1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-16 16:55:51 +00:00
youtube-dl/youtube_dl/extractor/ehow.py
2015-07-17 23:54:38 +06:00

38 lines
1.5 KiB
Python

from __future__ import unicode_literals
from .common import InfoExtractor
from ..compat import compat_urllib_parse_unquote
class EHowIE(InfoExtractor):
IE_NAME = 'eHow'
_VALID_URL = r'https?://(?:www\.)?ehow\.com/[^/_?]*_(?P<id>[0-9]+)'
_TEST = {
'url': 'http://www.ehow.com/video_12245069_hardwood-flooring-basics.html',
'md5': '9809b4e3f115ae2088440bcb4efbf371',
'info_dict': {
'id': '12245069',
'ext': 'flv',
'title': 'Hardwood Flooring Basics',
'description': 'Hardwood flooring may be time consuming, but its ultimately a pretty straightforward concept. Learn about hardwood flooring basics with help from a hardware flooring business owner in this free video...',
'uploader': 'Erick Nathan',
}
}
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
video_url = self._search_regex(
r'(?:file|source)=(http[^\'"&]*)', webpage, 'video URL')
final_url = compat_urllib_parse_unquote(video_url)
uploader = self._html_search_meta('uploader', webpage)
title = self._og_search_title(webpage).replace(' | eHow', '')
return {
'id': video_id,
'url': final_url,
'title': title,
'thumbnail': self._og_search_thumbnail(webpage),
'description': self._og_search_description(webpage),
'uploader': uploader,
}