You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
youtube-dl/youtube_dl/extractor/nhk.py

194 lines
7.2 KiB
Python

This file contains ambiguous Unicode characters!

This file contains ambiguous Unicode characters that may be confused with others in your current locale. If your use case is intentional and legitimate, you can safely ignore this warning. Use the Escape button to highlight these characters.

# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..utils import urljoin
class NhkBaseIE(InfoExtractor):
_API_URL_TEMPLATE = 'https://nwapi.nhk.jp/nhkworld/%sod%slist/v7b/%s/%s/%s/all%s.json'
_BASE_URL_REGEX = r'https?://www3\.nhk\.or\.jp/nhkworld/(?P<lang>[a-z]{2})/ondemand'
_TYPE_REGEX = r'/(?P<type>video|audio)/'
def _call_api(self, m_id, lang, is_video, is_episode, is_clip):
return self._download_json(
self._API_URL_TEMPLATE % (
'v' if is_video else 'r',
'clip' if is_clip else 'esd',
'episode' if is_episode else 'program',
m_id, lang, '/all' if is_video else ''),
m_id, query={'apikey': 'EJfK8jdS57GqlupFgAfAAwr573q01y6k'})['data']['episodes'] or []
def _extract_episode_info(self, url, episode=None):
fetch_episode = episode is None
lang, m_type, episode_id = re.match(NhkVodIE._VALID_URL, url).groups()
if len(episode_id) == 7:
episode_id = episode_id[:4] + '-' + episode_id[4:]
is_video = m_type == 'video'
if fetch_episode:
episode = self._call_api(
episode_id, lang, is_video, True, episode_id[:4] == '9999')[0]
title = episode.get('sub_title_clean') or episode['sub_title']
def get_clean_field(key):
return episode.get(key + '_clean') or episode.get(key)
series = get_clean_field('title')
thumbnails = []
for s, w, h in [('', 640, 360), ('_l', 1280, 720)]:
img_path = episode.get('image' + s)
if not img_path:
continue
thumbnails.append({
'id': '%dp' % h,
'height': h,
'width': w,
'url': 'https://www3.nhk.or.jp' + img_path,
})
info = {
'id': episode_id + '-' + lang,
'title': '%s - %s' % (series, title) if series and title else title,
'description': get_clean_field('description'),
'thumbnails': thumbnails,
'series': series,
'episode': title,
}
if is_video:
vod_id = episode['vod_id']
info.update({
'_type': 'url_transparent',
'ie_key': 'Piksel',
'url': 'https://player.piksel.com/v/refid/nhkworld/prefid/' + vod_id,
'id': vod_id,
})
else:
if fetch_episode:
audio_path = episode['audio']['audio']
info['formats'] = self._extract_m3u8_formats(
'https://nhkworld-vh.akamaihd.net/i%s/master.m3u8' % audio_path,
episode_id, 'm4a', entry_protocol='m3u8_native',
m3u8_id='hls', fatal=False)
for f in info['formats']:
f['language'] = lang
else:
info.update({
'_type': 'url_transparent',
'ie_key': NhkVodIE.ie_key(),
'url': url,
})
return info
class NhkVodIE(NhkBaseIE):
# the 7-character IDs can have alphabetic chars too: assume [a-z] rather than just [a-f], eg
_VALID_URL = r'%s%s(?P<id>[0-9a-z]{7}|[^/]+?-\d{8}-[0-9a-z]+)' % (NhkBaseIE._BASE_URL_REGEX, NhkBaseIE._TYPE_REGEX)
# Content available only for a limited period of time. Visit
# https://www3.nhk.or.jp/nhkworld/en/ondemand/ for working samples.
_TESTS = [{
# video clip
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/video/9999011/',
'md5': '7a90abcfe610ec22a6bfe15bd46b30ca',
'info_dict': {
'id': 'a95j5iza',
'ext': 'mp4',
'title': "Dining with the Chef - Chef Saito's Family recipe: MENCHI-KATSU",
'description': 'md5:5aee4a9f9d81c26281862382103b0ea5',
'timestamp': 1565965194,
'upload_date': '20190816',
},
}, {
# audio clip
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/audio/r_inventions-20201104-1/',
'info_dict': {
'id': 'r_inventions-20201104-1-en',
'ext': 'm4a',
'title': "Japan's Top Inventions - Miniature Video Cameras",
'description': 'md5:07ea722bdbbb4936fdd360b6a480c25b',
},
'params': {
# m3u8 download
'skip_download': True,
},
}, {
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/video/2015173/',
'only_matching': True,
}, {
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/audio/plugin-20190404-1/',
'only_matching': True,
}, {
'url': 'https://www3.nhk.or.jp/nhkworld/fr/ondemand/audio/plugin-20190404-1/',
'only_matching': True,
}, {
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/audio/j_art-20150903-1/',
'only_matching': True,
}, {
# video, alphabetic character in ID #29670
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/video/9999a34/',
'only_matching': True,
'info_dict': {
'id': 'qfjay6cg',
'ext': 'mp4',
'title': 'DESIGN TALKS plus - Fishermens Finery',
'description': 'md5:8a8f958aaafb0d7cb59d38de53f1e448',
'thumbnail': r're:^https?:/(/[a-z0-9.-]+)+\.jpg\?w=1920&h=1080$',
'upload_date': '20210615',
'timestamp': 1623722008,
}
}]
def _real_extract(self, url):
return self._extract_episode_info(url)
class NhkVodProgramIE(NhkBaseIE):
_VALID_URL = r'%s/program%s(?P<id>[0-9a-z]+)(?:.+?\btype=(?P<episode_type>clip|(?:radio|tv)Episode))?' % (NhkBaseIE._BASE_URL_REGEX, NhkBaseIE._TYPE_REGEX)
_TESTS = [{
# video program episodes
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/program/video/japanrailway',
'info_dict': {
'id': 'japanrailway',
'title': 'Japan Railway Journal',
},
'playlist_mincount': 1,
}, {
# video program clips
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/program/video/japanrailway/?type=clip',
'info_dict': {
'id': 'japanrailway',
'title': 'Japan Railway Journal',
},
'playlist_mincount': 5,
}, {
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/program/video/10yearshayaomiyazaki/',
'only_matching': True,
}, {
# audio program
'url': 'https://www3.nhk.or.jp/nhkworld/en/ondemand/program/audio/listener/',
'only_matching': True,
}]
def _real_extract(self, url):
lang, m_type, program_id, episode_type = re.match(self._VALID_URL, url).groups()
episodes = self._call_api(
program_id, lang, m_type == 'video', False, episode_type == 'clip')
entries = []
for episode in episodes:
episode_path = episode.get('url')
if not episode_path:
continue
entries.append(self._extract_episode_info(
urljoin(url, episode_path), episode))
program_title = None
if entries:
program_title = entries[0].get('series')
return self.playlist_result(entries, program_id, program_title)