1
0
mirror of https://codeberg.org/polarisfm/youtube-dl synced 2024-11-14 21:54:33 +01:00
youtube-dl/youtube_dl/extractor/jeuxvideo.py
2017-10-09 23:50:53 +07:00

57 lines
2.0 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
class JeuxVideoIE(InfoExtractor):
_VALID_URL = r'https?://.*?\.jeuxvideo\.com/.*/(.*?)\.htm'
_TESTS = [{
'url': 'http://www.jeuxvideo.com/reportages-videos-jeux/0004/00046170/tearaway-playstation-vita-gc-2013-tearaway-nous-presente-ses-papiers-d-identite-00115182.htm',
'md5': '046e491afb32a8aaac1f44dd4ddd54ee',
'info_dict': {
'id': '114765',
'ext': 'mp4',
'title': 'Tearaway : GC 2013 : Tearaway nous présente ses papiers d\'identité',
'description': 'Lorsque les développeurs de LittleBigPlanet proposent un nouveau titre, on ne peut que s\'attendre à un résultat original et fort attrayant.',
},
}, {
'url': 'http://www.jeuxvideo.com/videos/chroniques/434220/l-histoire-du-jeu-video-la-saturn.htm',
'only_matching': True,
}]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
title = mobj.group(1)
webpage = self._download_webpage(url, title)
title = self._html_search_meta('name', webpage) or self._og_search_title(webpage)
config_url = self._html_search_regex(
r'data-src(?:set-video)?="(/contenu/medias/video\.php.*?)"',
webpage, 'config URL')
config_url = 'http://www.jeuxvideo.com' + config_url
video_id = self._search_regex(
r'id=(\d+)',
config_url, 'video ID')
config = self._download_json(
config_url, title, 'Downloading JSON config')
formats = [{
'url': source['file'],
'format_id': source['label'],
'resolution': source['label'],
} for source in reversed(config['sources'])]
return {
'id': video_id,
'title': title,
'formats': formats,
'description': self._og_search_description(webpage),
'thumbnail': config.get('image'),
}