kopia lustrzana https://github.com/yt-dlp/yt-dlp
[vlive] add: support video post urls
rodzic
0536e60b48
commit
5dcfd2508a
|
@ -18,10 +18,10 @@ from ..utils import (
|
|||
|
||||
class VLiveIE(NaverBaseIE):
|
||||
IE_NAME = 'vlive'
|
||||
_VALID_URL = r'https?://(?:(?:www|m)\.)?vlive\.tv/video/(?P<id>[0-9]+)'
|
||||
_VALID_URL = r'https?://(?:(?:www|m)\.)?vlive\.tv/(?:video|post)/(?P<id>(?:\d-)?[0-9]+)'
|
||||
_NETRC_MACHINE = 'vlive'
|
||||
_TESTS = [{
|
||||
'url': 'http://www.vlive.tv/video/1326',
|
||||
'url': 'https://www.vlive.tv/video/1326',
|
||||
'md5': 'cc7314812855ce56de70a06a27314983',
|
||||
'info_dict': {
|
||||
'id': '1326',
|
||||
|
@ -31,8 +31,21 @@ class VLiveIE(NaverBaseIE):
|
|||
'view_count': int,
|
||||
'uploader_id': 'muploader_a',
|
||||
},
|
||||
}, {
|
||||
'url': 'http://www.vlive.tv/video/16937',
|
||||
},
|
||||
{
|
||||
'url': 'https://vlive.tv/post/1-18244258',
|
||||
'md5': 'cc7314812855ce56de70a06a27314983',
|
||||
'info_dict': {
|
||||
'id': '1326',
|
||||
'ext': 'mp4',
|
||||
'title': "[V LIVE] Girl's Day's Broadcast",
|
||||
'creator': "Girl's Day",
|
||||
'view_count': int,
|
||||
'uploader_id': 'muploader_a',
|
||||
},
|
||||
},
|
||||
{
|
||||
'url': 'https://www.vlive.tv/video/16937',
|
||||
'info_dict': {
|
||||
'id': '16937',
|
||||
'ext': 'mp4',
|
||||
|
@ -95,24 +108,30 @@ class VLiveIE(NaverBaseIE):
|
|||
raise ExtractorError('Unable to log in', expected=True)
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
webpage = self._download_webpage(url, video_id)
|
||||
# url may match on a post or a video url with a post_id potentially matching a video_id
|
||||
working_id = self._match_id(url)
|
||||
webpage = self._download_webpage(url, working_id)
|
||||
|
||||
PARAMS_RE = r'window\.__PRELOADED_STATE__\s*=\s*({.*});?\s*</script>'
|
||||
PARAMS_FIELD = 'params'
|
||||
|
||||
params = self._search_regex(
|
||||
PARAMS_RE, webpage, PARAMS_FIELD, default='', flags=re.DOTALL)
|
||||
params = self._parse_json(params, video_id, fatal=False)
|
||||
params = self._parse_json(params, working_id, fatal=False)
|
||||
|
||||
video_params = params["postDetail"]["post"].get("officialVideo")
|
||||
video_params = try_get(params, lambda x: x["postDetail"]["post"]["officialVideo"])
|
||||
if video_params is None:
|
||||
raise ExtractorError('Invalid key: Failed to extract video parameters.')
|
||||
if 'post' in url:
|
||||
raise ExtractorError('Url does not appear to be a video post.')
|
||||
else:
|
||||
raise ExtractorError('Failed to extract video parameters.')
|
||||
|
||||
video_id = working_id if 'video' in url else str(video_params["videoSeq"])
|
||||
long_video_id = video_params["vodId"]
|
||||
video_type = video_params["type"]
|
||||
KEY_ENDPOINT = 'https://www.vlive.tv/globalv-web/vam-web/video/v1.0/vod/%s/inkey' % video_id
|
||||
key_json = self._download_json(KEY_ENDPOINT, video_id,
|
||||
|
||||
VOD_KEY_ENDPOINT = 'https://www.vlive.tv/globalv-web/vam-web/video/v1.0/vod/%s/inkey' % video_id
|
||||
key_json = self._download_json(VOD_KEY_ENDPOINT, video_id,
|
||||
headers={"referer": "https://www.vlive.tv"})
|
||||
key = key_json["inkey"]
|
||||
|
||||
|
|
Ładowanie…
Reference in New Issue