mirror of https://github.com/yt-dlp/yt-dlp.git
[ie/bilibili] Support courses and interactive videos (#8343)
Closes #6135, Closes #8428 Authored by: c-basalt
This commit is contained in:
parent
f124fa4588
commit
9f09bdcfcb
|
@ -216,6 +216,8 @@ from .bilibili import (
|
||||||
BiliBiliBangumiIE,
|
BiliBiliBangumiIE,
|
||||||
BiliBiliBangumiSeasonIE,
|
BiliBiliBangumiSeasonIE,
|
||||||
BiliBiliBangumiMediaIE,
|
BiliBiliBangumiMediaIE,
|
||||||
|
BilibiliCheeseIE,
|
||||||
|
BilibiliCheeseSeasonIE,
|
||||||
BiliBiliSearchIE,
|
BiliBiliSearchIE,
|
||||||
BilibiliCategoryIE,
|
BilibiliCategoryIE,
|
||||||
BilibiliAudioIE,
|
BilibiliAudioIE,
|
||||||
|
|
|
@ -2,6 +2,7 @@ import base64
|
||||||
import functools
|
import functools
|
||||||
import hashlib
|
import hashlib
|
||||||
import itertools
|
import itertools
|
||||||
|
import json
|
||||||
import math
|
import math
|
||||||
import re
|
import re
|
||||||
import time
|
import time
|
||||||
|
@ -16,9 +17,11 @@ from ..utils import (
|
||||||
InAdvancePagedList,
|
InAdvancePagedList,
|
||||||
OnDemandPagedList,
|
OnDemandPagedList,
|
||||||
bool_or_none,
|
bool_or_none,
|
||||||
|
clean_html,
|
||||||
filter_dict,
|
filter_dict,
|
||||||
float_or_none,
|
float_or_none,
|
||||||
format_field,
|
format_field,
|
||||||
|
get_element_by_class,
|
||||||
int_or_none,
|
int_or_none,
|
||||||
join_nonempty,
|
join_nonempty,
|
||||||
make_archive_id,
|
make_archive_id,
|
||||||
|
@ -88,6 +91,12 @@ class BilibiliBaseIE(InfoExtractor):
|
||||||
|
|
||||||
return formats
|
return formats
|
||||||
|
|
||||||
|
def _download_playinfo(self, video_id, cid):
|
||||||
|
return self._download_json(
|
||||||
|
'https://api.bilibili.com/x/player/playurl', video_id,
|
||||||
|
query={'bvid': video_id, 'cid': cid, 'fnval': 4048},
|
||||||
|
note=f'Downloading video formats for cid {cid}')['data']
|
||||||
|
|
||||||
def json2srt(self, json_data):
|
def json2srt(self, json_data):
|
||||||
srt_data = ''
|
srt_data = ''
|
||||||
for idx, line in enumerate(json_data.get('body') or []):
|
for idx, line in enumerate(json_data.get('body') or []):
|
||||||
|
@ -96,7 +105,7 @@ class BilibiliBaseIE(InfoExtractor):
|
||||||
f'{line["content"]}\n\n')
|
f'{line["content"]}\n\n')
|
||||||
return srt_data
|
return srt_data
|
||||||
|
|
||||||
def _get_subtitles(self, video_id, aid, cid):
|
def _get_subtitles(self, video_id, cid, aid=None):
|
||||||
subtitles = {
|
subtitles = {
|
||||||
'danmaku': [{
|
'danmaku': [{
|
||||||
'ext': 'xml',
|
'ext': 'xml',
|
||||||
|
@ -104,8 +113,15 @@ class BilibiliBaseIE(InfoExtractor):
|
||||||
}]
|
}]
|
||||||
}
|
}
|
||||||
|
|
||||||
video_info_json = self._download_json(f'https://api.bilibili.com/x/player/v2?aid={aid}&cid={cid}', video_id)
|
subtitle_info = traverse_obj(self._download_json(
|
||||||
for s in traverse_obj(video_info_json, ('data', 'subtitle', 'subtitles', ...)):
|
'https://api.bilibili.com/x/player/v2', video_id,
|
||||||
|
query={'aid': aid, 'cid': cid} if aid else {'bvid': video_id, 'cid': cid},
|
||||||
|
note=f'Extracting subtitle info {cid}'), ('data', 'subtitle'))
|
||||||
|
subs_list = traverse_obj(subtitle_info, ('subtitles', lambda _, v: v['subtitle_url'] and v['lan']))
|
||||||
|
if not subs_list and traverse_obj(subtitle_info, 'allow_submit'):
|
||||||
|
if not self._get_cookies('https://api.bilibili.com').get('SESSDATA'): # no login session cookie
|
||||||
|
self.report_warning(f'CC subtitles (if any) are only visible when logged in. {self._login_hint()}', only_once=True)
|
||||||
|
for s in subs_list:
|
||||||
subtitles.setdefault(s['lan'], []).append({
|
subtitles.setdefault(s['lan'], []).append({
|
||||||
'ext': 'srt',
|
'ext': 'srt',
|
||||||
'data': self.json2srt(self._download_json(s['subtitle_url'], video_id))
|
'data': self.json2srt(self._download_json(s['subtitle_url'], video_id))
|
||||||
|
@ -155,7 +171,54 @@ class BilibiliBaseIE(InfoExtractor):
|
||||||
for entry in traverse_obj(season_info, (
|
for entry in traverse_obj(season_info, (
|
||||||
'result', 'main_section', 'episodes',
|
'result', 'main_section', 'episodes',
|
||||||
lambda _, v: url_or_none(v['share_url']) and v['id'])):
|
lambda _, v: url_or_none(v['share_url']) and v['id'])):
|
||||||
yield self.url_result(entry['share_url'], BiliBiliBangumiIE, f'ep{entry["id"]}')
|
yield self.url_result(entry['share_url'], BiliBiliBangumiIE, str_or_none(entry.get('id')))
|
||||||
|
|
||||||
|
def _get_divisions(self, video_id, graph_version, edges, edge_id, cid_edges=None):
|
||||||
|
cid_edges = cid_edges or {}
|
||||||
|
division_data = self._download_json(
|
||||||
|
'https://api.bilibili.com/x/stein/edgeinfo_v2', video_id,
|
||||||
|
query={'graph_version': graph_version, 'edge_id': edge_id, 'bvid': video_id},
|
||||||
|
note=f'Extracting divisions from edge {edge_id}')
|
||||||
|
edges.setdefault(edge_id, {}).update(
|
||||||
|
traverse_obj(division_data, ('data', 'story_list', lambda _, v: v['edge_id'] == edge_id, {
|
||||||
|
'title': ('title', {str}),
|
||||||
|
'cid': ('cid', {int_or_none}),
|
||||||
|
}), get_all=False))
|
||||||
|
|
||||||
|
edges[edge_id].update(traverse_obj(division_data, ('data', {
|
||||||
|
'title': ('title', {str}),
|
||||||
|
'choices': ('edges', 'questions', ..., 'choices', ..., {
|
||||||
|
'edge_id': ('id', {int_or_none}),
|
||||||
|
'cid': ('cid', {int_or_none}),
|
||||||
|
'text': ('option', {str}),
|
||||||
|
}),
|
||||||
|
})))
|
||||||
|
# use dict to combine edges that use the same video section (same cid)
|
||||||
|
cid_edges.setdefault(edges[edge_id]['cid'], {})[edge_id] = edges[edge_id]
|
||||||
|
for choice in traverse_obj(edges, (edge_id, 'choices', ...)):
|
||||||
|
if choice['edge_id'] not in edges:
|
||||||
|
edges[choice['edge_id']] = {'cid': choice['cid']}
|
||||||
|
self._get_divisions(video_id, graph_version, edges, choice['edge_id'], cid_edges=cid_edges)
|
||||||
|
return cid_edges
|
||||||
|
|
||||||
|
def _get_interactive_entries(self, video_id, cid, metainfo):
|
||||||
|
graph_version = traverse_obj(
|
||||||
|
self._download_json(
|
||||||
|
'https://api.bilibili.com/x/player/wbi/v2', video_id,
|
||||||
|
'Extracting graph version', query={'bvid': video_id, 'cid': cid}),
|
||||||
|
('data', 'interaction', 'graph_version', {int_or_none}))
|
||||||
|
cid_edges = self._get_divisions(video_id, graph_version, {1: {'cid': cid}}, 1)
|
||||||
|
for cid, edges in cid_edges.items():
|
||||||
|
play_info = self._download_playinfo(video_id, cid)
|
||||||
|
yield {
|
||||||
|
**metainfo,
|
||||||
|
'id': f'{video_id}_{cid}',
|
||||||
|
'title': f'{metainfo.get("title")} - {list(edges.values())[0].get("title")}',
|
||||||
|
'formats': self.extract_formats(play_info),
|
||||||
|
'description': f'{json.dumps(edges, ensure_ascii=False)}\n{metainfo.get("description", "")}',
|
||||||
|
'duration': float_or_none(play_info.get('timelength'), scale=1000),
|
||||||
|
'subtitles': self.extract_subtitles(video_id, cid),
|
||||||
|
}
|
||||||
|
|
||||||
|
|
||||||
class BiliBiliIE(BilibiliBaseIE):
|
class BiliBiliIE(BilibiliBaseIE):
|
||||||
|
@ -180,7 +243,7 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
'view_count': int,
|
'view_count': int,
|
||||||
},
|
},
|
||||||
}, {
|
}, {
|
||||||
# old av URL version
|
'note': 'old av URL version',
|
||||||
'url': 'http://www.bilibili.com/video/av1074402/',
|
'url': 'http://www.bilibili.com/video/av1074402/',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'thumbnail': r're:^https?://.*\.(jpg|jpeg)$',
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg)$',
|
||||||
|
@ -212,7 +275,7 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
'id': 'BV1bK411W797_p1',
|
'id': 'BV1bK411W797_p1',
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'title': '物语中的人物是如何吐槽自己的OP的 p01 Staple Stable/战场原+羽川',
|
'title': '物语中的人物是如何吐槽自己的OP的 p01 Staple Stable/战场原+羽川',
|
||||||
'tags': 'count:11',
|
'tags': 'count:10',
|
||||||
'timestamp': 1589601697,
|
'timestamp': 1589601697,
|
||||||
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
'uploader': '打牌还是打桩',
|
'uploader': '打牌还是打桩',
|
||||||
|
@ -232,7 +295,7 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
'id': 'BV1bK411W797_p1',
|
'id': 'BV1bK411W797_p1',
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'title': '物语中的人物是如何吐槽自己的OP的 p01 Staple Stable/战场原+羽川',
|
'title': '物语中的人物是如何吐槽自己的OP的 p01 Staple Stable/战场原+羽川',
|
||||||
'tags': 'count:11',
|
'tags': 'count:10',
|
||||||
'timestamp': 1589601697,
|
'timestamp': 1589601697,
|
||||||
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
'uploader': '打牌还是打桩',
|
'uploader': '打牌还是打桩',
|
||||||
|
@ -343,18 +406,120 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
},
|
},
|
||||||
'params': {'skip_download': True},
|
'params': {'skip_download': True},
|
||||||
|
}, {
|
||||||
|
'note': 'interactive/split-path video',
|
||||||
|
'url': 'https://www.bilibili.com/video/BV1af4y1H7ga/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'BV1af4y1H7ga',
|
||||||
|
'title': '【互动游戏】花了大半年时间做的自我介绍~请查收!!',
|
||||||
|
'timestamp': 1630500414,
|
||||||
|
'upload_date': '20210901',
|
||||||
|
'description': 'md5:01113e39ab06e28042d74ac356a08786',
|
||||||
|
'tags': list,
|
||||||
|
'uploader': '钉宫妮妮Ninico',
|
||||||
|
'duration': 1503,
|
||||||
|
'uploader_id': '8881297',
|
||||||
|
'comment_count': int,
|
||||||
|
'view_count': int,
|
||||||
|
'like_count': int,
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
|
},
|
||||||
|
'playlist_count': 33,
|
||||||
|
'playlist': [{
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'BV1af4y1H7ga_400950101',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'title': '【互动游戏】花了大半年时间做的自我介绍~请查收!! - 听见猫猫叫~',
|
||||||
|
'timestamp': 1630500414,
|
||||||
|
'upload_date': '20210901',
|
||||||
|
'description': 'md5:db66ac7a2813a94b8291dbce990cc5b2',
|
||||||
|
'tags': list,
|
||||||
|
'uploader': '钉宫妮妮Ninico',
|
||||||
|
'duration': 11.605,
|
||||||
|
'uploader_id': '8881297',
|
||||||
|
'comment_count': int,
|
||||||
|
'view_count': int,
|
||||||
|
'like_count': int,
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
|
},
|
||||||
|
}],
|
||||||
|
}, {
|
||||||
|
'note': '301 redirect to bangumi link',
|
||||||
|
'url': 'https://www.bilibili.com/video/BV1TE411f7f1',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '288525',
|
||||||
|
'title': '李永乐老师 钱学森弹道和乘波体飞行器是什么?',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'series': '我和我的祖国',
|
||||||
|
'series_id': '4780',
|
||||||
|
'season': '幕后纪实',
|
||||||
|
'season_id': '28609',
|
||||||
|
'season_number': 1,
|
||||||
|
'episode': '钱学森弹道和乘波体飞行器是什么?',
|
||||||
|
'episode_id': '288525',
|
||||||
|
'episode_number': 105,
|
||||||
|
'duration': 1183.957,
|
||||||
|
'timestamp': 1571648124,
|
||||||
|
'upload_date': '20191021',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
|
},
|
||||||
|
}, {
|
||||||
|
'url': 'https://www.bilibili.com/video/BV1jL41167ZG/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'BV1jL41167ZG',
|
||||||
|
'title': '一场大火引发的离奇死亡!古典推理经典短篇集《不可能犯罪诊断书》!',
|
||||||
|
'ext': 'mp4',
|
||||||
|
},
|
||||||
|
'skip': 'supporter-only video',
|
||||||
|
}, {
|
||||||
|
'url': 'https://www.bilibili.com/video/BV1Ks411f7aQ/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'BV1Ks411f7aQ',
|
||||||
|
'title': '【BD1080P】狼与香辛料I【华盟】',
|
||||||
|
'ext': 'mp4',
|
||||||
|
},
|
||||||
|
'skip': 'login required',
|
||||||
|
}, {
|
||||||
|
'url': 'https://www.bilibili.com/video/BV1GJ411x7h7/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'BV1GJ411x7h7',
|
||||||
|
'title': '【官方 MV】Never Gonna Give You Up - Rick Astley',
|
||||||
|
'ext': 'mp4',
|
||||||
|
},
|
||||||
|
'skip': 'geo-restricted',
|
||||||
}]
|
}]
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
video_id = self._match_id(url)
|
||||||
webpage = self._download_webpage(url, video_id)
|
webpage, urlh = self._download_webpage_handle(url, video_id)
|
||||||
|
if not self._match_valid_url(urlh.url):
|
||||||
|
return self.url_result(urlh.url)
|
||||||
|
|
||||||
initial_state = self._search_json(r'window\.__INITIAL_STATE__\s*=', webpage, 'initial state', video_id)
|
initial_state = self._search_json(r'window\.__INITIAL_STATE__\s*=', webpage, 'initial state', video_id)
|
||||||
|
|
||||||
is_festival = 'videoData' not in initial_state
|
is_festival = 'videoData' not in initial_state
|
||||||
if is_festival:
|
if is_festival:
|
||||||
video_data = initial_state['videoInfo']
|
video_data = initial_state['videoInfo']
|
||||||
else:
|
else:
|
||||||
play_info = self._search_json(r'window\.__playinfo__\s*=', webpage, 'play info', video_id)['data']
|
play_info_obj = self._search_json(
|
||||||
|
r'window\.__playinfo__\s*=', webpage, 'play info', video_id, fatal=False)
|
||||||
|
if not play_info_obj:
|
||||||
|
if traverse_obj(initial_state, ('error', 'trueCode')) == -403:
|
||||||
|
self.raise_login_required()
|
||||||
|
if traverse_obj(initial_state, ('error', 'trueCode')) == -404:
|
||||||
|
raise ExtractorError(
|
||||||
|
'This video may be deleted or geo-restricted. '
|
||||||
|
'You might want to try a VPN or a proxy server (with --proxy)', expected=True)
|
||||||
|
play_info = traverse_obj(play_info_obj, ('data', {dict}))
|
||||||
|
if not play_info:
|
||||||
|
if traverse_obj(play_info_obj, 'code') == 87007:
|
||||||
|
toast = get_element_by_class('tips-toast', webpage) or ''
|
||||||
|
msg = clean_html(
|
||||||
|
f'{get_element_by_class("belongs-to", toast) or ""},'
|
||||||
|
+ (get_element_by_class('level', toast) or ''))
|
||||||
|
raise ExtractorError(
|
||||||
|
f'This is a supporter-only video: {msg}. {self._login_hint()}', expected=True)
|
||||||
|
raise ExtractorError('Failed to extract play info')
|
||||||
video_data = initial_state['videoData']
|
video_data = initial_state['videoData']
|
||||||
|
|
||||||
video_id, title = video_data['bvid'], video_data.get('title')
|
video_id, title = video_data['bvid'], video_data.get('title')
|
||||||
|
@ -385,10 +550,7 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
|
|
||||||
festival_info = {}
|
festival_info = {}
|
||||||
if is_festival:
|
if is_festival:
|
||||||
play_info = self._download_json(
|
play_info = self._download_playinfo(video_id, cid)
|
||||||
'https://api.bilibili.com/x/player/playurl', video_id,
|
|
||||||
query={'bvid': video_id, 'cid': cid, 'fnval': 4048},
|
|
||||||
note='Extracting festival video formats')['data']
|
|
||||||
|
|
||||||
festival_info = traverse_obj(initial_state, {
|
festival_info = traverse_obj(initial_state, {
|
||||||
'uploader': ('videoInfo', 'upName'),
|
'uploader': ('videoInfo', 'upName'),
|
||||||
|
@ -397,7 +559,7 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
'thumbnail': ('sectionEpisodes', lambda _, v: v['bvid'] == video_id, 'cover'),
|
'thumbnail': ('sectionEpisodes', lambda _, v: v['bvid'] == video_id, 'cover'),
|
||||||
}, get_all=False)
|
}, get_all=False)
|
||||||
|
|
||||||
return {
|
metainfo = {
|
||||||
**traverse_obj(initial_state, {
|
**traverse_obj(initial_state, {
|
||||||
'uploader': ('upData', 'name'),
|
'uploader': ('upData', 'name'),
|
||||||
'uploader_id': ('upData', 'mid', {str_or_none}),
|
'uploader_id': ('upData', 'mid', {str_or_none}),
|
||||||
|
@ -413,28 +575,59 @@ class BiliBiliIE(BilibiliBaseIE):
|
||||||
'comment_count': ('stat', 'reply', {int_or_none}),
|
'comment_count': ('stat', 'reply', {int_or_none}),
|
||||||
}, get_all=False),
|
}, get_all=False),
|
||||||
'id': f'{video_id}{format_field(part_id, None, "_p%d")}',
|
'id': f'{video_id}{format_field(part_id, None, "_p%d")}',
|
||||||
'formats': self.extract_formats(play_info),
|
|
||||||
'_old_archive_ids': [make_archive_id(self, old_video_id)] if old_video_id else None,
|
'_old_archive_ids': [make_archive_id(self, old_video_id)] if old_video_id else None,
|
||||||
'title': title,
|
'title': title,
|
||||||
|
'http_headers': {'Referer': url},
|
||||||
|
}
|
||||||
|
|
||||||
|
is_interactive = traverse_obj(video_data, ('rights', 'is_stein_gate'))
|
||||||
|
if is_interactive:
|
||||||
|
return self.playlist_result(
|
||||||
|
self._get_interactive_entries(video_id, cid, metainfo), **metainfo, **{
|
||||||
|
'duration': traverse_obj(initial_state, ('videoData', 'duration', {int_or_none})),
|
||||||
|
'__post_extractor': self.extract_comments(aid),
|
||||||
|
})
|
||||||
|
else:
|
||||||
|
return {
|
||||||
|
**metainfo,
|
||||||
'duration': float_or_none(play_info.get('timelength'), scale=1000),
|
'duration': float_or_none(play_info.get('timelength'), scale=1000),
|
||||||
'chapters': self._get_chapters(aid, cid),
|
'chapters': self._get_chapters(aid, cid),
|
||||||
'subtitles': self.extract_subtitles(video_id, aid, cid),
|
'subtitles': self.extract_subtitles(video_id, cid),
|
||||||
|
'formats': self.extract_formats(play_info),
|
||||||
'__post_extractor': self.extract_comments(aid),
|
'__post_extractor': self.extract_comments(aid),
|
||||||
'http_headers': {'Referer': url},
|
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
||||||
class BiliBiliBangumiIE(BilibiliBaseIE):
|
class BiliBiliBangumiIE(BilibiliBaseIE):
|
||||||
_VALID_URL = r'https?://(?:www\.)?bilibili\.com/bangumi/play/(?P<id>ep\d+)'
|
_VALID_URL = r'https?://(?:www\.)?bilibili\.com/bangumi/play/ep(?P<id>\d+)'
|
||||||
|
|
||||||
_TESTS = [{
|
_TESTS = [{
|
||||||
|
'url': 'https://www.bilibili.com/bangumi/play/ep21495/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '21495',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'series': '悠久之翼',
|
||||||
|
'series_id': '774',
|
||||||
|
'season': '第二季',
|
||||||
|
'season_id': '1182',
|
||||||
|
'season_number': 2,
|
||||||
|
'episode': 'forever/ef',
|
||||||
|
'episode_id': '21495',
|
||||||
|
'episode_number': 12,
|
||||||
|
'title': '12 forever/ef',
|
||||||
|
'duration': 1420.791,
|
||||||
|
'timestamp': 1320412200,
|
||||||
|
'upload_date': '20111104',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
|
},
|
||||||
|
}, {
|
||||||
'url': 'https://www.bilibili.com/bangumi/play/ep267851',
|
'url': 'https://www.bilibili.com/bangumi/play/ep267851',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': '267851',
|
'id': '267851',
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'series': '鬼灭之刃',
|
'series': '鬼灭之刃',
|
||||||
'series_id': '4358',
|
'series_id': '4358',
|
||||||
'season': '鬼灭之刃',
|
'season': '立志篇',
|
||||||
'season_id': '26801',
|
'season_id': '26801',
|
||||||
'season_number': 1,
|
'season_number': 1,
|
||||||
'episode': '残酷',
|
'episode': '残酷',
|
||||||
|
@ -446,13 +639,32 @@ class BiliBiliBangumiIE(BilibiliBaseIE):
|
||||||
'upload_date': '20190406',
|
'upload_date': '20190406',
|
||||||
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$'
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$'
|
||||||
},
|
},
|
||||||
'skip': 'According to the copyright owner\'s request, you may only watch the video after you are premium member.'
|
'skip': 'Geo-restricted',
|
||||||
|
}, {
|
||||||
|
'note': 'a making-of which falls outside main section',
|
||||||
|
'url': 'https://www.bilibili.com/bangumi/play/ep345120',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '345120',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'series': '鬼灭之刃',
|
||||||
|
'series_id': '4358',
|
||||||
|
'season': '立志篇',
|
||||||
|
'season_id': '26801',
|
||||||
|
'season_number': 1,
|
||||||
|
'episode': '炭治郎篇',
|
||||||
|
'episode_id': '345120',
|
||||||
|
'episode_number': 27,
|
||||||
|
'title': '#1 炭治郎篇',
|
||||||
|
'duration': 1922.129,
|
||||||
|
'timestamp': 1602853860,
|
||||||
|
'upload_date': '20201016',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$'
|
||||||
|
},
|
||||||
}]
|
}]
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
episode_id = self._match_id(url)
|
||||||
episode_id = video_id[2:]
|
webpage = self._download_webpage(url, episode_id)
|
||||||
webpage = self._download_webpage(url, video_id)
|
|
||||||
|
|
||||||
if '您所在的地区无法观看本片' in webpage:
|
if '您所在的地区无法观看本片' in webpage:
|
||||||
raise GeoRestrictedError('This video is restricted')
|
raise GeoRestrictedError('This video is restricted')
|
||||||
|
@ -461,7 +673,7 @@ class BiliBiliBangumiIE(BilibiliBaseIE):
|
||||||
|
|
||||||
headers = {'Referer': url, **self.geo_verification_headers()}
|
headers = {'Referer': url, **self.geo_verification_headers()}
|
||||||
play_info = self._download_json(
|
play_info = self._download_json(
|
||||||
'https://api.bilibili.com/pgc/player/web/v2/playurl', video_id,
|
'https://api.bilibili.com/pgc/player/web/v2/playurl', episode_id,
|
||||||
'Extracting episode', query={'fnval': '4048', 'ep_id': episode_id},
|
'Extracting episode', query={'fnval': '4048', 'ep_id': episode_id},
|
||||||
headers=headers)
|
headers=headers)
|
||||||
premium_only = play_info.get('code') == -10403
|
premium_only = play_info.get('code') == -10403
|
||||||
|
@ -472,40 +684,43 @@ class BiliBiliBangumiIE(BilibiliBaseIE):
|
||||||
self.raise_login_required('This video is for premium members only')
|
self.raise_login_required('This video is for premium members only')
|
||||||
|
|
||||||
bangumi_info = self._download_json(
|
bangumi_info = self._download_json(
|
||||||
'https://api.bilibili.com/pgc/view/web/season', video_id, 'Get episode details',
|
'https://api.bilibili.com/pgc/view/web/season', episode_id, 'Get episode details',
|
||||||
query={'ep_id': episode_id}, headers=headers)['result']
|
query={'ep_id': episode_id}, headers=headers)['result']
|
||||||
|
|
||||||
episode_number, episode_info = next((
|
episode_number, episode_info = next((
|
||||||
(idx, ep) for idx, ep in enumerate(traverse_obj(
|
(idx, ep) for idx, ep in enumerate(traverse_obj(
|
||||||
bangumi_info, ('episodes', ..., {dict})), 1)
|
bangumi_info, (('episodes', ('section', ..., 'episodes')), ..., {dict})), 1)
|
||||||
if str_or_none(ep.get('id')) == episode_id), (1, {}))
|
if str_or_none(ep.get('id')) == episode_id), (1, {}))
|
||||||
|
|
||||||
season_id = bangumi_info.get('season_id')
|
season_id = bangumi_info.get('season_id')
|
||||||
season_number = season_id and next((
|
season_number, season_title = season_id and next((
|
||||||
idx + 1 for idx, e in enumerate(
|
(idx + 1, e.get('season_title')) for idx, e in enumerate(
|
||||||
traverse_obj(bangumi_info, ('seasons', ...)))
|
traverse_obj(bangumi_info, ('seasons', ...)))
|
||||||
if e.get('season_id') == season_id
|
if e.get('season_id') == season_id
|
||||||
), None)
|
), (None, None))
|
||||||
|
|
||||||
aid = episode_info.get('aid')
|
aid = episode_info.get('aid')
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'id': video_id,
|
'id': episode_id,
|
||||||
'formats': formats,
|
'formats': formats,
|
||||||
**traverse_obj(bangumi_info, {
|
**traverse_obj(bangumi_info, {
|
||||||
'series': ('series', 'series_title', {str}),
|
'series': ('series', 'series_title', {str}),
|
||||||
'series_id': ('series', 'series_id', {str_or_none}),
|
'series_id': ('series', 'series_id', {str_or_none}),
|
||||||
'thumbnail': ('square_cover', {url_or_none}),
|
'thumbnail': ('square_cover', {url_or_none}),
|
||||||
}),
|
}),
|
||||||
'title': join_nonempty('title', 'long_title', delim=' ', from_dict=episode_info),
|
**traverse_obj(episode_info, {
|
||||||
'episode': episode_info.get('long_title'),
|
'episode': ('long_title', {str}),
|
||||||
|
'episode_number': ('title', {int_or_none}, {lambda x: x or episode_number}),
|
||||||
|
'timestamp': ('pub_time', {int_or_none}),
|
||||||
|
'title': {lambda v: v and join_nonempty('title', 'long_title', delim=' ', from_dict=v)},
|
||||||
|
}),
|
||||||
'episode_id': episode_id,
|
'episode_id': episode_id,
|
||||||
'episode_number': int_or_none(episode_info.get('title')) or episode_number,
|
'season': str_or_none(season_title),
|
||||||
'season_id': str_or_none(season_id),
|
'season_id': str_or_none(season_id),
|
||||||
'season_number': season_number,
|
'season_number': season_number,
|
||||||
'timestamp': int_or_none(episode_info.get('pub_time')),
|
|
||||||
'duration': float_or_none(play_info.get('timelength'), scale=1000),
|
'duration': float_or_none(play_info.get('timelength'), scale=1000),
|
||||||
'subtitles': self.extract_subtitles(video_id, aid, episode_info.get('cid')),
|
'subtitles': self.extract_subtitles(episode_id, episode_info.get('cid'), aid=aid),
|
||||||
'__post_extractor': self.extract_comments(aid),
|
'__post_extractor': self.extract_comments(aid),
|
||||||
'http_headers': headers,
|
'http_headers': headers,
|
||||||
}
|
}
|
||||||
|
@ -517,17 +732,53 @@ class BiliBiliBangumiMediaIE(BilibiliBaseIE):
|
||||||
'url': 'https://www.bilibili.com/bangumi/media/md24097891',
|
'url': 'https://www.bilibili.com/bangumi/media/md24097891',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': '24097891',
|
'id': '24097891',
|
||||||
|
'title': 'CAROLE & TUESDAY',
|
||||||
|
'description': 'md5:42417ad33d1eaa1c93bfd2dd1626b829',
|
||||||
},
|
},
|
||||||
'playlist_mincount': 25,
|
'playlist_mincount': 25,
|
||||||
|
}, {
|
||||||
|
'url': 'https://www.bilibili.com/bangumi/media/md1565/',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '1565',
|
||||||
|
'title': '攻壳机动队 S.A.C. 2nd GIG',
|
||||||
|
'description': 'md5:46cac00bafd645b97f4d6df616fc576d',
|
||||||
|
},
|
||||||
|
'playlist_count': 26,
|
||||||
|
'playlist': [{
|
||||||
|
'info_dict': {
|
||||||
|
'id': '68540',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'series': '攻壳机动队',
|
||||||
|
'series_id': '1077',
|
||||||
|
'season': '第二季',
|
||||||
|
'season_id': '1565',
|
||||||
|
'season_number': 2,
|
||||||
|
'episode': '再启动 REEMBODY',
|
||||||
|
'episode_id': '68540',
|
||||||
|
'episode_number': 1,
|
||||||
|
'title': '1 再启动 REEMBODY',
|
||||||
|
'duration': 1525.777,
|
||||||
|
'timestamp': 1425074413,
|
||||||
|
'upload_date': '20150227',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$'
|
||||||
|
},
|
||||||
|
}],
|
||||||
}]
|
}]
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
media_id = self._match_id(url)
|
media_id = self._match_id(url)
|
||||||
webpage = self._download_webpage(url, media_id)
|
webpage = self._download_webpage(url, media_id)
|
||||||
ss_id = self._search_json(
|
|
||||||
r'window\.__INITIAL_STATE__\s*=', webpage, 'initial_state', media_id)['mediaInfo']['season_id']
|
|
||||||
|
|
||||||
return self.playlist_result(self._get_episodes_from_season(ss_id, url), media_id)
|
initial_state = self._search_json(
|
||||||
|
r'window\.__INITIAL_STATE__\s*=', webpage, 'initial_state', media_id)
|
||||||
|
ss_id = initial_state['mediaInfo']['season_id']
|
||||||
|
|
||||||
|
return self.playlist_result(
|
||||||
|
self._get_episodes_from_season(ss_id, url), media_id,
|
||||||
|
**traverse_obj(initial_state, ('mediaInfo', {
|
||||||
|
'title': ('title', {str}),
|
||||||
|
'description': ('evaluate', {str}),
|
||||||
|
})))
|
||||||
|
|
||||||
|
|
||||||
class BiliBiliBangumiSeasonIE(BilibiliBaseIE):
|
class BiliBiliBangumiSeasonIE(BilibiliBaseIE):
|
||||||
|
@ -535,15 +786,183 @@ class BiliBiliBangumiSeasonIE(BilibiliBaseIE):
|
||||||
_TESTS = [{
|
_TESTS = [{
|
||||||
'url': 'https://www.bilibili.com/bangumi/play/ss26801',
|
'url': 'https://www.bilibili.com/bangumi/play/ss26801',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': '26801'
|
'id': '26801',
|
||||||
|
'title': '鬼灭之刃',
|
||||||
|
'description': 'md5:e2cc9848b6f69be6db79fc2a82d9661b',
|
||||||
},
|
},
|
||||||
'playlist_mincount': 26
|
'playlist_mincount': 26
|
||||||
|
}, {
|
||||||
|
'url': 'https://www.bilibili.com/bangumi/play/ss2251',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '2251',
|
||||||
|
'title': '玲音',
|
||||||
|
'description': 'md5:1fd40e3df4c08d4d9d89a6a34844bdc4',
|
||||||
|
},
|
||||||
|
'playlist_count': 13,
|
||||||
|
'playlist': [{
|
||||||
|
'info_dict': {
|
||||||
|
'id': '50188',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'series': '玲音',
|
||||||
|
'series_id': '1526',
|
||||||
|
'season': 'TV',
|
||||||
|
'season_id': '2251',
|
||||||
|
'season_number': 1,
|
||||||
|
'episode': 'WEIRD',
|
||||||
|
'episode_id': '50188',
|
||||||
|
'episode_number': 1,
|
||||||
|
'title': '1 WEIRD',
|
||||||
|
'duration': 1436.992,
|
||||||
|
'timestamp': 1343185080,
|
||||||
|
'upload_date': '20120725',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$'
|
||||||
|
},
|
||||||
|
}],
|
||||||
}]
|
}]
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
ss_id = self._match_id(url)
|
ss_id = self._match_id(url)
|
||||||
|
webpage = self._download_webpage(url, ss_id)
|
||||||
|
metainfo = traverse_obj(
|
||||||
|
self._search_json(r'<script[^>]+type="application/ld\+json"[^>]*>', webpage, 'info', ss_id),
|
||||||
|
('itemListElement', ..., {
|
||||||
|
'title': ('name', {str}),
|
||||||
|
'description': ('description', {str}),
|
||||||
|
}), get_all=False)
|
||||||
|
|
||||||
return self.playlist_result(self._get_episodes_from_season(ss_id, url), ss_id)
|
return self.playlist_result(self._get_episodes_from_season(ss_id, url), ss_id, **metainfo)
|
||||||
|
|
||||||
|
|
||||||
|
class BilibiliCheeseBaseIE(BilibiliBaseIE):
|
||||||
|
_HEADERS = {'Referer': 'https://www.bilibili.com/'}
|
||||||
|
|
||||||
|
def _extract_episode(self, season_info, ep_id):
|
||||||
|
episode_info = traverse_obj(season_info, (
|
||||||
|
'episodes', lambda _, v: v['id'] == int(ep_id)), get_all=False)
|
||||||
|
aid, cid = episode_info['aid'], episode_info['cid']
|
||||||
|
|
||||||
|
if traverse_obj(episode_info, 'ep_status') == -1:
|
||||||
|
raise ExtractorError('This course episode is not yet available.', expected=True)
|
||||||
|
if not traverse_obj(episode_info, 'playable'):
|
||||||
|
self.raise_login_required('You need to purchase the course to download this episode')
|
||||||
|
|
||||||
|
play_info = self._download_json(
|
||||||
|
'https://api.bilibili.com/pugv/player/web/playurl', ep_id,
|
||||||
|
query={'avid': aid, 'cid': cid, 'ep_id': ep_id, 'fnval': 16, 'fourk': 1},
|
||||||
|
headers=self._HEADERS, note='Downloading playinfo')['data']
|
||||||
|
|
||||||
|
return {
|
||||||
|
'id': str_or_none(ep_id),
|
||||||
|
'episode_id': str_or_none(ep_id),
|
||||||
|
'formats': self.extract_formats(play_info),
|
||||||
|
'extractor_key': BilibiliCheeseIE.ie_key(),
|
||||||
|
'extractor': BilibiliCheeseIE.IE_NAME,
|
||||||
|
'webpage_url': f'https://www.bilibili.com/cheese/play/ep{ep_id}',
|
||||||
|
**traverse_obj(episode_info, {
|
||||||
|
'episode': ('title', {str}),
|
||||||
|
'title': {lambda v: v and join_nonempty('index', 'title', delim=' - ', from_dict=v)},
|
||||||
|
'alt_title': ('subtitle', {str}),
|
||||||
|
'duration': ('duration', {int_or_none}),
|
||||||
|
'episode_number': ('index', {int_or_none}),
|
||||||
|
'thumbnail': ('cover', {url_or_none}),
|
||||||
|
'timestamp': ('release_date', {int_or_none}),
|
||||||
|
'view_count': ('play', {int_or_none}),
|
||||||
|
}),
|
||||||
|
**traverse_obj(season_info, {
|
||||||
|
'uploader': ('up_info', 'uname', {str}),
|
||||||
|
'uploader_id': ('up_info', 'mid', {str_or_none}),
|
||||||
|
}),
|
||||||
|
'subtitles': self.extract_subtitles(ep_id, cid, aid=aid),
|
||||||
|
'__post_extractor': self.extract_comments(aid),
|
||||||
|
'http_headers': self._HEADERS,
|
||||||
|
}
|
||||||
|
|
||||||
|
def _download_season_info(self, query_key, video_id):
|
||||||
|
return self._download_json(
|
||||||
|
f'https://api.bilibili.com/pugv/view/web/season?{query_key}={video_id}', video_id,
|
||||||
|
headers=self._HEADERS, note='Downloading season info')['data']
|
||||||
|
|
||||||
|
|
||||||
|
class BilibiliCheeseIE(BilibiliCheeseBaseIE):
|
||||||
|
_VALID_URL = r'https?://(?:www\.)?bilibili\.com/cheese/play/ep(?P<id>\d+)'
|
||||||
|
_TESTS = [{
|
||||||
|
'url': 'https://www.bilibili.com/cheese/play/ep229832',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '229832',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'title': '1 - 课程先导片',
|
||||||
|
'alt_title': '视频课 · 3分41秒',
|
||||||
|
'uploader': '马督工',
|
||||||
|
'uploader_id': '316568752',
|
||||||
|
'episode': '课程先导片',
|
||||||
|
'episode_id': '229832',
|
||||||
|
'episode_number': 1,
|
||||||
|
'duration': 221,
|
||||||
|
'timestamp': 1695549606,
|
||||||
|
'upload_date': '20230924',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
|
'view_count': int,
|
||||||
|
}
|
||||||
|
}]
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
ep_id = self._match_id(url)
|
||||||
|
return self._extract_episode(self._download_season_info('ep_id', ep_id), ep_id)
|
||||||
|
|
||||||
|
|
||||||
|
class BilibiliCheeseSeasonIE(BilibiliCheeseBaseIE):
|
||||||
|
_VALID_URL = r'https?://(?:www\.)?bilibili\.com/cheese/play/ss(?P<id>\d+)'
|
||||||
|
_TESTS = [{
|
||||||
|
'url': 'https://www.bilibili.com/cheese/play/ss5918',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '5918',
|
||||||
|
'title': '【限时五折】新闻系学不到:马督工教你做自媒体',
|
||||||
|
'description': '帮普通人建立世界模型,降低人与人的沟通门槛',
|
||||||
|
},
|
||||||
|
'playlist': [{
|
||||||
|
'info_dict': {
|
||||||
|
'id': '229832',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'title': '1 - 课程先导片',
|
||||||
|
'alt_title': '视频课 · 3分41秒',
|
||||||
|
'uploader': '马督工',
|
||||||
|
'uploader_id': '316568752',
|
||||||
|
'episode': '课程先导片',
|
||||||
|
'episode_id': '229832',
|
||||||
|
'episode_number': 1,
|
||||||
|
'duration': 221,
|
||||||
|
'timestamp': 1695549606,
|
||||||
|
'upload_date': '20230924',
|
||||||
|
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)$',
|
||||||
|
'view_count': int,
|
||||||
|
}
|
||||||
|
}],
|
||||||
|
'params': {'playlist_items': '1'},
|
||||||
|
}, {
|
||||||
|
'url': 'https://www.bilibili.com/cheese/play/ss5918',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '5918',
|
||||||
|
'title': '【限时五折】新闻系学不到:马督工教你做自媒体',
|
||||||
|
'description': '帮普通人建立世界模型,降低人与人的沟通门槛',
|
||||||
|
},
|
||||||
|
'playlist_mincount': 5,
|
||||||
|
'skip': 'paid video in list',
|
||||||
|
}]
|
||||||
|
|
||||||
|
def _get_cheese_entries(self, season_info):
|
||||||
|
for ep_id in traverse_obj(season_info, ('episodes', lambda _, v: v['episode_can_view'], 'id')):
|
||||||
|
yield self._extract_episode(season_info, ep_id)
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
season_id = self._match_id(url)
|
||||||
|
season_info = self._download_season_info('season_id', season_id)
|
||||||
|
|
||||||
|
return self.playlist_result(
|
||||||
|
self._get_cheese_entries(season_info), season_id,
|
||||||
|
**traverse_obj(season_info, {
|
||||||
|
'title': ('title', {str}),
|
||||||
|
'description': ('subtitle', {str}),
|
||||||
|
}))
|
||||||
|
|
||||||
|
|
||||||
class BilibiliSpaceBaseIE(InfoExtractor):
|
class BilibiliSpaceBaseIE(InfoExtractor):
|
||||||
|
|
Loading…
Reference in New Issue