mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-14 04:13:20 +00:00
315b354429
Closes #6497 Authored by: bashonly
429 lines
17 KiB
Python
429 lines
17 KiB
Python
import functools
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
ExtractorError,
|
|
OnDemandPagedList,
|
|
UserNotLive,
|
|
determine_ext,
|
|
filter_dict,
|
|
int_or_none,
|
|
orderedSet,
|
|
unified_timestamp,
|
|
url_or_none,
|
|
urlencode_postdata,
|
|
urljoin,
|
|
)
|
|
from ..utils.traversal import traverse_obj
|
|
|
|
|
|
class AfreecaTVBaseIE(InfoExtractor):
|
|
_NETRC_MACHINE = 'afreecatv'
|
|
|
|
def _perform_login(self, username, password):
|
|
login_form = {
|
|
'szWork': 'login',
|
|
'szType': 'json',
|
|
'szUid': username,
|
|
'szPassword': password,
|
|
'isSaveId': 'false',
|
|
'szScriptVar': 'oLoginRet',
|
|
'szAction': '',
|
|
}
|
|
|
|
response = self._download_json(
|
|
'https://login.afreecatv.com/app/LoginAction.php', None,
|
|
'Logging in', data=urlencode_postdata(login_form))
|
|
|
|
_ERRORS = {
|
|
-4: 'Your account has been suspended due to a violation of our terms and policies.',
|
|
-5: 'https://member.afreecatv.com/app/user_delete_progress.php',
|
|
-6: 'https://login.afreecatv.com/membership/changeMember.php',
|
|
-8: "Hello! AfreecaTV here.\nThe username you have entered belongs to \n an account that requires a legal guardian's consent. \nIf you wish to use our services without restriction, \nplease make sure to go through the necessary verification process.",
|
|
-9: 'https://member.afreecatv.com/app/pop_login_block.php',
|
|
-11: 'https://login.afreecatv.com/afreeca/second_login.php',
|
|
-12: 'https://member.afreecatv.com/app/user_security.php',
|
|
0: 'The username does not exist or you have entered the wrong password.',
|
|
-1: 'The username does not exist or you have entered the wrong password.',
|
|
-3: 'You have entered your username/password incorrectly.',
|
|
-7: 'You cannot use your Global AfreecaTV account to access Korean AfreecaTV.',
|
|
-10: 'Sorry for the inconvenience. \nYour account has been blocked due to an unauthorized access. \nPlease contact our Help Center for assistance.',
|
|
-32008: 'You have failed to log in. Please contact our Help Center.',
|
|
}
|
|
|
|
result = int_or_none(response.get('RESULT'))
|
|
if result != 1:
|
|
error = _ERRORS.get(result, 'You have failed to log in.')
|
|
raise ExtractorError(
|
|
'Unable to login: %s said: %s' % (self.IE_NAME, error),
|
|
expected=True)
|
|
|
|
|
|
class AfreecaTVIE(AfreecaTVBaseIE):
|
|
IE_NAME = 'afreecatv'
|
|
IE_DESC = 'afreecatv.com'
|
|
_VALID_URL = r'''(?x)
|
|
https?://
|
|
(?:
|
|
(?:(?:live|afbbs|www)\.)?afreeca(?:tv)?\.com(?::\d+)?
|
|
(?:
|
|
/app/(?:index|read_ucc_bbs)\.cgi|
|
|
/player/[Pp]layer\.(?:swf|html)
|
|
)\?.*?\bnTitleNo=|
|
|
vod\.afreecatv\.com/(PLAYER/STATION|player)/
|
|
)
|
|
(?P<id>\d+)
|
|
'''
|
|
_TESTS = [{
|
|
'url': 'http://live.afreecatv.com:8079/app/index.cgi?szType=read_ucc_bbs&szBjId=dailyapril&nStationNo=16711924&nBbsNo=18605867&nTitleNo=36164052&szSkin=',
|
|
'md5': 'f72c89fe7ecc14c1b5ce506c4996046e',
|
|
'info_dict': {
|
|
'id': '36164052',
|
|
'ext': 'mp4',
|
|
'title': '데일리 에이프릴 요정들의 시상식!',
|
|
'thumbnail': 're:^https?://(?:video|st)img.afreecatv.com/.*$',
|
|
'uploader': 'dailyapril',
|
|
'uploader_id': 'dailyapril',
|
|
'upload_date': '20160503',
|
|
},
|
|
'skip': 'Video is gone',
|
|
}, {
|
|
'url': 'http://afbbs.afreecatv.com:8080/app/read_ucc_bbs.cgi?nStationNo=16711924&nTitleNo=36153164&szBjId=dailyapril&nBbsNo=18605867',
|
|
'info_dict': {
|
|
'id': '36153164',
|
|
'title': "BJ유트루와 함께하는 '팅커벨 메이크업!'",
|
|
'thumbnail': 're:^https?://(?:video|st)img.afreecatv.com/.*$',
|
|
'uploader': 'dailyapril',
|
|
'uploader_id': 'dailyapril',
|
|
},
|
|
'playlist_count': 2,
|
|
'playlist': [{
|
|
'md5': 'd8b7c174568da61d774ef0203159bf97',
|
|
'info_dict': {
|
|
'id': '36153164_1',
|
|
'ext': 'mp4',
|
|
'title': "BJ유트루와 함께하는 '팅커벨 메이크업!'",
|
|
'upload_date': '20160502',
|
|
},
|
|
}, {
|
|
'md5': '58f2ce7f6044e34439ab2d50612ab02b',
|
|
'info_dict': {
|
|
'id': '36153164_2',
|
|
'ext': 'mp4',
|
|
'title': "BJ유트루와 함께하는 '팅커벨 메이크업!'",
|
|
'upload_date': '20160502',
|
|
},
|
|
}],
|
|
'skip': 'Video is gone',
|
|
}, {
|
|
# non standard key
|
|
'url': 'http://vod.afreecatv.com/PLAYER/STATION/20515605',
|
|
'info_dict': {
|
|
'id': '20170411_BE689A0E_190960999_1_2_h',
|
|
'ext': 'mp4',
|
|
'title': '혼자사는여자집',
|
|
'thumbnail': 're:^https?://(?:video|st)img.afreecatv.com/.*$',
|
|
'uploader': '♥이슬이',
|
|
'uploader_id': 'dasl8121',
|
|
'upload_date': '20170411',
|
|
'timestamp': 1491929865,
|
|
'duration': 213,
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
# adult content
|
|
'url': 'https://vod.afreecatv.com/player/97267690',
|
|
'info_dict': {
|
|
'id': '20180327_27901457_202289533_1',
|
|
'ext': 'mp4',
|
|
'title': '[생]빨개요♥ (part 1)',
|
|
'thumbnail': 're:^https?://(?:video|st)img.afreecatv.com/.*$',
|
|
'uploader': '[SA]서아',
|
|
'uploader_id': 'bjdyrksu',
|
|
'upload_date': '20180327',
|
|
'duration': 3601,
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
'skip': 'The VOD does not exist',
|
|
}, {
|
|
'url': 'http://www.afreecatv.com/player/Player.swf?szType=szBjId=djleegoon&nStationNo=11273158&nBbsNo=13161095&nTitleNo=36327652',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'https://vod.afreecatv.com/player/96753363',
|
|
'info_dict': {
|
|
'id': '20230108_9FF5BEE1_244432674_1',
|
|
'ext': 'mp4',
|
|
'uploader_id': 'rlantnghks',
|
|
'uploader': '페이즈으',
|
|
'duration': 10840,
|
|
'thumbnail': r're:https?://videoimg\.afreecatv\.com/.+',
|
|
'upload_date': '20230108',
|
|
'timestamp': 1673218805,
|
|
'title': '젠지 페이즈',
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
# adult content
|
|
'url': 'https://vod.afreecatv.com/player/70395877',
|
|
'only_matching': True,
|
|
}, {
|
|
# subscribers only
|
|
'url': 'https://vod.afreecatv.com/player/104647403',
|
|
'only_matching': True,
|
|
}, {
|
|
# private
|
|
'url': 'https://vod.afreecatv.com/player/81669846',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
data = self._download_json(
|
|
'https://api.m.afreecatv.com/station/video/a/view', video_id,
|
|
headers={'Referer': url}, data=urlencode_postdata({
|
|
'nTitleNo': video_id,
|
|
'nApiLevel': 10,
|
|
}))['data']
|
|
|
|
error_code = traverse_obj(data, ('code', {int}))
|
|
if error_code == -6221:
|
|
raise ExtractorError('The VOD does not exist', expected=True)
|
|
elif error_code == -6205:
|
|
raise ExtractorError('This VOD is private', expected=True)
|
|
|
|
common_info = traverse_obj(data, {
|
|
'title': ('title', {str}),
|
|
'uploader': ('writer_nick', {str}),
|
|
'uploader_id': ('bj_id', {str}),
|
|
'duration': ('total_file_duration', {functools.partial(int_or_none, scale=1000)}),
|
|
'thumbnail': ('thumb', {url_or_none}),
|
|
})
|
|
|
|
entries = []
|
|
for file_num, file_element in enumerate(
|
|
traverse_obj(data, ('files', lambda _, v: url_or_none(v['file']))), start=1):
|
|
file_url = file_element['file']
|
|
if determine_ext(file_url) == 'm3u8':
|
|
formats = self._extract_m3u8_formats(
|
|
file_url, video_id, 'mp4', m3u8_id='hls',
|
|
note=f'Downloading part {file_num} m3u8 information')
|
|
else:
|
|
formats = [{
|
|
'url': file_url,
|
|
'format_id': 'http',
|
|
}]
|
|
|
|
entries.append({
|
|
**common_info,
|
|
'id': file_element.get('file_info_key') or f'{video_id}_{file_num}',
|
|
'title': f'{common_info.get("title") or "Untitled"} (part {file_num})',
|
|
'formats': formats,
|
|
**traverse_obj(file_element, {
|
|
'duration': ('duration', {functools.partial(int_or_none, scale=1000)}),
|
|
'timestamp': ('file_start', {unified_timestamp}),
|
|
})
|
|
})
|
|
|
|
if traverse_obj(data, ('adult_status', {str})) == 'notLogin':
|
|
if not entries:
|
|
self.raise_login_required(
|
|
'Only users older than 19 are able to watch this video', method='password')
|
|
self.report_warning(
|
|
'In accordance with local laws and regulations, underage users are '
|
|
'restricted from watching adult content. Only content suitable for all '
|
|
f'ages will be downloaded. {self._login_hint("password")}')
|
|
|
|
if not entries and traverse_obj(data, ('sub_upload_type', {str})):
|
|
self.raise_login_required('This VOD is for subscribers only', method='password')
|
|
|
|
if len(entries) == 1:
|
|
return {
|
|
**entries[0],
|
|
'title': common_info.get('title'),
|
|
}
|
|
|
|
common_info['timestamp'] = traverse_obj(entries, (..., 'timestamp'), get_all=False)
|
|
|
|
return self.playlist_result(entries, video_id, multi_video=True, **common_info)
|
|
|
|
|
|
class AfreecaTVLiveIE(AfreecaTVBaseIE):
|
|
IE_NAME = 'afreecatv:live'
|
|
IE_DESC = 'afreecatv.com livestreams'
|
|
_VALID_URL = r'https?://play\.afreeca(?:tv)?\.com/(?P<id>[^/]+)(?:/(?P<bno>\d+))?'
|
|
_TESTS = [{
|
|
'url': 'https://play.afreecatv.com/pyh3646/237852185',
|
|
'info_dict': {
|
|
'id': '237852185',
|
|
'ext': 'mp4',
|
|
'title': '【 우루과이 오늘은 무슨일이? 】',
|
|
'uploader': '박진우[JINU]',
|
|
'uploader_id': 'pyh3646',
|
|
'timestamp': 1640661495,
|
|
'is_live': True,
|
|
},
|
|
'skip': 'Livestream has ended',
|
|
}, {
|
|
'url': 'https://play.afreecatv.com/pyh3646/237852185',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'https://play.afreecatv.com/pyh3646',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
_LIVE_API_URL = 'https://live.afreecatv.com/afreeca/player_live_api.php'
|
|
_WORKING_CDNS = [
|
|
'gcp_cdn', # live-global-cdn-v02.afreecatv.com
|
|
'gs_cdn_pc_app', # pc-app.stream.afreecatv.com
|
|
'gs_cdn_mobile_web', # mobile-web.stream.afreecatv.com
|
|
'gs_cdn_pc_web', # pc-web.stream.afreecatv.com
|
|
]
|
|
_BAD_CDNS = [
|
|
'gs_cdn', # chromecast.afreeca.gscdn.com (cannot resolve)
|
|
'gs_cdn_chromecast', # chromecast.stream.afreecatv.com (HTTP Error 400)
|
|
'azure_cdn', # live-global-cdn-v01.afreecatv.com (cannot resolve)
|
|
'aws_cf', # live-global-cdn-v03.afreecatv.com (cannot resolve)
|
|
'kt_cdn', # kt.stream.afreecatv.com (HTTP Error 400)
|
|
]
|
|
|
|
def _extract_formats(self, channel_info, broadcast_no, aid):
|
|
stream_base_url = channel_info.get('RMD') or 'https://livestream-manager.afreecatv.com'
|
|
|
|
# If user has not passed CDN IDs, try API-provided CDN ID followed by other working CDN IDs
|
|
default_cdn_ids = orderedSet([
|
|
*traverse_obj(channel_info, ('CDN', {str}, all, lambda _, v: v not in self._BAD_CDNS)),
|
|
*self._WORKING_CDNS,
|
|
])
|
|
cdn_ids = self._configuration_arg('cdn', default_cdn_ids)
|
|
|
|
for attempt, cdn_id in enumerate(cdn_ids, start=1):
|
|
m3u8_url = traverse_obj(self._download_json(
|
|
urljoin(stream_base_url, 'broad_stream_assign.html'), broadcast_no,
|
|
f'Downloading {cdn_id} stream info', f'Unable to download {cdn_id} stream info',
|
|
fatal=False, query={
|
|
'return_type': cdn_id,
|
|
'broad_key': f'{broadcast_no}-common-master-hls',
|
|
}), ('view_url', {url_or_none}))
|
|
try:
|
|
return self._extract_m3u8_formats(
|
|
m3u8_url, broadcast_no, 'mp4', m3u8_id='hls', query={'aid': aid},
|
|
headers={'Referer': 'https://play.afreecatv.com/'})
|
|
except ExtractorError as e:
|
|
if attempt == len(cdn_ids):
|
|
raise
|
|
self.report_warning(
|
|
f'{e.cause or e.msg}. Retrying... (attempt {attempt} of {len(cdn_ids)})')
|
|
|
|
def _real_extract(self, url):
|
|
broadcaster_id, broadcast_no = self._match_valid_url(url).group('id', 'bno')
|
|
channel_info = traverse_obj(self._download_json(
|
|
self._LIVE_API_URL, broadcaster_id, data=urlencode_postdata({'bid': broadcaster_id})),
|
|
('CHANNEL', {dict})) or {}
|
|
|
|
broadcaster_id = channel_info.get('BJID') or broadcaster_id
|
|
broadcast_no = channel_info.get('BNO') or broadcast_no
|
|
if not broadcast_no:
|
|
raise UserNotLive(video_id=broadcaster_id)
|
|
|
|
password = self.get_param('videopassword')
|
|
if channel_info.get('BPWD') == 'Y' and password is None:
|
|
raise ExtractorError(
|
|
'This livestream is protected by a password, use the --video-password option',
|
|
expected=True)
|
|
|
|
token_info = traverse_obj(self._download_json(
|
|
self._LIVE_API_URL, broadcast_no, 'Downloading access token for stream',
|
|
'Unable to download access token for stream', data=urlencode_postdata(filter_dict({
|
|
'bno': broadcast_no,
|
|
'stream_type': 'common',
|
|
'type': 'aid',
|
|
'quality': 'master',
|
|
'pwd': password,
|
|
}))), ('CHANNEL', {dict})) or {}
|
|
aid = token_info.get('AID')
|
|
if not aid:
|
|
result = token_info.get('RESULT')
|
|
if result == 0:
|
|
raise ExtractorError('This livestream has ended', expected=True)
|
|
elif result == -6:
|
|
self.raise_login_required('This livestream is for subscribers only', method='password')
|
|
raise ExtractorError('Unable to extract access token')
|
|
|
|
formats = self._extract_formats(channel_info, broadcast_no, aid)
|
|
|
|
station_info = traverse_obj(self._download_json(
|
|
'https://st.afreecatv.com/api/get_station_status.php', broadcast_no,
|
|
'Downloading channel metadata', 'Unable to download channel metadata',
|
|
query={'szBjId': broadcaster_id}, fatal=False), {dict}) or {}
|
|
|
|
return {
|
|
'id': broadcast_no,
|
|
'title': channel_info.get('TITLE') or station_info.get('station_title'),
|
|
'uploader': channel_info.get('BJNICK') or station_info.get('station_name'),
|
|
'uploader_id': broadcaster_id,
|
|
'timestamp': unified_timestamp(station_info.get('broad_start')),
|
|
'formats': formats,
|
|
'is_live': True,
|
|
'http_headers': {'Referer': url},
|
|
}
|
|
|
|
|
|
class AfreecaTVUserIE(InfoExtractor):
|
|
IE_NAME = 'afreecatv:user'
|
|
_VALID_URL = r'https?://bj\.afreeca(?:tv)?\.com/(?P<id>[^/]+)/vods/?(?P<slug_type>[^/]+)?'
|
|
_TESTS = [{
|
|
'url': 'https://bj.afreecatv.com/ryuryu24/vods/review',
|
|
'info_dict': {
|
|
'_type': 'playlist',
|
|
'id': 'ryuryu24',
|
|
'title': 'ryuryu24 - review',
|
|
},
|
|
'playlist_count': 218,
|
|
}, {
|
|
'url': 'https://bj.afreecatv.com/parang1995/vods/highlight',
|
|
'info_dict': {
|
|
'_type': 'playlist',
|
|
'id': 'parang1995',
|
|
'title': 'parang1995 - highlight',
|
|
},
|
|
'playlist_count': 997,
|
|
}, {
|
|
'url': 'https://bj.afreecatv.com/ryuryu24/vods',
|
|
'info_dict': {
|
|
'_type': 'playlist',
|
|
'id': 'ryuryu24',
|
|
'title': 'ryuryu24 - all',
|
|
},
|
|
'playlist_count': 221,
|
|
}, {
|
|
'url': 'https://bj.afreecatv.com/ryuryu24/vods/balloonclip',
|
|
'info_dict': {
|
|
'_type': 'playlist',
|
|
'id': 'ryuryu24',
|
|
'title': 'ryuryu24 - balloonclip',
|
|
},
|
|
'playlist_count': 0,
|
|
}]
|
|
_PER_PAGE = 60
|
|
|
|
def _fetch_page(self, user_id, user_type, page):
|
|
page += 1
|
|
info = self._download_json(f'https://bjapi.afreecatv.com/api/{user_id}/vods/{user_type}', user_id,
|
|
query={'page': page, 'per_page': self._PER_PAGE, 'orderby': 'reg_date'},
|
|
note=f'Downloading {user_type} video page {page}')
|
|
for item in info['data']:
|
|
yield self.url_result(
|
|
f'https://vod.afreecatv.com/player/{item["title_no"]}/', AfreecaTVIE, item['title_no'])
|
|
|
|
def _real_extract(self, url):
|
|
user_id, user_type = self._match_valid_url(url).group('id', 'slug_type')
|
|
user_type = user_type or 'all'
|
|
entries = OnDemandPagedList(functools.partial(self._fetch_page, user_id, user_type), self._PER_PAGE)
|
|
return self.playlist_result(entries, user_id, f'{user_id} - {user_type}')
|