mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-25 00:00:48 +01:00
[animeondemand] Add support for full length films (Closes #10031)
This commit is contained in:
parent
0af985069b
commit
1f7258a367
@ -22,6 +22,7 @@ class AnimeOnDemandIE(InfoExtractor):
|
||||
_APPLY_HTML5_URL = 'https://www.anime-on-demand.de/html5apply'
|
||||
_NETRC_MACHINE = 'animeondemand'
|
||||
_TESTS = [{
|
||||
# jap, OmU
|
||||
'url': 'https://www.anime-on-demand.de/anime/161',
|
||||
'info_dict': {
|
||||
'id': '161',
|
||||
@ -30,17 +31,21 @@ class AnimeOnDemandIE(InfoExtractor):
|
||||
},
|
||||
'playlist_mincount': 4,
|
||||
}, {
|
||||
# Film wording is used instead of Episode
|
||||
# Film wording is used instead of Episode, ger/jap, Dub/OmU
|
||||
'url': 'https://www.anime-on-demand.de/anime/39',
|
||||
'only_matching': True,
|
||||
}, {
|
||||
# Episodes without titles
|
||||
# Episodes without titles, jap, OmU
|
||||
'url': 'https://www.anime-on-demand.de/anime/162',
|
||||
'only_matching': True,
|
||||
}, {
|
||||
# ger/jap, Dub/OmU, account required
|
||||
'url': 'https://www.anime-on-demand.de/anime/169',
|
||||
'only_matching': True,
|
||||
}, {
|
||||
# Full length film, non-series, ger/jap, Dub/OmU, account required
|
||||
'url': 'https://www.anime-on-demand.de/anime/185',
|
||||
'only_matching': True,
|
||||
}]
|
||||
|
||||
def _login(self):
|
||||
@ -110,35 +115,12 @@ def _real_extract(self, url):
|
||||
|
||||
entries = []
|
||||
|
||||
for num, episode_html in enumerate(re.findall(
|
||||
r'(?s)<h3[^>]+class="episodebox-title".+?>Episodeninhalt<', webpage), 1):
|
||||
episodebox_title = self._search_regex(
|
||||
(r'class="episodebox-title"[^>]+title=(["\'])(?P<title>.+?)\1',
|
||||
r'class="episodebox-title"[^>]+>(?P<title>.+?)<'),
|
||||
episode_html, 'episodebox title', default=None, group='title')
|
||||
if not episodebox_title:
|
||||
continue
|
||||
|
||||
episode_number = int(self._search_regex(
|
||||
r'(?:Episode|Film)\s*(\d+)',
|
||||
episodebox_title, 'episode number', default=num))
|
||||
episode_title = self._search_regex(
|
||||
r'(?:Episode|Film)\s*\d+\s*-\s*(.+)',
|
||||
episodebox_title, 'episode title', default=None)
|
||||
|
||||
video_id = 'episode-%d' % episode_number
|
||||
|
||||
common_info = {
|
||||
'id': video_id,
|
||||
'series': anime_title,
|
||||
'episode': episode_title,
|
||||
'episode_number': episode_number,
|
||||
}
|
||||
|
||||
def extract_info(html, video_id):
|
||||
title, description = [None] * 2
|
||||
formats = []
|
||||
|
||||
for input_ in re.findall(
|
||||
r'<input[^>]+class=["\'].*?streamstarter_html5[^>]+>', episode_html):
|
||||
r'<input[^>]+class=["\'].*?streamstarter_html5[^>]+>', html):
|
||||
attributes = extract_attributes(input_)
|
||||
playlist_urls = []
|
||||
for playlist_key in ('data-playlist', 'data-otherplaylist'):
|
||||
@ -215,28 +197,74 @@ def _real_extract(self, url):
|
||||
})
|
||||
formats.extend(file_formats)
|
||||
|
||||
if formats:
|
||||
self._sort_formats(formats)
|
||||
return {
|
||||
'title': title,
|
||||
'description': description,
|
||||
'formats': formats,
|
||||
}
|
||||
|
||||
def extract_entries(html, video_id, common_info):
|
||||
info = extract_info(html, video_id)
|
||||
|
||||
if info['formats']:
|
||||
self._sort_formats(info['formats'])
|
||||
f = common_info.copy()
|
||||
f.update({
|
||||
'title': title,
|
||||
'description': description,
|
||||
'formats': formats,
|
||||
})
|
||||
f.update(info)
|
||||
entries.append(f)
|
||||
|
||||
# Extract teaser only when full episode is not available
|
||||
if not formats:
|
||||
# Extract teaser/trailer only when full episode is not available
|
||||
if not info['formats']:
|
||||
m = re.search(
|
||||
r'data-dialog-header=(["\'])(?P<title>.+?)\1[^>]+href=(["\'])(?P<href>.+?)\3[^>]*>Teaser<',
|
||||
episode_html)
|
||||
r'data-dialog-header=(["\'])(?P<title>.+?)\1[^>]+href=(["\'])(?P<href>.+?)\3[^>]*>(?P<kind>Teaser|Trailer)<',
|
||||
html)
|
||||
if m:
|
||||
f = common_info.copy()
|
||||
f.update({
|
||||
'id': '%s-teaser' % f['id'],
|
||||
'id': '%s-%s' % (f['id'], m.group('kind').lower()),
|
||||
'title': m.group('title'),
|
||||
'url': compat_urlparse.urljoin(url, m.group('href')),
|
||||
})
|
||||
entries.append(f)
|
||||
|
||||
def extract_episodes(html):
|
||||
for num, episode_html in enumerate(re.findall(
|
||||
r'(?s)<h3[^>]+class="episodebox-title".+?>Episodeninhalt<', html), 1):
|
||||
episodebox_title = self._search_regex(
|
||||
(r'class="episodebox-title"[^>]+title=(["\'])(?P<title>.+?)\1',
|
||||
r'class="episodebox-title"[^>]+>(?P<title>.+?)<'),
|
||||
episode_html, 'episodebox title', default=None, group='title')
|
||||
if not episodebox_title:
|
||||
continue
|
||||
|
||||
episode_number = int(self._search_regex(
|
||||
r'(?:Episode|Film)\s*(\d+)',
|
||||
episodebox_title, 'episode number', default=num))
|
||||
episode_title = self._search_regex(
|
||||
r'(?:Episode|Film)\s*\d+\s*-\s*(.+)',
|
||||
episodebox_title, 'episode title', default=None)
|
||||
|
||||
video_id = 'episode-%d' % episode_number
|
||||
|
||||
common_info = {
|
||||
'id': video_id,
|
||||
'series': anime_title,
|
||||
'episode': episode_title,
|
||||
'episode_number': episode_number,
|
||||
}
|
||||
|
||||
extract_entries(episode_html, video_id, common_info)
|
||||
|
||||
def extract_film(html, video_id):
|
||||
common_info = {
|
||||
'id': anime_id,
|
||||
'title': anime_title,
|
||||
'description': anime_description,
|
||||
}
|
||||
extract_entries(html, video_id, common_info)
|
||||
|
||||
extract_episodes(webpage)
|
||||
|
||||
if not entries:
|
||||
extract_film(webpage, anime_id)
|
||||
|
||||
return self.playlist_result(entries, anime_id, anime_title, anime_description)
|
||||
|
Loading…
Reference in New Issue
Block a user