge in der Liga.',
}
}
def _real_extract(self, url):
video_id = self._match_id(url)
video_data = self._download_json(
'http://www.t-online.de/tv/id_%s/tid_json_video' % video_id, video_id)
title = video_data['subtitle']
formats = []
for asset in video_data.get('assets', []):
asset_source = asset.get('source') or asset.get('source2')
if not asset_source:
continue
formats_id = []
for field_key in ('type', 'profile'):
field_value = asset.get(field_key)
if field_value:
formats_id.append(field_value)
formats.append({
'format_id': '-'.join(formats_id),
'url': asset_source,
})
thumbnails = []
for image in video_data.get('images', []):
image_source = image.get('source')
if not image_source:
continue
thumbnails.append({
'url': image_source,
})
return {
'id': video_id,
'title': title,
'description': video_data.get('description'),
'duration': int_or_none(video_data.get('duration')),
'thumbnails': thumbnails,
'formats': formats,
}