mirror of
https://codeberg.org/polarisfm/youtube-dl
synced 2024-11-29 19:47:54 +01:00
[ninateka] Add new extractor
This commit is contained in:
parent
ce7db64bf1
commit
725ab1196d
@ -729,6 +729,7 @@ from .nick import (
|
|||||||
NickRuIE,
|
NickRuIE,
|
||||||
)
|
)
|
||||||
from .niconico import NiconicoIE, NiconicoPlaylistIE
|
from .niconico import NiconicoIE, NiconicoPlaylistIE
|
||||||
|
from .ninateka import NinatekaIE
|
||||||
from .ninecninemedia import NineCNineMediaIE
|
from .ninecninemedia import NineCNineMediaIE
|
||||||
from .ninegag import NineGagIE
|
from .ninegag import NineGagIE
|
||||||
from .ninenow import NineNowIE
|
from .ninenow import NineNowIE
|
||||||
|
92
youtube_dl/extractor/ninateka.py
Normal file
92
youtube_dl/extractor/ninateka.py
Normal file
@ -0,0 +1,92 @@
|
|||||||
|
# coding: utf-8
|
||||||
|
from __future__ import unicode_literals
|
||||||
|
|
||||||
|
from .common import InfoExtractor
|
||||||
|
from ..utils import (
|
||||||
|
determine_ext,
|
||||||
|
js_to_json,
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
|
class NinatekaIE(InfoExtractor):
|
||||||
|
IE_NAME = 'ninateka'
|
||||||
|
IE_DESC = 'Ninateka'
|
||||||
|
_VALID_URL = r'https?://ninateka\.pl/film/(?P<id>[^/\?#]+)'
|
||||||
|
_TEST = {
|
||||||
|
'url': 'https://ninateka.pl/film/dziwne-przygody-kota-filemona-7',
|
||||||
|
'md5': 'f39eebfad3a609df9c90a45a3155393d',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'dziwne-przygody-kota-filemona-7',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'title': 'Dziwny świat kota Filemona | Poważne zmartwienie',
|
||||||
|
'description': 'Filemon ma kłopot z własnym wyglądem, czy uda mu się z nim uporać?',
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
def decode_url(self, encoded):
|
||||||
|
xor_val = ord('h') ^ ord(encoded[0])
|
||||||
|
return ''.join(chr(ord(c) ^ xor_val) for c in encoded)
|
||||||
|
|
||||||
|
def extract_formats(self, data, video_id, name):
|
||||||
|
info = self._parse_json(data, video_id, transform_source=js_to_json)
|
||||||
|
formats = []
|
||||||
|
|
||||||
|
for source_info in info['sources']:
|
||||||
|
url = self.decode_url(source_info['src'])
|
||||||
|
type_ = source_info.get('type')
|
||||||
|
|
||||||
|
if type_ == 'application/vnd.ms-sstr+xml' or url.endswith('/Manifest'):
|
||||||
|
formats.extend(self._extract_ism_formats(
|
||||||
|
url, video_id, ism_id='mss-{}'.format(name), fatal=False))
|
||||||
|
|
||||||
|
elif type_ == 'application/x-mpegURL' or url.endswith('.m3u8'):
|
||||||
|
formats.extend(self._extract_m3u8_formats(
|
||||||
|
url, video_id, ext='mp4', m3u8_id='hls-{}'.format(name), fatal=False))
|
||||||
|
|
||||||
|
elif type_ == 'application/dash+xml' or url.endswith('.mpd'):
|
||||||
|
formats.extend(self._extract_mpd_formats(
|
||||||
|
url, video_id, mpd_id='dash-{}'.format(name), fatal=False))
|
||||||
|
|
||||||
|
elif url.endswith('.f4m'):
|
||||||
|
formats.extend(self._extract_f4m_formats(
|
||||||
|
url, video_id, f4m_id='hds-{}'.format(name), fatal=False))
|
||||||
|
|
||||||
|
else:
|
||||||
|
formats.append({
|
||||||
|
'format_id': 'direct-{}'.format(name),
|
||||||
|
'url': url,
|
||||||
|
'ext': determine_ext(url, 'mp4'),
|
||||||
|
})
|
||||||
|
|
||||||
|
return formats
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
video_id = self._match_id(url)
|
||||||
|
webpage = self._download_webpage(url, video_id)
|
||||||
|
|
||||||
|
main = self._search_regex(
|
||||||
|
r'(?m)(?:var|let|const)\s+playerOptionsWithMainSource\s*=\s*(\{.*?\})\s*;\s*?$',
|
||||||
|
webpage, 'main source')
|
||||||
|
formats = self.extract_formats(main, video_id, 'main')
|
||||||
|
|
||||||
|
audiodesc = self._search_regex(
|
||||||
|
r'(?m)(?:var|let|const)\s+playerOptionsWithAudioDescriptionSource\s*=\s*(\{.*?\})\s*;\s*?$',
|
||||||
|
webpage, 'audio description', default=None)
|
||||||
|
if audiodesc:
|
||||||
|
formats.extend(self.extract_formats(audiodesc, video_id, 'audiodescription'))
|
||||||
|
|
||||||
|
english_ver = self._search_regex(
|
||||||
|
r'(?m)(?:var|let|const)\s+playerOptionsWithEnglishVersion\s*=\s*(\{.*?\})\s*;\s*?$',
|
||||||
|
webpage, 'english version', default=None)
|
||||||
|
if english_ver:
|
||||||
|
formats.extend(self.extract_formats(english_ver, video_id, 'english'))
|
||||||
|
|
||||||
|
self._sort_formats(formats)
|
||||||
|
|
||||||
|
return {
|
||||||
|
'id': video_id,
|
||||||
|
'title': self._og_search_title(webpage),
|
||||||
|
'formats': formats,
|
||||||
|
'description': self._og_search_description(webpage),
|
||||||
|
'thumbnail': self._og_search_thumbnail(webpage),
|
||||||
|
}
|
Loading…
Reference in New Issue
Block a user