|
|
|
@ -1,4 +1,8 @@
|
|
|
|
|
import base64
|
|
|
|
|
import codecs
|
|
|
|
|
import datetime
|
|
|
|
|
import hashlib
|
|
|
|
|
import hmac
|
|
|
|
|
import json
|
|
|
|
|
import re
|
|
|
|
|
|
|
|
|
@ -12,6 +16,8 @@ from ..utils import (
|
|
|
|
|
multipart_encode,
|
|
|
|
|
parse_duration,
|
|
|
|
|
random_birthday,
|
|
|
|
|
traverse_obj,
|
|
|
|
|
try_call,
|
|
|
|
|
try_get,
|
|
|
|
|
urljoin,
|
|
|
|
|
)
|
|
|
|
@ -19,7 +25,18 @@ from ..utils import (
|
|
|
|
|
|
|
|
|
|
class CDAIE(InfoExtractor):
|
|
|
|
|
_VALID_URL = r'https?://(?:(?:www\.)?cda\.pl/video|ebd\.cda\.pl/[0-9]+x[0-9]+)/(?P<id>[0-9a-z]+)'
|
|
|
|
|
_NETRC_MACHINE = 'cdapl'
|
|
|
|
|
|
|
|
|
|
_BASE_URL = 'http://www.cda.pl/'
|
|
|
|
|
_BASE_API_URL = 'https://api.cda.pl'
|
|
|
|
|
_API_HEADERS = {
|
|
|
|
|
'Accept': 'application/vnd.cda.public+json',
|
|
|
|
|
'User-Agent': 'pl.cda 1.0 (version 1.2.88 build 15306; Android 9; Xiaomi Redmi 3S)',
|
|
|
|
|
}
|
|
|
|
|
# hardcoded in the app
|
|
|
|
|
_LOGIN_REQUEST_AUTH = 'Basic YzU3YzBlZDUtYTIzOC00MWQwLWI2NjQtNmZmMWMxY2Y2YzVlOklBTm95QlhRRVR6U09MV1hnV3MwMW0xT2VyNWJNZzV4clRNTXhpNGZJUGVGZ0lWUlo5UGVYTDhtUGZaR1U1U3Q'
|
|
|
|
|
_BEARER_CACHE = 'cda-bearer'
|
|
|
|
|
|
|
|
|
|
_TESTS = [{
|
|
|
|
|
'url': 'http://www.cda.pl/video/5749950c',
|
|
|
|
|
'md5': '6f844bf51b15f31fae165365707ae970',
|
|
|
|
@ -83,8 +100,73 @@ class CDAIE(InfoExtractor):
|
|
|
|
|
'Content-Type': content_type,
|
|
|
|
|
}, **kwargs)
|
|
|
|
|
|
|
|
|
|
def _perform_login(self, username, password):
|
|
|
|
|
cached_bearer = self.cache.load(self._BEARER_CACHE, username) or {}
|
|
|
|
|
if cached_bearer.get('valid_until', 0) > datetime.datetime.now().timestamp() + 5:
|
|
|
|
|
self._API_HEADERS['Authorization'] = f'Bearer {cached_bearer["token"]}'
|
|
|
|
|
return
|
|
|
|
|
|
|
|
|
|
password_hash = base64.urlsafe_b64encode(hmac.new(
|
|
|
|
|
b's01m1Oer5IANoyBXQETzSOLWXgWs01m1Oer5bMg5xrTMMxRZ9Pi4fIPeFgIVRZ9PeXL8mPfXQETZGUAN5StRZ9P',
|
|
|
|
|
''.join(f'{bytes((bt & 255, )).hex():0>2}'
|
|
|
|
|
for bt in hashlib.md5(password.encode()).digest()).encode(),
|
|
|
|
|
hashlib.sha256).digest()).decode().replace('=', '')
|
|
|
|
|
|
|
|
|
|
token_res = self._download_json(
|
|
|
|
|
f'{self._BASE_API_URL}/oauth/token', None, 'Logging in', data=b'',
|
|
|
|
|
headers={**self._API_HEADERS, 'Authorization': self._LOGIN_REQUEST_AUTH},
|
|
|
|
|
query={
|
|
|
|
|
'grant_type': 'password',
|
|
|
|
|
'login': username,
|
|
|
|
|
'password': password_hash,
|
|
|
|
|
})
|
|
|
|
|
self.cache.store(self._BEARER_CACHE, username, {
|
|
|
|
|
'token': token_res['access_token'],
|
|
|
|
|
'valid_until': token_res['expires_in'] + datetime.datetime.now().timestamp(),
|
|
|
|
|
})
|
|
|
|
|
self._API_HEADERS['Authorization'] = f'Bearer {token_res["access_token"]}'
|
|
|
|
|
|
|
|
|
|
def _real_extract(self, url):
|
|
|
|
|
video_id = self._match_id(url)
|
|
|
|
|
|
|
|
|
|
if 'Authorization' in self._API_HEADERS:
|
|
|
|
|
return self._api_extract(video_id)
|
|
|
|
|
else:
|
|
|
|
|
return self._web_extract(video_id, url)
|
|
|
|
|
|
|
|
|
|
def _api_extract(self, video_id):
|
|
|
|
|
meta = self._download_json(
|
|
|
|
|
f'{self._BASE_API_URL}/video/{video_id}', video_id, headers=self._API_HEADERS)['video']
|
|
|
|
|
|
|
|
|
|
if meta.get('premium') and not meta.get('premium_free'):
|
|
|
|
|
self.report_drm(video_id)
|
|
|
|
|
|
|
|
|
|
uploader = traverse_obj(meta, 'author', 'login')
|
|
|
|
|
|
|
|
|
|
formats = [{
|
|
|
|
|
'url': quality['file'],
|
|
|
|
|
'format': quality.get('title'),
|
|
|
|
|
'resolution': quality.get('name'),
|
|
|
|
|
'height': try_call(lambda: int(quality['name'][:-1])),
|
|
|
|
|
'filesize': quality.get('length'),
|
|
|
|
|
} for quality in meta['qualities'] if quality.get('file')]
|
|
|
|
|
|
|
|
|
|
self._sort_formats(formats)
|
|
|
|
|
|
|
|
|
|
return {
|
|
|
|
|
'id': video_id,
|
|
|
|
|
'title': meta.get('title'),
|
|
|
|
|
'description': meta.get('description'),
|
|
|
|
|
'uploader': None if uploader == 'anonim' else uploader,
|
|
|
|
|
'average_rating': float_or_none(meta.get('rating')),
|
|
|
|
|
'thumbnail': meta.get('thumb'),
|
|
|
|
|
'formats': formats,
|
|
|
|
|
'duration': meta.get('duration'),
|
|
|
|
|
'age_limit': 18 if meta.get('for_adults') else 0,
|
|
|
|
|
'view_count': meta.get('views'),
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
def _web_extract(self, video_id, url):
|
|
|
|
|
self._set_cookie('cda.pl', 'cda.player', 'html5')
|
|
|
|
|
webpage = self._download_webpage(
|
|
|
|
|
self._BASE_URL + '/video/' + video_id, video_id)
|
|
|
|
|