0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-24 02:35:12 +00:00
yt-dlp/yt_dlp/extractor/acfun.py

201 lines
7.8 KiB
Python
Raw Normal View History

from .common import InfoExtractor
from ..utils import (
float_or_none,
format_field,
int_or_none,
parse_codecs,
parse_qs,
str_or_none,
traverse_obj,
)
class AcFunVideoBaseIE(InfoExtractor):
def _extract_metadata(self, video_id, video_info):
playjson = self._parse_json(video_info['ksPlayJson'], video_id)
formats, subtitles = [], {}
for video in traverse_obj(playjson, ('adaptationSet', 0, 'representation')):
fmts, subs = self._extract_m3u8_formats_and_subtitles(video['url'], video_id, 'mp4', fatal=False)
formats.extend(fmts)
self._merge_subtitles(subs, target=subtitles)
for f in fmts:
f.update({
'fps': float_or_none(video.get('frameRate')),
'width': int_or_none(video.get('width')),
'height': int_or_none(video.get('height')),
'tbr': float_or_none(video.get('avgBitrate')),
**parse_codecs(video.get('codecs', '')),
})
return {
'id': video_id,
'formats': formats,
'subtitles': subtitles,
'duration': float_or_none(video_info.get('durationMillis'), 1000),
'timestamp': int_or_none(video_info.get('uploadTime'), 1000),
'http_headers': {'Referer': 'https://www.acfun.cn/'},
}
class AcFunVideoIE(AcFunVideoBaseIE):
_VALID_URL = r'https?://www\.acfun\.cn/v/ac(?P<id>[_\d]+)'
_TESTS = [{
'url': 'https://www.acfun.cn/v/ac35457073',
'info_dict': {
'id': '35457073',
'ext': 'mp4',
'duration': 174.208,
'timestamp': 1656403967,
'title': '1 8 岁 现 状',
'description': '“赶紧回去!班主任查班了!”',
'uploader': '锤子game',
'uploader_id': '51246077',
'thumbnail': r're:^https?://.*\.(jpg|jpeg)',
'upload_date': '20220628',
'like_count': int,
'view_count': int,
'comment_count': int,
'tags': list,
},
}, {
# example for len(video_list) > 1
'url': 'https://www.acfun.cn/v/ac35468952_2',
'info_dict': {
'id': '35468952_2',
'ext': 'mp4',
'title': '【动画剧集】Rocket & Groot Season 12022/火箭浣熊与格鲁特第1季 P02 S01E02 十拿九穩',
'duration': 90.459,
'uploader': '比令',
'uploader_id': '37259967',
'upload_date': '20220629',
'timestamp': 1656479962,
'tags': list,
'like_count': int,
'view_count': int,
'comment_count': int,
'thumbnail': r're:^https?://.*\.(jpg|jpeg)',
'description': 'md5:67583aaf3a0f933bd606bc8a2d3ebb17',
},
}]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
json_all = self._search_json(r'window.videoInfo\s*=', webpage, 'videoInfo', video_id)
title = json_all.get('title')
video_list = json_all.get('videoList') or []
video_internal_id = traverse_obj(json_all, ('currentVideoInfo', 'id'))
if video_internal_id and len(video_list) > 1:
part_idx, part_video_info = next(
(idx + 1, v) for (idx, v) in enumerate(video_list)
if v['id'] == video_internal_id)
title = f'{title} P{part_idx:02d} {part_video_info["title"]}'
return {
**self._extract_metadata(video_id, json_all['currentVideoInfo']),
'title': title,
'thumbnail': json_all.get('coverUrl'),
'description': json_all.get('description'),
'uploader': traverse_obj(json_all, ('user', 'name')),
'uploader_id': traverse_obj(json_all, ('user', 'href')),
'tags': traverse_obj(json_all, ('tagList', ..., 'name')),
'view_count': int_or_none(json_all.get('viewCount')),
'like_count': int_or_none(json_all.get('likeCountShow')),
'comment_count': int_or_none(json_all.get('commentCountShow')),
}
class AcFunBangumiIE(AcFunVideoBaseIE):
_VALID_URL = r'https?://www\.acfun\.cn/bangumi/(?P<id>aa[_\d]+)'
_TESTS = [{
'url': 'https://www.acfun.cn/bangumi/aa6002917_36188_1745457?ac=2',
'info_dict': {
'id': 'aa6002917_36188_1745457__2',
'ext': 'mp4',
'title': '【7月】租借女友 水原千鹤角色曲『DATE』特别PV',
'upload_date': '20200916',
'timestamp': 1600243813,
'duration': 92.091,
},
}, {
'url': 'https://www.acfun.cn/bangumi/aa5023171_36188_1750645',
'info_dict': {
'id': 'aa5023171_36188_1750645',
'ext': 'mp4',
'title': '红孩儿之趴趴蛙寻石记 第5话 ',
'duration': 760.0,
'season': '红孩儿之趴趴蛙寻石记',
'season_id': '5023171',
'season_number': 1, # series has only 1 season
'episode': 'Episode 5',
'episode_number': 5,
'upload_date': '20181223',
'timestamp': 1545552185,
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)',
'comment_count': int,
},
}, {
'url': 'https://www.acfun.cn/bangumi/aa6065485_36188_1885061',
'info_dict': {
'id': 'aa6065485_36188_1885061',
'ext': 'mp4',
'title': '叽歪老表(第二季) 第5话 坚不可摧',
'season': '叽歪老表(第二季)',
'season_number': 2,
'season_id': '6065485',
'episode': '坚不可摧',
'episode_number': 5,
'upload_date': '20220324',
'timestamp': 1648082786,
'duration': 105.002,
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)',
'comment_count': int,
},
}]
def _real_extract(self, url):
video_id = self._match_id(url)
ac_idx = parse_qs(url).get('ac', [None])[-1]
2022-11-06 20:59:58 +00:00
video_id = f'{video_id}{format_field(ac_idx, None, "__%s")}'
webpage = self._download_webpage(url, video_id)
json_bangumi_data = self._search_json(r'window.bangumiData\s*=', webpage, 'bangumiData', video_id)
if ac_idx:
video_info = json_bangumi_data['hlVideoInfo']
return {
**self._extract_metadata(video_id, video_info),
'title': video_info.get('title'),
}
video_info = json_bangumi_data['currentVideoInfo']
season_id = json_bangumi_data.get('bangumiId')
season_number = season_id and next((
idx for idx, v in enumerate(json_bangumi_data.get('relatedBangumis') or [], 1)
if v.get('id') == season_id), 1)
json_bangumi_list = self._search_json(
r'window\.bangumiList\s*=', webpage, 'bangumiList', video_id, fatal=False)
video_internal_id = int_or_none(traverse_obj(json_bangumi_data, ('currentVideoInfo', 'id')))
episode_number = video_internal_id and next((
idx for idx, v in enumerate(json_bangumi_list.get('items') or [], 1)
if v.get('videoId') == video_internal_id), None)
return {
**self._extract_metadata(video_id, video_info),
'title': json_bangumi_data.get('showTitle'),
'thumbnail': json_bangumi_data.get('image'),
'season': json_bangumi_data.get('bangumiTitle'),
'season_id': str_or_none(season_id),
'season_number': season_number,
'episode': json_bangumi_data.get('title'),
'episode_number': episode_number,
'comment_count': int_or_none(json_bangumi_data.get('commentCount')),
}