0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-09 00:15:06 +00:00
yt-dlp/yt_dlp/extractor/gofile.py
2022-04-20 02:43:15 -07:00

106 lines
3.4 KiB
Python

import hashlib
from .common import InfoExtractor
from ..utils import (
ExtractorError,
try_get
)
class GofileIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?gofile\.io/d/(?P<id>[^/]+)'
_TESTS = [{
'url': 'https://gofile.io/d/AMZyDw',
'info_dict': {
'id': 'AMZyDw',
},
'playlist_mincount': 2,
'playlist': [{
'info_dict': {
'id': 'de571ac1-5edc-42e2-8ec2-bdac83ad4a31',
'filesize': 928116,
'ext': 'mp4',
'title': 'nuuh',
'release_timestamp': 1638338704,
'release_date': '20211201',
}
}]
}, {
'url': 'https://gofile.io/d/is8lKr',
'info_dict': {
'id': 'TMjXd9',
'ext': 'mp4',
},
'playlist_count': 0,
'skip': 'No video/audio found at provided URL.',
}, {
'url': 'https://gofile.io/d/TMjXd9',
'info_dict': {
'id': 'TMjXd9',
},
'playlist_count': 1,
}, {
'url': 'https://gofile.io/d/gqOtRf',
'info_dict': {
'id': 'gqOtRf',
},
'playlist_mincount': 1,
'params': {
'videopassword': 'password',
},
}]
_TOKEN = None
def _real_initialize(self):
token = self._get_cookies('https://gofile.io/').get('accountToken')
if token:
self._TOKEN = token.value
return
account_data = self._download_json(
'https://api.gofile.io/createAccount', None, note='Getting a new guest account')
self._TOKEN = account_data['data']['token']
self._set_cookie('gofile.io', 'accountToken', self._TOKEN)
def _entries(self, file_id):
query_params = {
'contentId': file_id,
'token': self._TOKEN,
'websiteToken': 12345,
}
password = self.get_param('videopassword')
if password:
query_params['password'] = hashlib.sha256(password.encode('utf-8')).hexdigest()
files = self._download_json(
'https://api.gofile.io/getContent', file_id, note='Getting filelist', query=query_params)
status = files['status']
if status == 'error-passwordRequired':
raise ExtractorError(
'This video is protected by a password, use the --video-password option', expected=True)
elif status != 'ok':
raise ExtractorError(f'{self.IE_NAME} said: status {status}', expected=True)
found_files = False
for file in (try_get(files, lambda x: x['data']['contents'], dict) or {}).values():
file_type, file_format = file.get('mimetype').split('/', 1)
if file_type not in ('video', 'audio') and file_format != 'vnd.mts':
continue
found_files = True
file_url = file.get('link')
if file_url:
yield {
'id': file['id'],
'title': file['name'].rsplit('.', 1)[0],
'url': file_url,
'filesize': file.get('size'),
'release_timestamp': file.get('createTime')
}
if not found_files:
raise ExtractorError('No video/audio found at provided URL.', expected=True)
def _real_extract(self, url):
file_id = self._match_id(url)
return self.playlist_result(self._entries(file_id), playlist_id=file_id)