1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-12-22 08:47:41 +00:00

[hitrecord] Add extractor

This commit is contained in:
J 2017-01-06 21:56:59 +01:00 committed by Sergey M․
parent 827961b122
commit 553c68bbd9
No known key found for this signature in database
GPG key ID: 2C393E0F18A9236D
2 changed files with 49 additions and 0 deletions

View file

@ -366,6 +366,7 @@ from .hgtv import (
)
from .historicfilms import HistoricFilmsIE
from .hitbox import HitboxIE, HitboxLiveIE
from .hitrecord import HitRecordIE
from .hornbunny import HornBunnyIE
from .hotnewhiphop import HotNewHipHopIE
from .hotstar import HotStarIE

View file

@ -0,0 +1,48 @@
from __future__ import unicode_literals
from .common import InfoExtractor
from ..utils import (
clean_html,
int_or_none,
unified_strdate,
)
from ..compat import compat_str
class HitRecordIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?hitrecord\.org/records/(?P<id>\d+)'
_TEST = {
'url': 'https://hitrecord.org/records/2954362',
'md5': 'fe1cdc2023bce0bbb95c39c57426aa71',
'info_dict': {
'id': '2954362',
'ext': 'mp4',
'title': 'A Very Different World (HITRECORD x ACLU)',
'description': 'md5:e62defaffab5075a5277736bead95a3d',
'release_date': '20160818',
'timestamp': 1471557582,
'uploader': 'Zuzi.C12',
'uploader_id': '362811',
}
}
def _real_extract(self, url):
video_id = self._match_id(url)
video_info = self._download_json('https://hitrecord.org/api/web/records/' + video_id, video_id)
user_info = video_info.get('user', {})
return {
'id': video_id,
'title': video_info['title'],
'url': video_info['source_url']['mp4_url'],
'description': clean_html(video_info.get('body')),
'uploader': user_info.get('username'),
'uploader_id': compat_str(user_info.get('id')),
'release_date': unified_strdate(video_info.get('created_at')),
'timestamp': video_info.get('created_at_i'),
'view_count': int_or_none(video_info.get('total_views_count')),
'like_count': int_or_none(video_info.get('hearts_count')),
'comment_count': int_or_none(video_info.get('comments_count')),
'tags': [tag.get('text') for tag in video_info.get('tags', [])],
}