mirror of
https://github.com/l1ving/youtube-dl
synced 2024-12-22 21:42:57 +08:00
54 lines
1.8 KiB
Python
54 lines
1.8 KiB
Python
from __future__ import unicode_literals
|
|
|
|
import json
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
|
|
|
|
class NineGagIE(InfoExtractor):
|
|
IE_NAME = '9gag'
|
|
_VALID_URL = r'^https?://(?:www\.)?9gag\.tv/v/(?P<id>[0-9]+)'
|
|
|
|
_TEST = {
|
|
"url": "http://9gag.tv/v/1912",
|
|
"file": "1912.mp4",
|
|
"info_dict": {
|
|
"description": "This 3-minute video will make you smile and then make you feel untalented and insignificant. Anyway, you should share this awesomeness. (Thanks, Dino!)",
|
|
"title": "\"People Are Awesome 2013\" Is Absolutely Awesome",
|
|
"view_count": int,
|
|
"thumbnail": "re:^https?://",
|
|
},
|
|
'add_ie': ['Youtube']
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
video_id = mobj.group('id')
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
youtube_id = self._html_search_regex(
|
|
r'(?s)id="jsid-video-post-container".*?data-external-id="([^"]+)"',
|
|
webpage, 'video ID')
|
|
description = self._html_search_regex(
|
|
r'(?s)<div class="video-caption">.*?<p>(.*?)</p>', webpage,
|
|
'description', fatal=False)
|
|
view_count_str = self._html_search_regex(
|
|
r'<p><b>([0-9][0-9,]*)</b> views</p>', webpage, 'view count',
|
|
fatal=False)
|
|
view_count = (
|
|
None if view_count_str is None
|
|
else int(view_count_str.replace(',', '')))
|
|
|
|
return {
|
|
'_type': 'url_transparent',
|
|
'url': youtube_id,
|
|
'ie_key': 'Youtube',
|
|
'id': video_id,
|
|
'title': self._og_search_title(webpage),
|
|
'description': description,
|
|
'view_count': view_count,
|
|
'thumbnail': self._og_search_thumbnail(webpage),
|
|
}
|