from __future__ import unicode_literals import re from .common import InfoExtractor from ..compat import compat_urlparse from ..utils import ( determine_ext, encode_dict, ExtractorError, sanitized_Request, urlencode_postdata, ) class AnimeOnDemandIE(InfoExtractor): _VALID_URL = r'https?://(?:www\.)?anime-on-demand\.de/anime/(?P\d+)' _LOGIN_URL = 'https://www.anime-on-demand.de/users/sign_in' _APPLY_HTML5_URL = 'https://www.anime-on-demand.de/html5apply' _NETRC_MACHINE = 'animeondemand' _TESTS = [{ 'url': 'https://www.anime-on-demand.de/anime/161', 'info_dict': { 'id': '161', 'title': 'Grimgar, Ashes and Illusions (OmU)', 'description': 'md5:6681ce3c07c7189d255ac6ab23812d31', }, 'playlist_mincount': 4, }, { # Film wording is used instead of Episode 'url': 'https://www.anime-on-demand.de/anime/39', 'only_matching': True, }] def _login(self): (username, password) = self._get_login_info() if username is None: return login_page = self._download_webpage( self._LOGIN_URL, None, 'Downloading login page') login_form = self._form_hidden_inputs('new_user', login_page) login_form.update({ 'user[login]': username, 'user[password]': password, }) post_url = self._search_regex( r']+action=(["\'])(?P.+?)\1', login_page, 'post url', default=self._LOGIN_URL, group='url') if not post_url.startswith('http'): post_url = compat_urlparse.urljoin(self._LOGIN_URL, post_url) request = sanitized_Request( post_url, urlencode_postdata(encode_dict(login_form))) request.add_header('Referer', self._LOGIN_URL) response = self._download_webpage( request, None, 'Logging in as %s' % username) if all(p not in response for p in ('>Logout<', 'href="/users/sign_out"')): error = self._search_regex( r'

(.+?)

', response, 'error', default=None) if error: raise ExtractorError('Unable to login: %s' % error, expected=True) raise ExtractorError('Unable to log in') def _real_initialize(self): self._login() def _real_extract(self, url): anime_id = self._match_id(url) webpage = self._download_webpage(url, anime_id) if 'data-playlist=' not in webpage: self._download_webpage( self._APPLY_HTML5_URL, anime_id, 'Activating HTML5 beta', 'Unable to apply HTML5 beta') webpage = self._download_webpage(url, anime_id) csrf_token = self._html_search_meta( 'csrf-token', webpage, 'csrf token', fatal=True) anime_title = self._html_search_regex( r'(?s)]+itemprop="name"[^>]*>(.+?)', webpage, 'anime name') anime_description = self._html_search_regex( r'(?s)]+itemprop="description"[^>]*>(.+?)', webpage, 'anime description', default=None) entries = [] for episode_html in re.findall(r'(?s)]+class="episodebox-title".+?>Episodeninhalt<', webpage): m = re.search( r'class="episodebox-title"[^>]+title="(?:Episode|Film)\s*(?P\d+)\s*-\s*(?P.+?)"', episode_html) if not m: continue episode_number = int(m.group('number')) episode_title = m.group('title') video_id = 'episode-%d' % episode_number common_info = { 'id': video_id, 'series': anime_title, 'episode': episode_title, 'episode_number': episode_number, } formats = [] playlist_url = self._search_regex( r'data-playlist=(["\'])(?P<url>.+?)\1', episode_html, 'data playlist', default=None, group='url') if playlist_url: request = sanitized_Request( compat_urlparse.urljoin(url, playlist_url), headers={ 'X-Requested-With': 'XMLHttpRequest', 'X-CSRF-Token': csrf_token, 'Referer': url, 'Accept': 'application/json, text/javascript, */*; q=0.01', }) playlist = self._download_json( request, video_id, 'Downloading playlist JSON', fatal=False) if playlist: playlist = playlist['playlist'][0] title = playlist['title'] description = playlist.get('description') for source in playlist.get('sources', []): file_ = source.get('file') if file_ and determine_ext(file_) == 'm3u8': formats = self._extract_m3u8_formats( file_, video_id, 'mp4', entry_protocol='m3u8_native', m3u8_id='hls') if formats: f = common_info.copy() f.update({ 'title': title, 'description': description, 'formats': formats, }) entries.append(f) m = re.search( r'data-dialog-header=(["\'])(?P<title>.+?)\1[^>]+href=(["\'])(?P<href>.+?)\3[^>]*>Teaser<', episode_html) if m: f = common_info.copy() f.update({ 'id': '%s-teaser' % f['id'], 'title': m.group('title'), 'url': compat_urlparse.urljoin(url, m.group('href')), }) entries.append(f) return self.playlist_result(entries, anime_id, anime_title, anime_description)