iyas/saya-search
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

first commit

Browse Source
This commit is contained in:
Iyas Altawil
2025-06-26 15:38:10 +03:30
commit e928faf6d2
899 changed files with 403713 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

266
searx/network/__init__.py Normal file
View File

@@ -0,0 +1,266 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# pylint: disable=missing-module-docstring, global-statement
import asyncio
import threading
import concurrent.futures
from queue import SimpleQueue
from types import MethodType
from timeit import default_timer
from typing import Iterable, NamedTuple, Tuple, List, Dict, Union
from contextlib import contextmanager
import httpx
import anyio
from searx.extended_types import SXNG_Response
from .network import get_network, initialize, check_network_configuration # pylint:disable=cyclic-import
from .client import get_loop
from .raise_for_httperror import raise_for_httperror
THREADLOCAL = threading.local()
"""Thread-local data is data for thread specific values."""
def reset_time_for_thread():
THREADLOCAL.total_time = 0
def get_time_for_thread():
"""returns thread's total time or None"""
return THREADLOCAL.__dict__.get('total_time')
def set_timeout_for_thread(timeout, start_time=None):
THREADLOCAL.timeout = timeout
THREADLOCAL.start_time = start_time
def set_context_network_name(network_name):
THREADLOCAL.network = get_network(network_name)
def get_context_network():
"""If set return thread's network.
If unset, return value from :py:obj:`get_network`.
"""
return THREADLOCAL.__dict__.get('network') or get_network()
@contextmanager
def _record_http_time():
# pylint: disable=too-many-branches
time_before_request = default_timer()
start_time = getattr(THREADLOCAL, 'start_time', time_before_request)
try:
yield start_time
finally:
# update total_time.
# See get_time_for_thread() and reset_time_for_thread()
if hasattr(THREADLOCAL, 'total_time'):
time_after_request = default_timer()
THREADLOCAL.total_time += time_after_request - time_before_request
def _get_timeout(start_time, kwargs):
# pylint: disable=too-many-branches
# timeout (httpx)
if 'timeout' in kwargs:
timeout = kwargs['timeout']
else:
timeout = getattr(THREADLOCAL, 'timeout', None)
if timeout is not None:
kwargs['timeout'] = timeout
# 2 minutes timeout for the requests without timeout
timeout = timeout or 120
# adjust actual timeout
timeout += 0.2 # overhead
if start_time:
timeout -= default_timer() - start_time
return timeout
def request(method, url, **kwargs) -> SXNG_Response:
"""same as requests/requests/api.py request(...)"""
with _record_http_time() as start_time:
network = get_context_network()
timeout = _get_timeout(start_time, kwargs)
future = asyncio.run_coroutine_threadsafe(network.request(method, url, **kwargs), get_loop())
try:
return future.result(timeout)
except concurrent.futures.TimeoutError as e:
raise httpx.TimeoutException('Timeout', request=None) from e
def multi_requests(request_list: List["Request"]) -> List[Union[httpx.Response, Exception]]:
"""send multiple HTTP requests in parallel. Wait for all requests to finish."""
with _record_http_time() as start_time:
# send the requests
network = get_context_network()
loop = get_loop()
future_list = []
for request_desc in request_list:
timeout = _get_timeout(start_time, request_desc.kwargs)
future = asyncio.run_coroutine_threadsafe(
network.request(request_desc.method, request_desc.url, **request_desc.kwargs), loop
)
future_list.append((future, timeout))
# read the responses
responses = []
for future, timeout in future_list:
try:
responses.append(future.result(timeout))
except concurrent.futures.TimeoutError:
responses.append(httpx.TimeoutException('Timeout', request=None))
except Exception as e: # pylint: disable=broad-except
responses.append(e)
return responses
class Request(NamedTuple):
"""Request description for the multi_requests function"""
method: str
url: str
kwargs: Dict[str, str] = {}
@staticmethod
def get(url, **kwargs):
return Request('GET', url, kwargs)
@staticmethod
def options(url, **kwargs):
return Request('OPTIONS', url, kwargs)
@staticmethod
def head(url, **kwargs):
return Request('HEAD', url, kwargs)
@staticmethod
def post(url, **kwargs):
return Request('POST', url, kwargs)
@staticmethod
def put(url, **kwargs):
return Request('PUT', url, kwargs)
@staticmethod
def patch(url, **kwargs):
return Request('PATCH', url, kwargs)
@staticmethod
def delete(url, **kwargs):
return Request('DELETE', url, kwargs)
def get(url, **kwargs) -> SXNG_Response:
kwargs.setdefault('allow_redirects', True)
return request('get', url, **kwargs)
def options(url, **kwargs) -> SXNG_Response:
kwargs.setdefault('allow_redirects', True)
return request('options', url, **kwargs)
def head(url, **kwargs) -> SXNG_Response:
kwargs.setdefault('allow_redirects', False)
return request('head', url, **kwargs)
def post(url, data=None, **kwargs) -> SXNG_Response:
return request('post', url, data=data, **kwargs)
def put(url, data=None, **kwargs) -> SXNG_Response:
return request('put', url, data=data, **kwargs)
def patch(url, data=None, **kwargs) -> SXNG_Response:
return request('patch', url, data=data, **kwargs)
def delete(url, **kwargs) -> SXNG_Response:
return request('delete', url, **kwargs)
async def stream_chunk_to_queue(network, queue, method, url, **kwargs):
try:
async with await network.stream(method, url, **kwargs) as response:
queue.put(response)
# aiter_raw: access the raw bytes on the response without applying any HTTP content decoding
# https://www.python-httpx.org/quickstart/#streaming-responses
async for chunk in response.aiter_raw(65536):
if len(chunk) > 0:
queue.put(chunk)
except (httpx.StreamClosed, anyio.ClosedResourceError):
# the response was queued before the exception.
# the exception was raised on aiter_raw.
# we do nothing here: in the finally block, None will be queued
# so stream(method, url, **kwargs) generator can stop
pass
except Exception as e: # pylint: disable=broad-except
# broad except to avoid this scenario:
# exception in network.stream(method, url, **kwargs)
# -> the exception is not catch here
# -> queue None (in finally)
# -> the function below steam(method, url, **kwargs) has nothing to return
queue.put(e)
finally:
queue.put(None)
def _stream_generator(method, url, **kwargs):
queue = SimpleQueue()
network = get_context_network()
future = asyncio.run_coroutine_threadsafe(stream_chunk_to_queue(network, queue, method, url, **kwargs), get_loop())
# yield chunks
obj_or_exception = queue.get()
while obj_or_exception is not None:
if isinstance(obj_or_exception, Exception):
raise obj_or_exception
yield obj_or_exception
obj_or_exception = queue.get()
future.result()
def _close_response_method(self):
asyncio.run_coroutine_threadsafe(self.aclose(), get_loop())
# reach the end of _self.generator ( _stream_generator ) to an avoid memory leak.
# it makes sure that :
# * the httpx response is closed (see the stream_chunk_to_queue function)
# * to call future.result() in _stream_generator
for _ in self._generator: # pylint: disable=protected-access
continue
def stream(method, url, **kwargs) -> Tuple[httpx.Response, Iterable[bytes]]:
"""Replace httpx.stream.
Usage:
response, stream = poolrequests.stream(...)
for chunk in stream:
...
httpx.Client.stream requires to write the httpx.HTTPTransport version of the
the httpx.AsyncHTTPTransport declared above.
"""
generator = _stream_generator(method, url, **kwargs)
# yield response
response = next(generator) # pylint: disable=stop-iteration-return
if isinstance(response, Exception):
raise response
response._generator = generator # pylint: disable=protected-access
response.close = MethodType(_close_response_method, response)
return response, generator

232
searx/network/client.py Normal file
View File

@@ -0,0 +1,232 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# pylint: disable=missing-module-docstring, global-statement
import asyncio
import logging
import random
from ssl import SSLContext
import threading
from typing import Any, Dict
import httpx
from httpx_socks import AsyncProxyTransport
from python_socks import parse_proxy_url, ProxyConnectionError, ProxyTimeoutError, ProxyError
import uvloop
from searx import logger
uvloop.install()
logger = logger.getChild('searx.network.client')
LOOP = None
SSLCONTEXTS: Dict[Any, SSLContext] = {}
def shuffle_ciphers(ssl_context: SSLContext):
"""Shuffle httpx's default ciphers of a SSL context randomly.
From `What Is TLS Fingerprint and How to Bypass It`_
> When implementing TLS fingerprinting, servers can't operate based on a
> locked-in whitelist database of fingerprints. New fingerprints appear
> when web clients or TLS libraries release new versions. So, they have to
> live off a blocklist database instead.
> ...
> It's safe to leave the first three as is but shuffle the remaining ciphers
> and you can bypass the TLS fingerprint check.
.. _What Is TLS Fingerprint and How to Bypass It:
https://www.zenrows.com/blog/what-is-tls-fingerprint#how-to-bypass-tls-fingerprinting
"""
c_list = [cipher["name"] for cipher in ssl_context.get_ciphers()]
sc_list, c_list = c_list[:3], c_list[3:]
random.shuffle(c_list)
ssl_context.set_ciphers(":".join(sc_list + c_list))
def get_sslcontexts(proxy_url=None, cert=None, verify=True, trust_env=True):
key = (proxy_url, cert, verify, trust_env)
if key not in SSLCONTEXTS:
SSLCONTEXTS[key] = httpx.create_ssl_context(verify, cert, trust_env)
shuffle_ciphers(SSLCONTEXTS[key])
return SSLCONTEXTS[key]
class AsyncHTTPTransportNoHttp(httpx.AsyncHTTPTransport):
"""Block HTTP request
The constructor is blank because httpx.AsyncHTTPTransport.__init__ creates an SSLContext unconditionally:
https://github.com/encode/httpx/blob/0f61aa58d66680c239ce43c8cdd453e7dc532bfc/httpx/_transports/default.py#L271
Each SSLContext consumes more than 500kb of memory, since there is about one network per engine.
In consequence, this class overrides all public methods
For reference: https://github.com/encode/httpx/issues/2298
"""
def __init__(self, *args, **kwargs):
# pylint: disable=super-init-not-called
# this on purpose if the base class is not called
pass
async def handle_async_request(self, request):
raise httpx.UnsupportedProtocol('HTTP protocol is disabled')
async def aclose(self) -> None:
pass
async def __aenter__(self):
return self
async def __aexit__(
self,
exc_type=None,
exc_value=None,
traceback=None,
) -> None:
pass
class AsyncProxyTransportFixed(AsyncProxyTransport):
"""Fix httpx_socks.AsyncProxyTransport
Map python_socks exceptions to httpx.ProxyError exceptions
"""
async def handle_async_request(self, request):
try:
return await super().handle_async_request(request)
except ProxyConnectionError as e:
raise httpx.ProxyError("ProxyConnectionError: " + e.strerror, request=request) from e
except ProxyTimeoutError as e:
raise httpx.ProxyError("ProxyTimeoutError: " + e.args[0], request=request) from e
except ProxyError as e:
raise httpx.ProxyError("ProxyError: " + e.args[0], request=request) from e
def get_transport_for_socks_proxy(verify, http2, local_address, proxy_url, limit, retries):
# support socks5h (requests compatibility):
# https://requests.readthedocs.io/en/master/user/advanced/#socks
# socks5:// hostname is resolved on client side
# socks5h:// hostname is resolved on proxy side
rdns = False
socks5h = 'socks5h://'
if proxy_url.startswith(socks5h):
proxy_url = 'socks5://' + proxy_url[len(socks5h) :]
rdns = True
proxy_type, proxy_host, proxy_port, proxy_username, proxy_password = parse_proxy_url(proxy_url)
verify = get_sslcontexts(proxy_url, None, verify, True) if verify is True else verify
return AsyncProxyTransportFixed(
proxy_type=proxy_type,
proxy_host=proxy_host,
proxy_port=proxy_port,
username=proxy_username,
password=proxy_password,
rdns=rdns,
loop=get_loop(),
verify=verify,
http2=http2,
local_address=local_address,
limits=limit,
retries=retries,
)
def get_transport(verify, http2, local_address, proxy_url, limit, retries):
verify = get_sslcontexts(None, None, verify, True) if verify is True else verify
return httpx.AsyncHTTPTransport(
# pylint: disable=protected-access
verify=verify,
http2=http2,
limits=limit,
proxy=httpx._config.Proxy(proxy_url) if proxy_url else None,
local_address=local_address,
retries=retries,
)
def new_client(
# pylint: disable=too-many-arguments
enable_http,
verify,
enable_http2,
max_connections,
max_keepalive_connections,
keepalive_expiry,
proxies,
local_address,
retries,
max_redirects,
hook_log_response,
):
limit = httpx.Limits(
max_connections=max_connections,
max_keepalive_connections=max_keepalive_connections,
keepalive_expiry=keepalive_expiry,
)
# See https://www.python-httpx.org/advanced/#routing
mounts = {}
for pattern, proxy_url in proxies.items():
if not enable_http and pattern.startswith('http://'):
continue
if proxy_url.startswith('socks4://') or proxy_url.startswith('socks5://') or proxy_url.startswith('socks5h://'):
mounts[pattern] = get_transport_for_socks_proxy(
verify, enable_http2, local_address, proxy_url, limit, retries
)
else:
mounts[pattern] = get_transport(verify, enable_http2, local_address, proxy_url, limit, retries)
if not enable_http:
mounts['http://'] = AsyncHTTPTransportNoHttp()
transport = get_transport(verify, enable_http2, local_address, None, limit, retries)
event_hooks = None
if hook_log_response:
event_hooks = {'response': [hook_log_response]}
return httpx.AsyncClient(
transport=transport,
mounts=mounts,
max_redirects=max_redirects,
event_hooks=event_hooks,
)
def get_loop():
return LOOP
def init():
# log
for logger_name in (
'httpx',
'httpcore.proxy',
'httpcore.connection',
'httpcore.http11',
'httpcore.http2',
'hpack.hpack',
'hpack.table',
):
logging.getLogger(logger_name).setLevel(logging.WARNING)
# loop
def loop_thread():
global LOOP
LOOP = asyncio.new_event_loop()
LOOP.run_forever()
thread = threading.Thread(
target=loop_thread,
name='asyncio_loop',
daemon=True,
)
thread.start()
init()

433
searx/network/network.py Normal file
View File

@@ -0,0 +1,433 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
# pylint: disable=global-statement
# pylint: disable=missing-module-docstring, missing-class-docstring
from __future__ import annotations
import typing
import atexit
import asyncio
import ipaddress
from itertools import cycle
from typing import Dict
import httpx
from searx import logger, sxng_debug
from searx.extended_types import SXNG_Response
from .client import new_client, get_loop, AsyncHTTPTransportNoHttp
from .raise_for_httperror import raise_for_httperror
logger = logger.getChild('network')
DEFAULT_NAME = '__DEFAULT__'
NETWORKS: Dict[str, 'Network'] = {}
# requests compatibility when reading proxy settings from settings.yml
PROXY_PATTERN_MAPPING = {
'http': 'http://',
'https': 'https://',
'socks4': 'socks4://',
'socks5': 'socks5://',
'socks5h': 'socks5h://',
'http:': 'http://',
'https:': 'https://',
'socks4:': 'socks4://',
'socks5:': 'socks5://',
'socks5h:': 'socks5h://',
}
ADDRESS_MAPPING = {'ipv4': '0.0.0.0', 'ipv6': '::'}
class Network:
__slots__ = (
'enable_http',
'verify',
'enable_http2',
'max_connections',
'max_keepalive_connections',
'keepalive_expiry',
'local_addresses',
'proxies',
'using_tor_proxy',
'max_redirects',
'retries',
'retry_on_http_error',
'_local_addresses_cycle',
'_proxies_cycle',
'_clients',
'_logger',
)
_TOR_CHECK_RESULT = {}
def __init__(
# pylint: disable=too-many-arguments
self,
enable_http=True,
verify=True,
enable_http2=False,
max_connections=None,
max_keepalive_connections=None,
keepalive_expiry=None,
proxies=None,
using_tor_proxy=False,
local_addresses=None,
retries=0,
retry_on_http_error=None,
max_redirects=30,
logger_name=None,
):
self.enable_http = enable_http
self.verify = verify
self.enable_http2 = enable_http2
self.max_connections = max_connections
self.max_keepalive_connections = max_keepalive_connections
self.keepalive_expiry = keepalive_expiry
self.proxies = proxies
self.using_tor_proxy = using_tor_proxy
self.local_addresses = local_addresses
self.retries = retries
self.retry_on_http_error = retry_on_http_error
self.max_redirects = max_redirects
self._local_addresses_cycle = self.get_ipaddress_cycle()
self._proxies_cycle = self.get_proxy_cycles()
self._clients = {}
self._logger = logger.getChild(logger_name) if logger_name else logger
self.check_parameters()
def check_parameters(self):
for address in self.iter_ipaddresses():
if '/' in address:
ipaddress.ip_network(address, False)
else:
ipaddress.ip_address(address)
if self.proxies is not None and not isinstance(self.proxies, (str, dict)):
raise ValueError('proxies type has to be str, dict or None')
def iter_ipaddresses(self):
local_addresses = self.local_addresses
if not local_addresses:
return
if isinstance(local_addresses, str):
local_addresses = [local_addresses]
yield from local_addresses
def get_ipaddress_cycle(self):
while True:
count = 0
for address in self.iter_ipaddresses():
if '/' in address:
for a in ipaddress.ip_network(address, False).hosts():
yield str(a)
count += 1
else:
a = ipaddress.ip_address(address)
yield str(a)
count += 1
if count == 0:
yield None
def iter_proxies(self):
if not self.proxies:
return
# https://www.python-httpx.org/compatibility/#proxy-keys
if isinstance(self.proxies, str):
yield 'all://', [self.proxies]
else:
for pattern, proxy_url in self.proxies.items():
pattern = PROXY_PATTERN_MAPPING.get(pattern, pattern)
if isinstance(proxy_url, str):
proxy_url = [proxy_url]
yield pattern, proxy_url
def get_proxy_cycles(self):
proxy_settings = {}
for pattern, proxy_urls in self.iter_proxies():
proxy_settings[pattern] = cycle(proxy_urls)
while True:
# pylint: disable=stop-iteration-return
yield tuple((pattern, next(proxy_url_cycle)) for pattern, proxy_url_cycle in proxy_settings.items())
async def log_response(self, response: httpx.Response):
request = response.request
status = f"{response.status_code} {response.reason_phrase}"
response_line = f"{response.http_version} {status}"
content_type = response.headers.get("Content-Type")
content_type = f' ({content_type})' if content_type else ''
self._logger.debug(f'HTTP Request: {request.method} {request.url} "{response_line}"{content_type}')
@staticmethod
async def check_tor_proxy(client: httpx.AsyncClient, proxies) -> bool:
if proxies in Network._TOR_CHECK_RESULT:
return Network._TOR_CHECK_RESULT[proxies]
result = True
# ignore client._transport because it is not used with all://
for transport in client._mounts.values(): # pylint: disable=protected-access
if isinstance(transport, AsyncHTTPTransportNoHttp):
continue
if getattr(transport, "_pool") and getattr(
transport._pool, "_rdns", False # pylint: disable=protected-access
):
continue
return False
response = await client.get("https://check.torproject.org/api/ip", timeout=60)
if not response.json()["IsTor"]:
result = False
Network._TOR_CHECK_RESULT[proxies] = result
return result
async def get_client(self, verify=None, max_redirects=None) -> httpx.AsyncClient:
verify = self.verify if verify is None else verify
max_redirects = self.max_redirects if max_redirects is None else max_redirects
local_address = next(self._local_addresses_cycle)
proxies = next(self._proxies_cycle) # is a tuple so it can be part of the key
key = (verify, max_redirects, local_address, proxies)
hook_log_response = self.log_response if sxng_debug else None
if key not in self._clients or self._clients[key].is_closed:
client = new_client(
self.enable_http,
verify,
self.enable_http2,
self.max_connections,
self.max_keepalive_connections,
self.keepalive_expiry,
dict(proxies),
local_address,
0,
max_redirects,
hook_log_response,
)
if self.using_tor_proxy and not await self.check_tor_proxy(client, proxies):
await client.aclose()
raise httpx.ProxyError('Network configuration problem: not using Tor')
self._clients[key] = client
return self._clients[key]
async def aclose(self):
async def close_client(client):
try:
await client.aclose()
except httpx.HTTPError:
pass
await asyncio.gather(*[close_client(client) for client in self._clients.values()], return_exceptions=False)
@staticmethod
def extract_kwargs_clients(kwargs):
kwargs_clients = {}
if 'verify' in kwargs:
kwargs_clients['verify'] = kwargs.pop('verify')
if 'max_redirects' in kwargs:
kwargs_clients['max_redirects'] = kwargs.pop('max_redirects')
if 'allow_redirects' in kwargs:
# see https://github.com/encode/httpx/pull/1808
kwargs['follow_redirects'] = kwargs.pop('allow_redirects')
return kwargs_clients
@staticmethod
def extract_do_raise_for_httperror(kwargs):
do_raise_for_httperror = True
if 'raise_for_httperror' in kwargs:
do_raise_for_httperror = kwargs['raise_for_httperror']
del kwargs['raise_for_httperror']
return do_raise_for_httperror
def patch_response(self, response, do_raise_for_httperror) -> SXNG_Response:
if isinstance(response, httpx.Response):
response = typing.cast(SXNG_Response, response)
# requests compatibility (response is not streamed)
# see also https://www.python-httpx.org/compatibility/#checking-for-4xx5xx-responses
response.ok = not response.is_error
# raise an exception
if do_raise_for_httperror:
try:
raise_for_httperror(response)
except:
self._logger.warning(f"HTTP Request failed: {response.request.method} {response.request.url}")
raise
return response
def is_valid_response(self, response):
# pylint: disable=too-many-boolean-expressions
if (
(self.retry_on_http_error is True and 400 <= response.status_code <= 599)
or (isinstance(self.retry_on_http_error, list) and response.status_code in self.retry_on_http_error)
or (isinstance(self.retry_on_http_error, int) and response.status_code == self.retry_on_http_error)
):
return False
return True
async def call_client(self, stream, method, url, **kwargs) -> SXNG_Response:
retries = self.retries
was_disconnected = False
do_raise_for_httperror = Network.extract_do_raise_for_httperror(kwargs)
kwargs_clients = Network.extract_kwargs_clients(kwargs)
while retries >= 0: # pragma: no cover
client = await self.get_client(**kwargs_clients)
cookies = kwargs.pop("cookies", None)
client.cookies = httpx.Cookies(cookies)
try:
if stream:
response = client.stream(method, url, **kwargs)
else:
response = await client.request(method, url, **kwargs)
if self.is_valid_response(response) or retries <= 0:
return self.patch_response(response, do_raise_for_httperror)
except httpx.RemoteProtocolError as e:
if not was_disconnected:
# the server has closed the connection:
# try again without decreasing the retries variable & with a new HTTP client
was_disconnected = True
await client.aclose()
self._logger.warning('httpx.RemoteProtocolError: the server has disconnected, retrying')
continue
if retries <= 0:
raise e
except (httpx.RequestError, httpx.HTTPStatusError) as e:
if retries <= 0:
raise e
retries -= 1
async def request(self, method, url, **kwargs):
return await self.call_client(False, method, url, **kwargs)
async def stream(self, method, url, **kwargs):
return await self.call_client(True, method, url, **kwargs)
@classmethod
async def aclose_all(cls):
await asyncio.gather(*[network.aclose() for network in NETWORKS.values()], return_exceptions=False)
def get_network(name=None):
return NETWORKS.get(name or DEFAULT_NAME)
def check_network_configuration():
async def check():
exception_count = 0
for network in NETWORKS.values():
if network.using_tor_proxy:
try:
await network.get_client()
except Exception: # pylint: disable=broad-except
network._logger.exception('Error') # pylint: disable=protected-access
exception_count += 1
return exception_count
future = asyncio.run_coroutine_threadsafe(check(), get_loop())
exception_count = future.result()
if exception_count > 0:
raise RuntimeError("Invalid network configuration")
def initialize(settings_engines=None, settings_outgoing=None):
# pylint: disable=import-outside-toplevel)
from searx.engines import engines
from searx import settings
# pylint: enable=import-outside-toplevel)
settings_engines = settings_engines or settings['engines']
settings_outgoing = settings_outgoing or settings['outgoing']
# default parameters for AsyncHTTPTransport
# see https://github.com/encode/httpx/blob/e05a5372eb6172287458b37447c30f650047e1b8/httpx/_transports/default.py#L108-L121 # pylint: disable=line-too-long
default_params = {
'enable_http': False,
'verify': settings_outgoing['verify'],
'enable_http2': settings_outgoing['enable_http2'],
'max_connections': settings_outgoing['pool_connections'],
'max_keepalive_connections': settings_outgoing['pool_maxsize'],
'keepalive_expiry': settings_outgoing['keepalive_expiry'],
'local_addresses': settings_outgoing['source_ips'],
'using_tor_proxy': settings_outgoing['using_tor_proxy'],
'proxies': settings_outgoing['proxies'],
'max_redirects': settings_outgoing['max_redirects'],
'retries': settings_outgoing['retries'],
'retry_on_http_error': None,
}
def new_network(params, logger_name=None):
nonlocal default_params
result = {}
result.update(default_params)
result.update(params)
if logger_name:
result['logger_name'] = logger_name
return Network(**result)
def iter_networks():
nonlocal settings_engines
for engine_spec in settings_engines:
engine_name = engine_spec['name']
engine = engines.get(engine_name)
if engine is None:
continue
network = getattr(engine, 'network', None)
yield engine_name, engine, network
if NETWORKS:
done()
NETWORKS.clear()
NETWORKS[DEFAULT_NAME] = new_network({}, logger_name='default')
NETWORKS['ipv4'] = new_network({'local_addresses': '0.0.0.0'}, logger_name='ipv4')
NETWORKS['ipv6'] = new_network({'local_addresses': '::'}, logger_name='ipv6')
# define networks from outgoing.networks
for network_name, network in settings_outgoing['networks'].items():
NETWORKS[network_name] = new_network(network, logger_name=network_name)
# define networks from engines.[i].network (except references)
for engine_name, engine, network in iter_networks():
if network is None:
network = {}
for attribute_name, attribute_value in default_params.items():
if hasattr(engine, attribute_name):
network[attribute_name] = getattr(engine, attribute_name)
else:
network[attribute_name] = attribute_value
NETWORKS[engine_name] = new_network(network, logger_name=engine_name)
elif isinstance(network, dict):
NETWORKS[engine_name] = new_network(network, logger_name=engine_name)
# define networks from engines.[i].network (references)
for engine_name, engine, network in iter_networks():
if isinstance(network, str):
NETWORKS[engine_name] = NETWORKS[network]
# the /image_proxy endpoint has a dedicated network.
# same parameters than the default network, but HTTP/2 is disabled.
# It decreases the CPU load average, and the total time is more or less the same
if 'image_proxy' not in NETWORKS:
image_proxy_params = default_params.copy()
image_proxy_params['enable_http2'] = False
NETWORKS['image_proxy'] = new_network(image_proxy_params, logger_name='image_proxy')
@atexit.register
def done():
"""Close all HTTP client
Avoid a warning at exit
See https://github.com/encode/httpx/pull/2026
Note: since Network.aclose has to be async, it is not possible to call this method on Network.__del__
So Network.aclose is called here using atexit.register
"""
try:
loop = get_loop()
if loop:
future = asyncio.run_coroutine_threadsafe(Network.aclose_all(), loop)
# wait 3 seconds to close the HTTP clients
future.result(3)
finally:
NETWORKS.clear()
NETWORKS[DEFAULT_NAME] = Network()

77
searx/network/raise_for_httperror.py Normal file
View File

@@ -0,0 +1,77 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
"""Raise exception for an HTTP response is an error.
"""
from searx.exceptions import (
SearxEngineCaptchaException,
SearxEngineTooManyRequestsException,
SearxEngineAccessDeniedException,
)
from searx import get_setting
def is_cloudflare_challenge(resp):
if resp.status_code in [429, 503]:
if ('__cf_chl_jschl_tk__=' in resp.text) or (
'/cdn-cgi/challenge-platform/' in resp.text
and 'orchestrate/jsch/v1' in resp.text
and 'window._cf_chl_enter(' in resp.text
):
return True
if resp.status_code == 403 and '__cf_chl_captcha_tk__=' in resp.text:
return True
return False
def is_cloudflare_firewall(resp):
return resp.status_code == 403 and '<span class="cf-error-code">1020</span>' in resp.text
def raise_for_cloudflare_captcha(resp):
if resp.headers.get('Server', '').startswith('cloudflare'):
if is_cloudflare_challenge(resp):
# https://support.cloudflare.com/hc/en-us/articles/200170136-Understanding-Cloudflare-Challenge-Passage-Captcha-
# suspend for 2 weeks
raise SearxEngineCaptchaException(
message='Cloudflare CAPTCHA', suspended_time=get_setting('search.suspended_times.cf_SearxEngineCaptcha')
)
if is_cloudflare_firewall(resp):
raise SearxEngineAccessDeniedException(
message='Cloudflare Firewall',
suspended_time=get_setting('search.suspended_times.cf_SearxEngineAccessDenied'),
)
def raise_for_recaptcha(resp):
if resp.status_code == 503 and '"https://www.google.com/recaptcha/' in resp.text:
raise SearxEngineCaptchaException(
message='ReCAPTCHA', suspended_time=get_setting('search.suspended_times.recaptcha_SearxEngineCaptcha')
)
def raise_for_captcha(resp):
raise_for_cloudflare_captcha(resp)
raise_for_recaptcha(resp)
def raise_for_httperror(resp):
"""Raise exception for an HTTP response is an error.
Args:
resp (requests.Response): Response to check
Raises:
requests.HTTPError: raise by resp.raise_for_status()
searx.exceptions.SearxEngineAccessDeniedException: raise when the HTTP status code is 402 or 403.
searx.exceptions.SearxEngineTooManyRequestsException: raise when the HTTP status code is 429.
searx.exceptions.SearxEngineCaptchaException: raise when if CATPCHA challenge is detected.
"""
if resp.status_code and resp.status_code >= 400:
raise_for_captcha(resp)
if resp.status_code in (402, 403):
raise SearxEngineAccessDeniedException(message='HTTP error ' + str(resp.status_code))
if resp.status_code == 429:
raise SearxEngineTooManyRequestsException()
resp.raise_for_status()