0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-18 01:35:12 +00:00
yt-dlp/yt_dlp/extractor/viously.py

61 lines
2.5 KiB
Python
Raw Normal View History

import base64
import re
from .common import InfoExtractor
from ..utils import (
extract_attributes,
int_or_none,
parse_iso8601,
)
from ..utils.traversal import traverse_obj
class ViouslyIE(InfoExtractor):
_VALID_URL = False
_WEBPAGE_TESTS = [{
'url': 'http://www.turbo.fr/videos-voiture/454443-turbo-du-07-09-2014-renault-twingo-3-bentley-continental-gt-speed-ces-guide-achat-dacia.html',
'md5': '37a6c3381599381ff53a7e1e0575c0bc',
'info_dict': {
'id': 'F_xQzS2jwb3',
'ext': 'mp4',
'title': 'Turbo du 07/09/2014\xa0: Renault Twingo 3, Bentley Continental GT Speed, CES, Guide Achat Dacia...',
'description': 'Turbo du 07/09/2014\xa0: Renault Twingo 3, Bentley Continental GT Speed, CES, Guide Achat Dacia...',
'age_limit': 0,
'upload_date': '20230328',
'timestamp': 1680037507,
'duration': 3716,
'categories': ['motors'],
}
}]
def _extract_from_webpage(self, url, webpage):
viously_players = re.findall(r'<div[^>]*class="(?:[^"]*\s)?v(?:iou)?sly-player(?:\s[^"]*)?"[^>]*>', webpage)
if not viously_players:
return
def custom_decode(text):
STANDARD_ALPHABET = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/='
CUSTOM_ALPHABET = 'VIOUSLYABCDEFGHJKMNPQRTWXZviouslyabcdefghjkmnpqrtwxz9876543210+/='
data = base64.b64decode(text.translate(str.maketrans(CUSTOM_ALPHABET, STANDARD_ALPHABET)))
return data.decode('utf-8').strip('\x00')
for video_id in traverse_obj(viously_players, (..., {extract_attributes}, 'id')):
formats = self._extract_m3u8_formats(
f'https://www.viously.com/video/hls/{video_id}/index.m3u8', video_id, fatal=False)
if not formats:
continue
data = self._download_json(
f'https://www.viously.com/export/json/{video_id}', video_id,
transform_source=custom_decode, fatal=False)
yield {
'id': video_id,
'formats': formats,
**traverse_obj(data, ('video', {
'title': ('title', {str}),
'description': ('description', {str}),
'duration': ('duration', {int_or_none}),
'timestamp': ('iso_date', {parse_iso8601}),
'categories': ('category', 'name', {str}, {lambda x: [x] if x else None}),
})),
}