mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-24 19:22:06 +00:00
[packtpub] add support for authentication(closes #12622)
This commit is contained in:
parent
8d65880e24
commit
5d0968f0af
1 changed files with 36 additions and 3 deletions
|
@ -3,7 +3,10 @@ from __future__ import unicode_literals
|
|||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..compat import compat_str
|
||||
from ..compat import (
|
||||
compat_str,
|
||||
compat_HTTPError,
|
||||
)
|
||||
from ..utils import (
|
||||
clean_html,
|
||||
ExtractorError,
|
||||
|
@ -11,6 +14,7 @@ from ..utils import (
|
|||
strip_or_none,
|
||||
unified_timestamp,
|
||||
urljoin,
|
||||
urlencode_postdata,
|
||||
)
|
||||
|
||||
|
||||
|
@ -34,6 +38,32 @@ class PacktPubIE(PacktPubBaseIE):
|
|||
'upload_date': '20170331',
|
||||
},
|
||||
}
|
||||
_NETRC_MACHINE = 'packtpub'
|
||||
_TOKEN = None
|
||||
|
||||
def _real_initialize(self):
|
||||
(username, password) = self._get_login_info()
|
||||
if username is None:
|
||||
return
|
||||
webpage = self._download_webpage(self._PACKT_BASE, None)
|
||||
login_form = self._form_hidden_inputs(
|
||||
'packt-user-login-form', webpage)
|
||||
login_form.update({
|
||||
'email': username,
|
||||
'password': password,
|
||||
})
|
||||
self._download_webpage(
|
||||
self._PACKT_BASE, None, 'Logging in as %s' % username,
|
||||
data=urlencode_postdata(login_form))
|
||||
try:
|
||||
self._TOKEN = self._download_json(
|
||||
'%s/users/tokens/sessions' % self._MAPT_REST, None,
|
||||
'Downloading Authorization Token')['data']['token']
|
||||
except ExtractorError as e:
|
||||
if isinstance(e.cause, compat_HTTPError) and e.cause.code in (401, 404):
|
||||
message = self._parse_json(e.cause.read().decode(), None)['message']
|
||||
raise ExtractorError(message, expected=True)
|
||||
raise
|
||||
|
||||
def _handle_error(self, response):
|
||||
if response.get('status') != 'success':
|
||||
|
@ -51,14 +81,17 @@ class PacktPubIE(PacktPubBaseIE):
|
|||
course_id, chapter_id, video_id = mobj.group(
|
||||
'course_id', 'chapter_id', 'id')
|
||||
|
||||
headers = {}
|
||||
if self._TOKEN:
|
||||
headers['Authorization'] = self._TOKEN
|
||||
video = self._download_json(
|
||||
'%s/users/me/products/%s/chapters/%s/sections/%s'
|
||||
% (self._MAPT_REST, course_id, chapter_id, video_id), video_id,
|
||||
'Downloading JSON video')['data']
|
||||
'Downloading JSON video', headers=headers)['data']
|
||||
|
||||
content = video.get('content')
|
||||
if not content:
|
||||
raise ExtractorError('This video is locked', expected=True)
|
||||
self.raise_login_required('This video is locked')
|
||||
|
||||
video_url = content['file']
|
||||
|
||||
|
|
Loading…
Reference in a new issue