mirror of
https://github.com/l1ving/youtube-dl
synced 2024-11-22 21:12:50 +08:00
67 lines
2.0 KiB
Python
67 lines
2.0 KiB
Python
import json
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
|
|
|
|
class TriluliluIE(InfoExtractor):
|
|
_VALID_URL = r'(?x)(?:https?://)?(?:www\.)?trilulilu\.ro/video-(?P<category>[^/]+)/(?P<video_id>[^/]+)'
|
|
_TEST = {
|
|
u"url": u"http://www.trilulilu.ro/video-animatie/big-buck-bunny-1",
|
|
u'file': u"big-buck-bunny-1.mp4",
|
|
u'info_dict': {
|
|
u"title": u"Big Buck Bunny",
|
|
u"description": u":) pentru copilul din noi",
|
|
},
|
|
# Server ignores Range headers (--test)
|
|
u"params": {
|
|
u"skip_download": True
|
|
}
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
video_id = mobj.group('video_id')
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
title = self._og_search_title(webpage)
|
|
thumbnail = self._og_search_thumbnail(webpage)
|
|
description = self._og_search_description(webpage)
|
|
|
|
log_str = self._search_regex(
|
|
r'block_flash_vars[ ]=[ ]({[^}]+})', webpage, u'log info')
|
|
log = json.loads(log_str)
|
|
|
|
format_url = (u'http://fs%(server)s.trilulilu.ro/%(hash)s/'
|
|
u'video-formats2' % log)
|
|
format_doc = self._download_xml(
|
|
format_url, video_id,
|
|
note=u'Downloading formats',
|
|
errnote=u'Error while downloading formats')
|
|
|
|
video_url_template = (
|
|
u'http://fs%(server)s.trilulilu.ro/stream.php?type=video'
|
|
u'&source=site&hash=%(hash)s&username=%(userid)s&'
|
|
u'key=ministhebest&format=%%s&sig=&exp=' %
|
|
log)
|
|
formats = [
|
|
{
|
|
'format': fnode.text,
|
|
'url': video_url_template % fnode.text,
|
|
'ext': fnode.text.partition('-')[0]
|
|
}
|
|
|
|
for fnode in format_doc.findall('./formats/format')
|
|
]
|
|
|
|
return {
|
|
'_type': 'video',
|
|
'id': video_id,
|
|
'formats': formats,
|
|
'title': title,
|
|
'description': description,
|
|
'thumbnail': thumbnail,
|
|
}
|
|
|