mirror of
https://github.com/morpheus65535/bazarr
synced 2024-12-23 00:03:33 +00:00
cb420628f8
* Upgraded cloudscraper to fix multiple issues with providers that uses antibot page. * Fixed subs4series provider. It now require anti-captcha provider to download subtitles. One captcha will have to be solved for each download. #1442
190 lines
5.6 KiB
Python
190 lines
5.6 KiB
Python
from __future__ import absolute_import
|
|
|
|
import requests
|
|
|
|
try:
|
|
from urlparse import urlparse
|
|
except ImportError:
|
|
from urllib.parse import urlparse
|
|
|
|
from ..exceptions import (
|
|
CaptchaServiceUnavailable,
|
|
CaptchaAPIError,
|
|
CaptchaTimeout,
|
|
CaptchaParameter,
|
|
CaptchaBadJobID
|
|
)
|
|
|
|
try:
|
|
import polling2
|
|
except ImportError:
|
|
raise ImportError("Please install the python module 'polling2' via pip")
|
|
|
|
from . import Captcha
|
|
|
|
|
|
class captchaSolver(Captcha):
|
|
|
|
def __init__(self):
|
|
super(captchaSolver, self).__init__('capmonster')
|
|
self.host = 'https://api.capmonster.cloud'
|
|
self.session = requests.Session()
|
|
|
|
# ------------------------------------------------------------------------------- #
|
|
|
|
@staticmethod
|
|
def checkErrorStatus(response):
|
|
if response.status_code in [500, 502]:
|
|
raise CaptchaServiceUnavailable(
|
|
f'CapMonster: Server Side Error {response.status_code}'
|
|
)
|
|
|
|
payload = response.json()
|
|
if payload['errorId'] == 1:
|
|
if 'errorDescription' in payload:
|
|
raise CaptchaAPIError(
|
|
payload['errorDescription']
|
|
)
|
|
else:
|
|
raise CaptchaAPIError(payload['errorCode'])
|
|
|
|
# ------------------------------------------------------------------------------- #
|
|
|
|
def requestJob(self, taskID):
|
|
if not taskID:
|
|
raise CaptchaBadJobID(
|
|
'CapMonster: Error bad task id to request Captcha.'
|
|
)
|
|
|
|
def _checkRequest(response):
|
|
self.checkErrorStatus(response)
|
|
|
|
if response.ok and response.json()['status'] == 'ready':
|
|
return True
|
|
|
|
return None
|
|
|
|
response = polling2.poll(
|
|
lambda: self.session.post(
|
|
f'{self.host}/getTaskResult',
|
|
json={
|
|
'clientKey': self.clientKey,
|
|
'taskId': taskID
|
|
},
|
|
timeout=30
|
|
),
|
|
check_success=_checkRequest,
|
|
step=5,
|
|
timeout=180
|
|
)
|
|
|
|
if response:
|
|
return response.json()['solution']['gRecaptchaResponse']
|
|
else:
|
|
raise CaptchaTimeout(
|
|
"CapMonster: Error failed to solve Captcha."
|
|
)
|
|
|
|
# ------------------------------------------------------------------------------- #
|
|
|
|
def requestSolve(self, captchaType, url, siteKey):
|
|
def _checkRequest(response):
|
|
self.checkErrorStatus(response)
|
|
|
|
if response.ok and response.json()['taskId']:
|
|
return True
|
|
|
|
return None
|
|
|
|
data = {
|
|
'clientKey': self.clientKey,
|
|
'task': {
|
|
'websiteURL': url,
|
|
'websiteKey': siteKey,
|
|
'softId': 37,
|
|
'type': 'NoCaptchaTask' if captchaType == 'reCaptcha' else 'HCaptchaTask'
|
|
}
|
|
}
|
|
|
|
if self.proxy:
|
|
data['task'].update(self.proxy)
|
|
else:
|
|
data['task']['type'] = f"{data['task']['type']}Proxyless"
|
|
|
|
response = polling2.poll(
|
|
lambda: self.session.post(
|
|
f'{self.host}/createTask',
|
|
json=data,
|
|
allow_redirects=False,
|
|
timeout=30
|
|
),
|
|
check_success=_checkRequest,
|
|
step=5,
|
|
timeout=180
|
|
)
|
|
|
|
if response:
|
|
return response.json()['taskId']
|
|
else:
|
|
raise CaptchaBadJobID(
|
|
'CapMonster: Error no task id was returned.'
|
|
)
|
|
|
|
# ------------------------------------------------------------------------------- #
|
|
|
|
def getCaptchaAnswer(self, captchaType, url, siteKey, captchaParams):
|
|
taskID = None
|
|
|
|
if not captchaParams.get('clientKey'):
|
|
raise CaptchaParameter(
|
|
"CapMonster: Missing clientKey parameter."
|
|
)
|
|
|
|
self.clientKey = captchaParams.get('clientKey')
|
|
|
|
if captchaParams.get('proxy') and not captchaParams.get('no_proxy'):
|
|
hostParsed = urlparse(captchaParams.get('proxy', {}).get('https'))
|
|
|
|
if not hostParsed.scheme:
|
|
raise CaptchaParameter('Cannot parse proxy correctly, bad scheme')
|
|
|
|
if not hostParsed.netloc:
|
|
raise CaptchaParameter('Cannot parse proxy correctly, bad netloc')
|
|
|
|
ports = {
|
|
'http': 80,
|
|
'https': 443
|
|
}
|
|
|
|
self.proxy = {
|
|
'proxyType': hostParsed.scheme,
|
|
'proxyAddress': hostParsed.hostname,
|
|
'proxyPort': hostParsed.port if hostParsed.port else ports[self.proxy['proxyType']],
|
|
'proxyLogin': hostParsed.username,
|
|
'proxyPassword': hostParsed.password,
|
|
}
|
|
else:
|
|
self.proxy = None
|
|
|
|
try:
|
|
taskID = self.requestSolve(captchaType, url, siteKey)
|
|
return self.requestJob(taskID)
|
|
except polling2.TimeoutException:
|
|
try:
|
|
if taskID:
|
|
self.reportJob(taskID)
|
|
except polling2.TimeoutException:
|
|
raise CaptchaTimeout(
|
|
"CapMonster: Captcha solve took to long and also failed "
|
|
f"reporting the task with task id {taskID}."
|
|
)
|
|
|
|
raise CaptchaTimeout(
|
|
"CapMonster: Captcha solve took to long to execute "
|
|
f"task id {taskID}, aborting."
|
|
)
|
|
|
|
|
|
# ------------------------------------------------------------------------------- #
|
|
|
|
captchaSolver()
|