1
0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-12-11 16:15:46 +01:00
yt-dlp/yt_dlp/extractor/indavideo.py
sepro add96eb9f8
[cleanup] Add more ruff rules (#10149)
Authored by: seproDev

Reviewed-by: bashonly <88596187+bashonly@users.noreply.github.com>
Reviewed-by: Simon Sawicki <contact@grub4k.xyz>
2024-06-12 01:09:58 +02:00

116 lines
4.1 KiB
Python

from .common import InfoExtractor
from ..utils import (
int_or_none,
parse_age_limit,
parse_iso8601,
time_seconds,
update_url_query,
)
class IndavideoEmbedIE(InfoExtractor):
_VALID_URL = r'https?://(?:(?:embed\.)?indavideo\.hu/player/video/|assets\.indavideo\.hu/swf/player\.swf\?.*\b(?:v(?:ID|id))=)(?P<id>[\da-f]+)'
# Some example URLs covered by generic extractor:
# https://indavideo.hu/video/Vicces_cica_1
# https://index.indavideo.hu/video/Hod_Nemetorszagban
# https://auto.indavideo.hu/video/Sajat_utanfutoban_a_kis_tacsko
# https://film.indavideo.hu/video/f_farkaslesen
# https://palyazat.indavideo.hu/video/Embertelen_dal_Dodgem_egyuttes
_EMBED_REGEX = [r'<iframe[^>]+\bsrc=["\'](?P<url>(?:https?:)//embed\.indavideo\.hu/player/video/[\da-f]+)']
_TESTS = [{
'url': 'https://indavideo.hu/player/video/1bdc3c6d80/',
'md5': 'c8a507a1c7410685f83a06eaeeaafeab',
'info_dict': {
'id': '1837039',
'ext': 'mp4',
'title': 'Cicatánc',
'description': '',
'thumbnail': r're:^https?://.*\.jpg$',
'uploader': 'cukiajanlo',
'uploader_id': '83729',
'timestamp': 1439193826,
'upload_date': '20150810',
'duration': 72,
'age_limit': 0,
'tags': ['tánc', 'cica', 'cuki', 'cukiajanlo', 'newsroom'],
},
}, {
'url': 'https://embed.indavideo.hu/player/video/1bdc3c6d80?autostart=1&hide=1',
'only_matching': True,
}]
_WEBPAGE_TESTS = [{
'url': 'https://indavideo.hu/video/Vicces_cica_1',
'info_dict': {
'id': '1335611',
'ext': 'mp4',
'title': 'Vicces cica',
'description': 'Játszik a tablettel. :D',
'thumbnail': r're:^https?://.*\.jpg$',
'uploader': 'Jet_Pack',
'uploader_id': '491217',
'timestamp': 1390821212,
'upload_date': '20140127',
'duration': 7,
'age_limit': 0,
'tags': ['cica', 'Jet_Pack'],
},
}]
def _real_extract(self, url):
video_id = self._match_id(url)
video = self._download_json(
f'https://amfphp.indavideo.hu/SYm0json.php/player.playerHandler.getVideoData/{video_id}/',
video_id, query={'_': time_seconds()})['data']
video_urls = []
video_files = video.get('video_files')
if isinstance(video_files, list):
video_urls.extend(video_files)
elif isinstance(video_files, dict):
video_urls.extend(video_files.values())
video_urls = list(set(video_urls))
filesh = video.get('filesh') or {}
formats = []
for video_url in video_urls:
height = int_or_none(self._search_regex(
r'\.(\d{3,4})\.mp4(?:\?|$)', video_url, 'height', default=None))
if not height and len(filesh) == 1:
height = int_or_none(next(iter(filesh.keys())))
token = filesh.get(str(height))
if token is None:
continue
formats.append({
'url': update_url_query(video_url, {'token': token}),
'height': height,
})
timestamp = video.get('date')
if timestamp:
# upload date is in CEST
timestamp = parse_iso8601(timestamp + ' +0200', ' ')
thumbnails = [{
'url': self._proto_relative_url(thumbnail),
} for thumbnail in video.get('thumbnails', [])]
tags = [tag['title'] for tag in video.get('tags') or []]
return {
'id': video.get('id') or video_id,
'title': video.get('title'),
'description': video.get('description'),
'thumbnails': thumbnails,
'uploader': video.get('user_name'),
'uploader_id': video.get('user_id'),
'timestamp': timestamp,
'duration': int_or_none(video.get('length')),
'age_limit': parse_age_limit(video.get('age_limit')),
'tags': tags,
'formats': formats,
}