1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-06-01 01:49:30 +00:00

[medaltv] Add new extractor (#27149)

This commit is contained in:
Joshua Lochner 2020-11-24 19:27:33 +02:00 committed by GitHub
parent 01c92973dd
commit e24ebeafd1
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 139 additions and 0 deletions

View file

@ -606,6 +606,7 @@ from .markiza import (
from .massengeschmacktv import MassengeschmackTVIE from .massengeschmacktv import MassengeschmackTVIE
from .matchtv import MatchTVIE from .matchtv import MatchTVIE
from .mdr import MDRIE from .mdr import MDRIE
from .medaltv import MedalTVIE
from .mediaset import MediasetIE from .mediaset import MediasetIE
from .mediasite import ( from .mediasite import (
MediasiteIE, MediasiteIE,

View file

@ -0,0 +1,138 @@
# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..utils import (
ExtractorError,
try_get,
float_or_none,
int_or_none
)
class MedalTVIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?medal\.tv/clips/(?P<id>[0-9]+)'
_TESTS = [{
'url': 'https://medal.tv/clips/34934644/3Is9zyGMoBMr',
'md5': '7b07b064331b1cf9e8e5c52a06ae68fa',
'info_dict': {
'id': '34934644',
'ext': 'mp4',
'title': 'Quad Cold',
'description': 'Medal,https://medal.tv/desktop/',
'uploader': 'MowgliSB',
'timestamp': 1603165266,
'upload_date': '20201020',
'uploader_id': 10619174,
}
}, {
'url': 'https://medal.tv/clips/36787208',
'md5': 'b6dc76b78195fff0b4f8bf4a33ec2148',
'info_dict': {
'id': '36787208',
'ext': 'mp4',
'title': 'u tk me i tk u bigger',
'description': 'Medal,https://medal.tv/desktop/',
'uploader': 'Mimicc',
'timestamp': 1605580939,
'upload_date': '20201117',
'uploader_id': 5156321,
}
}]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
hydration_data = self._search_regex(
r'<script[^>]*>\s*(?:var\s*)?hydrationData\s*=\s*({.+?})\s*</script>',
webpage, 'hydration data', default='{}')
parsed = self._parse_json(hydration_data, video_id)
clip_info = try_get(parsed, lambda x: x['clips'][video_id], dict) or {}
if not clip_info:
raise ExtractorError('Could not find video information.',
video_id=video_id)
width = int_or_none(clip_info.get('sourceWidth'))
height = int_or_none(clip_info.get('sourceHeight'))
aspect_ratio = (width / height) if(width and height) else (16 / 9)
# ordered from lowest to highest resolution
heights = (144, 240, 360, 480, 720, 1080)
formats = []
thumbnails = []
for height in heights:
format_key = '{0}p'.format(height)
video_key = 'contentUrl{0}'.format(format_key)
thumbnail_key = 'thumbnail{0}'.format(format_key)
width = int(round(aspect_ratio * height))
# Second condition needed as sometimes medal says
# they have a format when in fact it is another format.
format_url = clip_info.get(video_key)
if(format_url and format_key in format_url):
formats.append({
'url': format_url,
'format_id': format_key,
'width': width,
'height': height
})
thumbnail_url = clip_info.get(thumbnail_key)
if(thumbnail_url and format_key in thumbnail_url):
thumbnails.append({
'id': format_key,
'url': thumbnail_url,
'width': width,
'height': height
})
# add source to formats
source_url = clip_info.get('contentUrl')
if(source_url):
formats.append({
'url': source_url,
'format_id': 'source',
'width': width,
'height': height
})
error = clip_info.get('error')
if not formats and error:
if(error == 404):
raise ExtractorError('That clip does not exist.',
expected=True, video_id=video_id)
else:
raise ExtractorError('An unknown error occurred ({0}).'.format(error),
video_id=video_id)
# Necessary because the id of the author is not known in advance.
# Won't raise an issue if no profile can be found as this is optional.
author_info = try_get(parsed,
lambda x: list(x['profiles'].values())[0], dict
) or {}
author_id = author_info.get('id')
author_url = 'https://medal.tv/users/{0}'.format(author_id) if author_id else None
return {
'id': video_id,
'title': clip_info.get('contentTitle'),
'formats': formats,
'thumbnails': thumbnails,
'description': clip_info.get('contentDescription'),
'uploader': author_info.get('displayName'),
'timestamp': float_or_none(clip_info.get('created'), 1000),
'uploader_id': author_id,
'uploader_url': author_url,
'duration': float_or_none(clip_info.get('videoLengthSeconds')),
'view_count': int_or_none(clip_info.get('views')),
'like_count': int_or_none(clip_info.get('likes')),
'comment_count': int_or_none(clip_info.get('comments'))
}