yt-dlp/yt_dlp/extractor/naver.py

405 lines
16 KiB
Python

import base64
import hashlib
import hmac
import itertools
import json
import re
import time
import urllib.parse
from .common import InfoExtractor
from ..utils import (
ExtractorError,
dict_get,
int_or_none,
join_nonempty,
merge_dicts,
parse_iso8601,
traverse_obj,
try_get,
unified_timestamp,
update_url_query,
url_or_none,
)
class NaverBaseIE(InfoExtractor):
_CAPTION_EXT_RE = r'\.(?:ttml|vtt)'
@staticmethod # NB: Used in WeverseIE
def process_subtitles(vod_data, process_url):
ret = {'subtitles': {}, 'automatic_captions': {}}
for caption in traverse_obj(vod_data, ('captions', 'list', ...)):
caption_url = caption.get('source')
if not caption_url:
continue
type_ = 'automatic_captions' if caption.get('type') == 'auto' else 'subtitles'
lang = caption.get('locale') or join_nonempty('language', 'country', from_dict=caption) or 'und'
if caption.get('type') == 'fan':
lang += '_fan{}'.format(next(i for i in itertools.count(1) if f'{lang}_fan{i}' not in ret[type_]))
ret[type_].setdefault(lang, []).extend({
'url': sub_url,
'name': join_nonempty('label', 'fanName', from_dict=caption, delim=' - '),
} for sub_url in process_url(caption_url))
return ret
def _extract_video_info(self, video_id, vid, key):
video_data = self._download_json(
'http://play.rmcnmv.naver.com/vod/play/v2.0/' + vid,
video_id, query={
'key': key,
})
meta = video_data['meta']
title = meta['subject']
formats = []
get_list = lambda x: try_get(video_data, lambda y: y[x + 's']['list'], list) or []
def extract_formats(streams, stream_type, query={}):
for stream in streams:
stream_url = stream.get('source')
if not stream_url:
continue
stream_url = update_url_query(stream_url, query)
encoding_option = stream.get('encodingOption', {})
bitrate = stream.get('bitrate', {})
formats.append({
'format_id': '{}_{}'.format(stream.get('type') or stream_type, dict_get(encoding_option, ('name', 'id'))),
'url': stream_url,
'ext': 'mp4',
'width': int_or_none(encoding_option.get('width')),
'height': int_or_none(encoding_option.get('height')),
'vbr': int_or_none(bitrate.get('video')),
'abr': int_or_none(bitrate.get('audio')),
'filesize': int_or_none(stream.get('size')),
'protocol': 'm3u8_native' if stream_type == 'HLS' else None,
})
extract_formats(get_list('video'), 'H264')
for stream_set in video_data.get('streams', []):
query = {}
for param in stream_set.get('keys', []):
query[param['name']] = param['value']
stream_type = stream_set.get('type')
videos = stream_set.get('videos')
if videos:
extract_formats(videos, stream_type, query)
elif stream_type == 'HLS':
stream_url = stream_set.get('source')
if not stream_url:
continue
formats.extend(self._extract_m3u8_formats(
update_url_query(stream_url, query), video_id,
'mp4', 'm3u8_native', m3u8_id=stream_type, fatal=False))
replace_ext = lambda x, y: re.sub(self._CAPTION_EXT_RE, '.' + y, x)
def get_subs(caption_url):
if re.search(self._CAPTION_EXT_RE, caption_url):
return [
replace_ext(caption_url, 'ttml'),
replace_ext(caption_url, 'vtt'),
]
return [caption_url]
user = meta.get('user', {})
return {
'id': video_id,
'title': title,
'formats': formats,
'thumbnail': try_get(meta, lambda x: x['cover']['source']),
'view_count': int_or_none(meta.get('count')),
'uploader_id': user.get('id'),
'uploader': user.get('name'),
'uploader_url': user.get('url'),
**self.process_subtitles(video_data, get_subs),
}
def _call_api(self, path, video_id):
api_endpoint = f'https://apis.naver.com/now_web2/now_web_api/v1{path}'
key = b'nbxvs5nwNG9QKEWK0ADjYA4JZoujF4gHcIwvoCxFTPAeamq5eemvt5IWAYXxrbYM'
msgpad = int(time.time() * 1000)
md = base64.b64encode(hmac.HMAC(
key, f'{api_endpoint[:255]}{msgpad}'.encode(), digestmod=hashlib.sha1).digest()).decode()
return self._download_json(api_endpoint, video_id=video_id, headers=self.geo_verification_headers(), query={
'msgpad': msgpad,
'md': md,
})['result']
class NaverIE(NaverBaseIE):
_VALID_URL = r'https?://(?:m\.)?tv(?:cast)?\.naver\.com/(?:v|embed)/(?P<id>\d+)'
_GEO_BYPASS = False
_TESTS = [{
'url': 'http://tv.naver.com/v/81652',
'info_dict': {
'id': '81652',
'ext': 'mp4',
'title': '[9월 모의고사 해설강의][수학_김상희] 수학 A형 16~20번',
'description': '메가스터디 수학 김상희 선생님이 9월 모의고사 수학A형 16번에서 20번까지 해설강의를 공개합니다.',
'timestamp': 1378200754,
'upload_date': '20130903',
'uploader': '메가스터디, 합격불변의 법칙',
'uploader_id': 'megastudy',
'uploader_url': 'https://tv.naver.com/megastudy',
'view_count': int,
'like_count': int,
'comment_count': int,
'duration': 2118,
'thumbnail': r're:^https?://.*\.jpg',
},
}, {
'url': 'http://tv.naver.com/v/395837',
'md5': '7791205fa89dbed2f5e3eb16d287ff05',
'info_dict': {
'id': '395837',
'ext': 'mp4',
'title': '9년이 지나도 아픈 기억, 전효성의 아버지',
'description': 'md5:c76be23e21403a6473d8119678cdb5cb',
'timestamp': 1432030253,
'upload_date': '20150519',
'uploader': '4가지쇼',
'uploader_id': '4show',
'uploader_url': 'https://tv.naver.com/4show',
'view_count': int,
'like_count': int,
'comment_count': int,
'duration': 277,
'thumbnail': r're:^https?://.*\.jpg',
},
}, {
'url': 'http://tvcast.naver.com/v/81652',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
data = self._call_api(f'/clips/{video_id}/play-info', video_id)
vid = traverse_obj(data, ('clip', 'videoId', {str}))
in_key = traverse_obj(data, ('play', 'inKey', {str}))
if not vid or not in_key:
raise ExtractorError('Unable to extract video info')
info = self._extract_video_info(video_id, vid, in_key)
info.update(traverse_obj(data, ('clip', {
'title': 'title',
'description': 'description',
'timestamp': ('firstExposureDatetime', {parse_iso8601}),
'duration': ('playTime', {int_or_none}),
'like_count': ('likeItCount', {int_or_none}),
'view_count': ('playCount', {int_or_none}),
'comment_count': ('commentCount', {int_or_none}),
'thumbnail': ('thumbnailImageUrl', {url_or_none}),
'uploader': 'channelName',
'uploader_id': 'channelId',
'uploader_url': ('channelUrl', {url_or_none}),
'age_limit': ('adultVideo', {lambda x: 19 if x else None}),
})))
return info
class NaverLiveIE(NaverBaseIE):
IE_NAME = 'Naver:live'
_VALID_URL = r'https?://(?:m\.)?tv(?:cast)?\.naver\.com/l/(?P<id>\d+)'
_GEO_BYPASS = False
_TESTS = [{
'url': 'https://tv.naver.com/l/127062',
'info_dict': {
'id': '127062',
'ext': 'mp4',
'live_status': 'is_live',
'channel': '뉴스는 YTN',
'channel_id': 'ytnnews24',
'title': 're:^대한민국 24시간 뉴스 채널 [0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}$',
'description': 'md5:f938b5956711beab6f882314ffadf4d5',
'start_time': 1677752280,
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)',
'like_count': int,
},
}, {
'url': 'https://tv.naver.com/l/140535',
'info_dict': {
'id': '140535',
'ext': 'mp4',
'live_status': 'is_live',
'channel': 'KBS뉴스',
'channel_id': 'kbsnews',
'start_time': 1696867320,
'title': 're:^언제 어디서나! KBS 뉴스 24 [0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}$',
'description': 'md5:6ad419c0bf2f332829bda3f79c295284',
'thumbnail': r're:^https?://.*\.(jpg|jpeg|png)',
'like_count': int,
},
}, {
'url': 'https://tv.naver.com/l/54887',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
data = self._call_api(f'/live-end/normal/{video_id}/play-info?renewLastPlayDate=true', video_id)
status = traverse_obj(data, ('live', 'liveStatus'))
if status == 'CLOSED':
raise ExtractorError('Stream is offline.', expected=True)
elif status != 'OPENED':
raise ExtractorError(f'Unknown status {status!r}')
return {
'id': video_id,
'formats': self._extract_m3u8_formats(
traverse_obj(data, ('playbackBody', {json.loads}, 'media', 0, 'path')), video_id, live=True),
**traverse_obj(data, ('live', {
'title': 'title',
'channel': 'channelName',
'channel_id': 'channelId',
'description': 'description',
'like_count': (('likeCount', 'likeItCount'), {int_or_none}),
'thumbnail': ('thumbnailImageUrl', {url_or_none}),
'start_time': (('startTime', 'startDateTime', 'startYmdt'), {parse_iso8601}),
}), get_all=False),
'is_live': True,
}
class NaverNowIE(NaverBaseIE):
IE_NAME = 'navernow'
_VALID_URL = r'https?://now\.naver\.com/s/now\.(?P<id>\w+)'
_API_URL = 'https://apis.naver.com/now_web/oldnow_web/v4'
_TESTS = [{
'url': 'https://now.naver.com/s/now.4759?shareReplayId=26331132#replay=',
'md5': 'e05854162c21c221481de16b2944a0bc',
'info_dict': {
'id': '4759-26331132',
'title': '아이키X노제\r\n💖꽁냥꽁냥💖(1)',
'ext': 'mp4',
'thumbnail': r're:^https?://.*\.jpg',
'timestamp': 1650369600,
'upload_date': '20220419',
'uploader_id': 'now',
'view_count': int,
'uploader_url': 'https://now.naver.com/show/4759',
'uploader': '아이키의 떰즈업',
},
'params': {
'noplaylist': True,
},
}, {
'url': 'https://now.naver.com/s/now.4759?shareHightlight=26601461#highlight=',
'md5': '9f6118e398aa0f22b2152f554ea7851b',
'info_dict': {
'id': '4759-26601461',
'title': '아이키: 나 리정한테 흔들렸어,,, 질투 폭발하는 노제 여보😾 [아이키의 떰즈업]ㅣ네이버 NOW.',
'ext': 'mp4',
'thumbnail': r're:^https?://.*\.jpg',
'upload_date': '20220504',
'timestamp': 1651648311,
'uploader_id': 'now',
'view_count': int,
'uploader_url': 'https://now.naver.com/show/4759',
'uploader': '아이키의 떰즈업',
},
'params': {
'noplaylist': True,
},
}, {
'url': 'https://now.naver.com/s/now.4759',
'info_dict': {
'id': '4759',
'title': '아이키의 떰즈업',
},
'playlist_mincount': 101,
}, {
'url': 'https://now.naver.com/s/now.4759?shareReplayId=26331132#replay',
'info_dict': {
'id': '4759',
'title': '아이키의 떰즈업',
},
'playlist_mincount': 101,
}, {
'url': 'https://now.naver.com/s/now.4759?shareHightlight=26601461#highlight=',
'info_dict': {
'id': '4759',
'title': '아이키의 떰즈업',
},
'playlist_mincount': 101,
}, {
'url': 'https://now.naver.com/s/now.kihyunplay?shareReplayId=30573291#replay',
'only_matching': True,
}]
def _extract_replay(self, show_id, replay_id):
vod_info = self._download_json(f'{self._API_URL}/shows/now.{show_id}/vod/{replay_id}', replay_id)
in_key = self._download_json(f'{self._API_URL}/shows/now.{show_id}/vod/{replay_id}/inkey', replay_id)['inKey']
return merge_dicts({
'id': f'{show_id}-{replay_id}',
'title': traverse_obj(vod_info, ('episode', 'title')),
'timestamp': unified_timestamp(traverse_obj(vod_info, ('episode', 'start_time'))),
'thumbnail': vod_info.get('thumbnail_image_url'),
}, self._extract_video_info(replay_id, vod_info['video_id'], in_key))
def _extract_show_replays(self, show_id):
page_size = 15
page = 1
while True:
show_vod_info = self._download_json(
f'{self._API_URL}/vod-shows/now.{show_id}', show_id,
query={'page': page, 'page_size': page_size},
note=f'Downloading JSON vod list for show {show_id} - page {page}',
)['response']['result']
for v in show_vod_info.get('vod_list') or []:
yield self._extract_replay(show_id, v['id'])
if len(show_vod_info.get('vod_list') or []) < page_size:
break
page += 1
def _extract_show_highlights(self, show_id, highlight_id=None):
page_size = 10
page = 1
while True:
highlights_videos = self._download_json(
f'{self._API_URL}/shows/now.{show_id}/highlights/videos/', show_id,
query={'page': page, 'page_size': page_size},
note=f'Downloading JSON highlights for show {show_id} - page {page}')
for highlight in highlights_videos.get('results') or []:
if highlight_id and highlight.get('clip_no') != int(highlight_id):
continue
yield merge_dicts({
'id': f'{show_id}-{highlight["clip_no"]}',
'title': highlight.get('title'),
'timestamp': unified_timestamp(highlight.get('regdate')),
'thumbnail': highlight.get('thumbnail_url'),
}, self._extract_video_info(highlight['clip_no'], highlight['video_id'], highlight['video_inkey']))
if len(highlights_videos.get('results') or []) < page_size:
break
page += 1
def _extract_highlight(self, show_id, highlight_id):
try:
return next(self._extract_show_highlights(show_id, highlight_id))
except StopIteration:
raise ExtractorError(f'Unable to find highlight {highlight_id} for show {show_id}')
def _real_extract(self, url):
show_id = self._match_id(url)
qs = urllib.parse.parse_qs(urllib.parse.urlparse(url).query)
if not self._yes_playlist(show_id, qs.get('shareHightlight')):
return self._extract_highlight(show_id, qs['shareHightlight'][0])
elif not self._yes_playlist(show_id, qs.get('shareReplayId')):
return self._extract_replay(show_id, qs['shareReplayId'][0])
show_info = self._download_json(
f'{self._API_URL}/shows/now.{show_id}/', show_id,
note=f'Downloading JSON vod list for show {show_id}')
return self.playlist_result(
itertools.chain(self._extract_show_replays(show_id), self._extract_show_highlights(show_id)),
show_id, show_info.get('title'))