[animeondemand] Improve and modernize

This commit is contained in:
Sergey M․ 2017-09-11 04:32:35 +07:00
parent 018cc61549
commit 2f483758bc
No known key found for this signature in database
GPG Key ID: 2C393E0F18A9236D

View File

@ -3,16 +3,13 @@
import re
from .common import InfoExtractor
from ..compat import (
compat_urlparse,
compat_str,
)
from ..compat import compat_str
from ..utils import (
determine_ext,
extract_attributes,
ExtractorError,
sanitized_Request,
urlencode_postdata,
urljoin,
)
@ -78,14 +75,13 @@ def _login(self):
'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(login_form))
request.add_header('Referer', self._LOGIN_URL)
post_url = urljoin(self._LOGIN_URL, post_url)
response = self._download_webpage(
request, None, 'Logging in as %s' % username)
post_url, None, 'Logging in as %s' % username,
data=urlencode_postdata(login_form), headers={
'Referer': self._LOGIN_URL,
})
if all(p not in response for p in ('>Logout<', 'href="/users/sign_out"')):
error = self._search_regex(
@ -154,17 +150,19 @@ def extract_info(html, video_id, num=None):
format_id_list.append(compat_str(num))
format_id = '-'.join(format_id_list)
format_note = ', '.join(filter(None, (kind, lang_note)))
request = sanitized_Request(
compat_urlparse.urljoin(url, playlist_url),
item_id_list = []
if format_id:
item_id_list.append(format_id)
item_id_list.append('videomaterial')
playlist = self._download_json(
urljoin(url, playlist_url), video_id,
'Downloading %s JSON' % ' '.join(item_id_list),
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 %s playlist JSON' % format_id,
fatal=False)
}, fatal=False)
if not playlist:
continue
stream_url = playlist.get('streamurl')
@ -246,7 +244,7 @@ def extract_entries(html, video_id, common_info, num=None):
f.update({
'id': '%s-%s' % (f['id'], m.group('kind').lower()),
'title': m.group('title'),
'url': compat_urlparse.urljoin(url, m.group('href')),
'url': urljoin(url, m.group('href')),
})
entries.append(f)