yt-dlp/yt_dlp/extractor/kakao.py

153 lines
5.8 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

from .common import InfoExtractor
from ..networking.exceptions import HTTPError
from ..utils import (
ExtractorError,
int_or_none,
str_or_none,
strip_or_none,
traverse_obj,
unified_timestamp,
)
class KakaoIE(InfoExtractor):
_VALID_URL = r'https?://(?:play-)?tv\.kakao\.com/(?:channel/\d+|embed/player)/cliplink/(?P<id>\d+|[^?#&]+@my)'
_API_BASE_TMPL = 'http://tv.kakao.com/api/v1/ft/playmeta/cliplink/%s/'
_CDN_API = 'https://tv.kakao.com/katz/v1/ft/cliplink/%s/readyNplay?'
_TESTS = [{
'url': 'http://tv.kakao.com/channel/2671005/cliplink/301965083',
'md5': '702b2fbdeb51ad82f5c904e8c0766340',
'info_dict': {
'id': '301965083',
'ext': 'mp4',
'title': '乃木坂46 バナナマン 「3期生紹介コーナーが始動顔高低差GPも」 『乃木坂工事中』',
'description': '',
'uploader_id': '2671005',
'uploader': '그랑그랑이',
'timestamp': 1488160199,
'upload_date': '20170227',
'like_count': int,
'thumbnail': r're:http://.+/thumb\.png',
'tags': ['乃木坂'],
'view_count': int,
'duration': 1503,
'comment_count': int,
},
}, {
'url': 'http://tv.kakao.com/channel/2653210/cliplink/300103180',
'md5': 'a8917742069a4dd442516b86e7d66529',
'info_dict': {
'id': '300103180',
'ext': 'mp4',
'description': '러블리즈 - Destiny (나의 지구) (Lovelyz - Destiny)\r\n\r\n[쇼! 음악중심] 20160611, 507회',
'title': '러블리즈 - Destiny (나의 지구) (Lovelyz - Destiny)',
'uploader_id': '2653210',
'uploader': '쇼! 음악중심',
'timestamp': 1485684628,
'upload_date': '20170129',
'like_count': int,
'thumbnail': r're:http://.+/thumb\.png',
'tags': 'count:28',
'view_count': int,
'duration': 184,
'comment_count': int,
},
}, {
# geo restricted
'url': 'https://tv.kakao.com/channel/3643855/cliplink/412069491',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
api_base = self._API_BASE_TMPL % video_id
cdn_api_base = self._CDN_API % video_id
query = {
'player': 'monet_html5',
'referer': url,
'uuid': '',
'service': 'kakao_tv',
'section': '',
'dteType': 'PC',
'fields': ','.join([
'-*', 'tid', 'clipLink', 'displayTitle', 'clip', 'title',
'description', 'channelId', 'createTime', 'duration', 'playCount',
'likeCount', 'commentCount', 'tagList', 'channel', 'name',
'clipChapterThumbnailList', 'thumbnailUrl', 'timeInSec', 'isDefault',
'videoOutputList', 'width', 'height', 'kbps', 'profile', 'label']),
}
api_json = self._download_json(
api_base, video_id, 'Downloading video info')
clip_link = api_json['clipLink']
clip = clip_link['clip']
title = clip.get('title') or clip_link.get('displayTitle')
formats = []
for fmt in clip.get('videoOutputList') or []:
profile_name = fmt.get('profile')
if not profile_name or profile_name == 'AUDIO':
continue
query.update({
'profile': profile_name,
'fields': '-*,code,message,url',
})
try:
fmt_url_json = self._download_json(
cdn_api_base, video_id, query=query,
note=f'Downloading video URL for profile {profile_name}')
except ExtractorError as e:
if isinstance(e.cause, HTTPError) and e.cause.status == 403:
resp = self._parse_json(e.cause.response.read().decode(), video_id)
if resp.get('code') == 'GeoBlocked':
self.raise_geo_restricted()
raise
fmt_url = traverse_obj(fmt_url_json, ('videoLocation', 'url'))
if not fmt_url:
continue
formats.append({
'url': fmt_url,
'format_id': profile_name,
'width': int_or_none(fmt.get('width')),
'height': int_or_none(fmt.get('height')),
'format_note': fmt.get('label'),
'filesize': int_or_none(fmt.get('filesize')),
'tbr': int_or_none(fmt.get('kbps')),
})
thumbs = []
for thumb in clip.get('clipChapterThumbnailList') or []:
thumbs.append({
'url': thumb.get('thumbnailUrl'),
'id': str(thumb.get('timeInSec')),
'preference': -1 if thumb.get('isDefault') else 0,
})
top_thumbnail = clip.get('thumbnailUrl')
if top_thumbnail:
thumbs.append({
'url': top_thumbnail,
'preference': 10,
})
return {
'id': video_id,
'title': title,
'description': strip_or_none(clip.get('description')),
'uploader': traverse_obj(clip_link, ('channel', 'name')),
'uploader_id': str_or_none(clip_link.get('channelId')),
'thumbnails': thumbs,
'timestamp': unified_timestamp(clip_link.get('createTime')),
'duration': int_or_none(clip.get('duration')),
'view_count': int_or_none(clip.get('playCount')),
'like_count': int_or_none(clip.get('likeCount')),
'comment_count': int_or_none(clip.get('commentCount')),
'formats': formats,
'tags': clip.get('tagList'),
}