612 lines
16 KiB
Python
612 lines
16 KiB
Python
import asyncio
|
|
import logging
|
|
|
|
from collections import deque, namedtuple
|
|
|
|
import pytest
|
|
import uvicorn
|
|
|
|
from sanic import Sanic
|
|
from sanic.application.state import Mode
|
|
from sanic.asgi import ASGIApp, MockTransport
|
|
from sanic.exceptions import BadRequest, Forbidden, ServiceUnavailable
|
|
from sanic.request import Request
|
|
from sanic.response import json, text
|
|
from sanic.server.websockets.connection import WebSocketConnection
|
|
from sanic.signals import RESERVED_NAMESPACES
|
|
|
|
|
|
try:
|
|
from unittest.mock import AsyncMock
|
|
except ImportError:
|
|
from tests.asyncmock import AsyncMock # type: ignore
|
|
|
|
|
|
@pytest.fixture
|
|
def message_stack():
|
|
return deque()
|
|
|
|
|
|
@pytest.fixture
|
|
def receive(message_stack):
|
|
async def _receive():
|
|
return message_stack.popleft()
|
|
|
|
return _receive
|
|
|
|
|
|
@pytest.fixture
|
|
def send(message_stack):
|
|
async def _send(message):
|
|
message_stack.append(message)
|
|
|
|
return _send
|
|
|
|
|
|
@pytest.fixture
|
|
def transport(message_stack, receive, send):
|
|
return MockTransport({}, receive, send)
|
|
|
|
|
|
@pytest.fixture
|
|
def protocol(transport):
|
|
return transport.get_protocol()
|
|
|
|
|
|
def test_listeners_triggered(caplog):
|
|
app = Sanic("app")
|
|
before_server_start = False
|
|
after_server_start = False
|
|
before_server_stop = False
|
|
after_server_stop = False
|
|
|
|
@app.listener("before_server_start")
|
|
def do_before_server_start(*args, **kwargs):
|
|
nonlocal before_server_start
|
|
before_server_start = True
|
|
|
|
@app.listener("after_server_start")
|
|
def do_after_server_start(*args, **kwargs):
|
|
nonlocal after_server_start
|
|
after_server_start = True
|
|
|
|
@app.listener("before_server_stop")
|
|
def do_before_server_stop(*args, **kwargs):
|
|
nonlocal before_server_stop
|
|
before_server_stop = True
|
|
|
|
@app.listener("after_server_stop")
|
|
def do_after_server_stop(*args, **kwargs):
|
|
nonlocal after_server_stop
|
|
after_server_stop = True
|
|
|
|
@app.route("/")
|
|
def handler(request):
|
|
return text("...")
|
|
|
|
class CustomServer(uvicorn.Server):
|
|
def install_signal_handlers(self):
|
|
pass
|
|
|
|
config = uvicorn.Config(app=app, loop="asyncio", limit_max_requests=0)
|
|
server = CustomServer(config=config)
|
|
|
|
start_message = (
|
|
'You have set a listener for "before_server_start" in ASGI mode. '
|
|
"It will be executed as early as possible, but not before the ASGI "
|
|
"server is started."
|
|
)
|
|
stop_message = (
|
|
'You have set a listener for "after_server_stop" in ASGI mode. '
|
|
"It will be executed as late as possible, but not after the ASGI "
|
|
"server is stopped."
|
|
)
|
|
|
|
with caplog.at_level(logging.DEBUG):
|
|
server.run()
|
|
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
start_message,
|
|
) not in caplog.record_tuples
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
stop_message,
|
|
) not in caplog.record_tuples
|
|
|
|
all_tasks = asyncio.all_tasks(asyncio.get_event_loop())
|
|
for task in all_tasks:
|
|
task.cancel()
|
|
|
|
assert before_server_start
|
|
assert after_server_start
|
|
assert before_server_stop
|
|
assert after_server_stop
|
|
|
|
app.state.mode = Mode.DEBUG
|
|
with caplog.at_level(logging.DEBUG):
|
|
server.run()
|
|
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
start_message,
|
|
) not in caplog.record_tuples
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
stop_message,
|
|
) not in caplog.record_tuples
|
|
|
|
app.state.verbosity = 2
|
|
with caplog.at_level(logging.DEBUG):
|
|
server.run()
|
|
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
start_message,
|
|
) in caplog.record_tuples
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
stop_message,
|
|
) in caplog.record_tuples
|
|
|
|
|
|
def test_listeners_triggered_async(app, caplog):
|
|
before_server_start = False
|
|
after_server_start = False
|
|
before_server_stop = False
|
|
after_server_stop = False
|
|
|
|
@app.listener("before_server_start")
|
|
async def do_before_server_start(*args, **kwargs):
|
|
nonlocal before_server_start
|
|
before_server_start = True
|
|
|
|
@app.listener("after_server_start")
|
|
async def do_after_server_start(*args, **kwargs):
|
|
nonlocal after_server_start
|
|
after_server_start = True
|
|
|
|
@app.listener("before_server_stop")
|
|
async def do_before_server_stop(*args, **kwargs):
|
|
nonlocal before_server_stop
|
|
before_server_stop = True
|
|
|
|
@app.listener("after_server_stop")
|
|
async def do_after_server_stop(*args, **kwargs):
|
|
nonlocal after_server_stop
|
|
after_server_stop = True
|
|
|
|
@app.route("/")
|
|
def handler(request):
|
|
return text("...")
|
|
|
|
class CustomServer(uvicorn.Server):
|
|
def install_signal_handlers(self):
|
|
pass
|
|
|
|
config = uvicorn.Config(app=app, loop="asyncio", limit_max_requests=0)
|
|
server = CustomServer(config=config)
|
|
|
|
start_message = (
|
|
'You have set a listener for "before_server_start" in ASGI mode. '
|
|
"It will be executed as early as possible, but not before the ASGI "
|
|
"server is started."
|
|
)
|
|
stop_message = (
|
|
'You have set a listener for "after_server_stop" in ASGI mode. '
|
|
"It will be executed as late as possible, but not after the ASGI "
|
|
"server is stopped."
|
|
)
|
|
|
|
with caplog.at_level(logging.DEBUG):
|
|
server.run()
|
|
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
start_message,
|
|
) not in caplog.record_tuples
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
stop_message,
|
|
) not in caplog.record_tuples
|
|
|
|
all_tasks = asyncio.all_tasks(asyncio.get_event_loop())
|
|
for task in all_tasks:
|
|
task.cancel()
|
|
|
|
assert before_server_start
|
|
assert after_server_start
|
|
assert before_server_stop
|
|
assert after_server_stop
|
|
|
|
app.state.mode = Mode.DEBUG
|
|
app.state.verbosity = 0
|
|
with caplog.at_level(logging.DEBUG):
|
|
server.run()
|
|
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
start_message,
|
|
) not in caplog.record_tuples
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
stop_message,
|
|
) not in caplog.record_tuples
|
|
|
|
app.state.verbosity = 2
|
|
with caplog.at_level(logging.DEBUG):
|
|
server.run()
|
|
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
start_message,
|
|
) in caplog.record_tuples
|
|
assert (
|
|
"sanic.root",
|
|
logging.DEBUG,
|
|
stop_message,
|
|
) in caplog.record_tuples
|
|
|
|
|
|
def test_non_default_uvloop_config_raises_warning(app):
|
|
app.config.USE_UVLOOP = True
|
|
|
|
class CustomServer(uvicorn.Server):
|
|
def install_signal_handlers(self):
|
|
pass
|
|
|
|
config = uvicorn.Config(app=app, loop="asyncio", limit_max_requests=0)
|
|
server = CustomServer(config=config)
|
|
|
|
with pytest.warns(UserWarning) as records:
|
|
server.run()
|
|
|
|
all_tasks = asyncio.all_tasks(asyncio.get_event_loop())
|
|
for task in all_tasks:
|
|
task.cancel()
|
|
|
|
msg = ""
|
|
for record in records:
|
|
_msg = str(record.message)
|
|
if _msg.startswith("You have set the USE_UVLOOP configuration"):
|
|
msg = _msg
|
|
break
|
|
|
|
assert msg == (
|
|
"You have set the USE_UVLOOP configuration option, but Sanic "
|
|
"cannot control the event loop when running in ASGI mode."
|
|
"This option will be ignored."
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_mockprotocol_events(protocol):
|
|
assert protocol._not_paused.is_set()
|
|
protocol.pause_writing()
|
|
assert not protocol._not_paused.is_set()
|
|
protocol.resume_writing()
|
|
assert protocol._not_paused.is_set()
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_protocol_push_data(protocol, message_stack):
|
|
text = b"hello"
|
|
|
|
await protocol.push_data(text)
|
|
await protocol.complete()
|
|
|
|
assert len(message_stack) == 2
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "http.response.body"
|
|
assert message["more_body"]
|
|
assert message["body"] == text
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "http.response.body"
|
|
assert not message["more_body"]
|
|
assert message["body"] == b""
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_websocket_send(send, receive, message_stack):
|
|
text_string = "hello"
|
|
text_bytes = b"hello"
|
|
|
|
ws = WebSocketConnection(send, receive)
|
|
await ws.send(text_string)
|
|
await ws.send(text_bytes)
|
|
|
|
assert len(message_stack) == 2
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "websocket.send"
|
|
assert message["text"] == text_string
|
|
assert "bytes" not in message
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "websocket.send"
|
|
assert message["bytes"] == text_bytes
|
|
assert "text" not in message
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_websocket_text_receive(send, receive, message_stack):
|
|
msg = {"text": "hello", "type": "websocket.receive"}
|
|
message_stack.append(msg)
|
|
|
|
ws = WebSocketConnection(send, receive)
|
|
text = await ws.receive()
|
|
|
|
assert text == msg["text"]
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_websocket_bytes_receive(send, receive, message_stack):
|
|
msg = {"bytes": b"hello", "type": "websocket.receive"}
|
|
message_stack.append(msg)
|
|
|
|
ws = WebSocketConnection(send, receive)
|
|
data = await ws.receive()
|
|
|
|
assert data == msg["bytes"]
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_websocket_accept_with_no_subprotocols(
|
|
send, receive, message_stack
|
|
):
|
|
ws = WebSocketConnection(send, receive)
|
|
await ws.accept()
|
|
|
|
assert len(message_stack) == 1
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "websocket.accept"
|
|
assert message["subprotocol"] is None
|
|
assert "bytes" not in message
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_websocket_accept_with_subprotocol(send, receive, message_stack):
|
|
subprotocols = ["graphql-ws"]
|
|
|
|
ws = WebSocketConnection(send, receive, subprotocols)
|
|
await ws.accept(subprotocols)
|
|
|
|
assert len(message_stack) == 1
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "websocket.accept"
|
|
assert message["subprotocol"] == "graphql-ws"
|
|
assert "bytes" not in message
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_websocket_accept_with_multiple_subprotocols(
|
|
send, receive, message_stack
|
|
):
|
|
subprotocols = ["graphql-ws", "hello", "world"]
|
|
|
|
ws = WebSocketConnection(send, receive, subprotocols)
|
|
await ws.accept(["hello", "world"])
|
|
|
|
assert len(message_stack) == 1
|
|
|
|
message = message_stack.popleft()
|
|
assert message["type"] == "websocket.accept"
|
|
assert message["subprotocol"] == "hello"
|
|
assert "bytes" not in message
|
|
|
|
|
|
def test_improper_websocket_connection(transport, send, receive):
|
|
with pytest.raises(BadRequest):
|
|
transport.get_websocket_connection()
|
|
|
|
transport.create_websocket_connection(send, receive)
|
|
connection = transport.get_websocket_connection()
|
|
assert isinstance(connection, WebSocketConnection)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_request_class_regular(app):
|
|
@app.get("/regular")
|
|
def regular_request(request):
|
|
return text(request.__class__.__name__)
|
|
|
|
_, response = await app.asgi_client.get("/regular")
|
|
assert response.body == b"Request"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_request_class_custom():
|
|
class MyCustomRequest(Request):
|
|
pass
|
|
|
|
app = Sanic(name="Test", request_class=MyCustomRequest)
|
|
|
|
@app.get("/custom")
|
|
def custom_request(request):
|
|
return text(request.__class__.__name__)
|
|
|
|
_, response = await app.asgi_client.get("/custom")
|
|
assert response.body == b"MyCustomRequest"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_cookie_customization(app):
|
|
@app.get("/cookie")
|
|
def get_cookie(request):
|
|
response = text("There's a cookie up in this response")
|
|
response.cookies["test"] = "Cookie1"
|
|
response.cookies["test"]["httponly"] = True
|
|
|
|
response.cookies["c2"] = "Cookie2"
|
|
response.cookies["c2"]["httponly"] = False
|
|
|
|
return response
|
|
|
|
_, response = await app.asgi_client.get("/cookie")
|
|
|
|
CookieDef = namedtuple("CookieDef", ("value", "httponly"))
|
|
Cookie = namedtuple("Cookie", ("domain", "path", "value", "httponly"))
|
|
cookie_map = {
|
|
"test": CookieDef("Cookie1", True),
|
|
"c2": CookieDef("Cookie2", False),
|
|
}
|
|
|
|
cookies = {
|
|
c.name: Cookie(c.domain, c.path, c.value, "HttpOnly" in c._rest.keys())
|
|
for c in response.cookies.jar
|
|
}
|
|
|
|
for name, definition in cookie_map.items():
|
|
cookie = cookies.get(name)
|
|
assert cookie
|
|
assert cookie.value == definition.value
|
|
assert cookie.domain == "mockserver.local"
|
|
assert cookie.path == "/"
|
|
assert cookie.httponly == definition.httponly
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_content_type(app):
|
|
@app.get("/json")
|
|
def send_json(request):
|
|
return json({"foo": "bar"})
|
|
|
|
@app.get("/text")
|
|
def send_text(request):
|
|
return text("foobar")
|
|
|
|
@app.get("/custom")
|
|
def send_custom(request):
|
|
return text("foobar", content_type="somethingelse")
|
|
|
|
_, response = await app.asgi_client.get("/json")
|
|
assert response.headers.get("content-type") == "application/json"
|
|
|
|
_, response = await app.asgi_client.get("/text")
|
|
assert response.headers.get("content-type") == "text/plain; charset=utf-8"
|
|
|
|
_, response = await app.asgi_client.get("/custom")
|
|
assert response.headers.get("content-type") == "somethingelse"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_request_handle_exception(app):
|
|
@app.get("/error-prone")
|
|
def _request(request):
|
|
raise ServiceUnavailable(message="Service unavailable")
|
|
|
|
_, response = await app.asgi_client.get("/wrong-path")
|
|
assert response.status_code == 404
|
|
|
|
_, response = await app.asgi_client.get("/error-prone")
|
|
assert response.status_code == 503
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_request_exception_suppressed_by_middleware(app):
|
|
@app.get("/error-prone")
|
|
def _request(request):
|
|
raise ServiceUnavailable(message="Service unavailable")
|
|
|
|
@app.on_request
|
|
def forbidden(request):
|
|
raise Forbidden(message="forbidden")
|
|
|
|
_, response = await app.asgi_client.get("/wrong-path")
|
|
assert response.status_code == 403
|
|
|
|
_, response = await app.asgi_client.get("/error-prone")
|
|
assert response.status_code == 403
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_signals_triggered(app):
|
|
@app.get("/test_signals_triggered")
|
|
async def _request(request):
|
|
return text("test_signals_triggered")
|
|
|
|
signals_triggered = []
|
|
signals_expected = [
|
|
# "http.lifecycle.begin",
|
|
# "http.lifecycle.read_head",
|
|
"http.lifecycle.request",
|
|
"http.lifecycle.handle",
|
|
"http.routing.before",
|
|
"http.routing.after",
|
|
"http.handler.before",
|
|
"http.handler.after",
|
|
"http.lifecycle.response",
|
|
# "http.lifecycle.send",
|
|
# "http.lifecycle.complete",
|
|
]
|
|
|
|
def signal_handler(signal):
|
|
return lambda *a, **kw: signals_triggered.append(signal)
|
|
|
|
for signal in RESERVED_NAMESPACES["http"]:
|
|
app.signal(signal)(signal_handler(signal))
|
|
|
|
_, response = await app.asgi_client.get("/test_signals_triggered")
|
|
assert response.status_code == 200
|
|
assert response.text == "test_signals_triggered"
|
|
assert signals_triggered == signals_expected
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_asgi_serve_location(app):
|
|
@app.get("/")
|
|
def _request(request: Request):
|
|
return text(request.app.serve_location)
|
|
|
|
_, response = await app.asgi_client.get("/")
|
|
assert response.text == "http://<ASGI>"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_error_on_lifespan_exception_start(app, caplog):
|
|
@app.before_server_start
|
|
async def before_server_start(_):
|
|
1 / 0
|
|
|
|
recv = AsyncMock(return_value={"type": "lifespan.startup"})
|
|
send = AsyncMock()
|
|
app.asgi = True
|
|
|
|
with caplog.at_level(logging.ERROR):
|
|
await ASGIApp.create(app, {"type": "lifespan"}, recv, send)
|
|
|
|
send.assert_awaited_once_with(
|
|
{"type": "lifespan.startup.failed", "message": "division by zero"}
|
|
)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_error_on_lifespan_exception_stop(app: Sanic):
|
|
@app.before_server_stop
|
|
async def before_server_stop(_):
|
|
1 / 0
|
|
|
|
recv = AsyncMock(return_value={"type": "lifespan.shutdown"})
|
|
send = AsyncMock()
|
|
app.asgi = True
|
|
await app._startup()
|
|
|
|
await ASGIApp.create(app, {"type": "lifespan"}, recv, send)
|
|
|
|
send.assert_awaited_once_with(
|
|
{"type": "lifespan.shutdown.failed", "message": "division by zero"}
|
|
)
|