0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-18 01:35:12 +00:00
yt-dlp/youtube_dl/extractor/techtalks.py

66 lines
2 KiB
Python
Raw Normal View History

import re
from .common import InfoExtractor
from ..utils import (
get_element_by_attribute,
clean_html,
)
class TechTalksIE(InfoExtractor):
_VALID_URL = r'https?://techtalks\.tv/talks/[^/]*/(?P<id>\d+)/'
_TEST = {
u'url': u'http://techtalks.tv/talks/learning-topic-models-going-beyond-svd/57758/',
u'playlist': [
{
u'file': u'57758.flv',
u'info_dict': {
u'title': u'Learning Topic Models --- Going beyond SVD',
},
},
{
u'file': u'57758-slides.flv',
u'info_dict': {
u'title': u'Learning Topic Models --- Going beyond SVD',
},
},
],
u'params': {
# rtmp download
u'skip_download': True,
},
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
talk_id = mobj.group('id')
webpage = self._download_webpage(url, talk_id)
rtmp_url = self._search_regex(r'netConnectionUrl: \'(.*?)\'', webpage,
u'rtmp url')
play_path = self._search_regex(r'href=\'(.*?)\' [^>]*id="flowplayer_presenter"',
webpage, u'presenter play path')
title = clean_html(get_element_by_attribute('class', 'title', webpage))
video_info = {
'id': talk_id,
'title': title,
'url': rtmp_url,
'play_path': play_path,
'ext': 'flv',
}
m_slides = re.search(r'<a class="slides" href=\'(.*?)\'', webpage)
if m_slides is None:
return video_info
else:
return [
video_info,
# The slides video
{
'id': talk_id + '-slides',
'title': title,
'url': rtmp_url,
'play_path': m_slides.group(1),
'ext': 'flv',
},
]