an asyncio testing server for mocking external services
- Fast mocks using actual network connections
- allows mocking some types of network issues
- use regular expression matching for domain, path, method, or body
- works with https requests as well (by switching them to http requests)
- works with callables
Add routes and responses via the aresponses.add
method:
def add(
host_pattern=ANY,
path_pattern=ANY,
method_pattern=ANY,
response="",
*,
route=None,
body_pattern=ANY, m
match_querystring=False,
repeat=1
)
When a request is received the first matching response will be returned
and removed from the routing table. The response
argument can be
either a string, Response, dict, or list. Use aresponses.Response
when you need to do something more complex.
Note that version >=2.0 requires explicit assertions!
@pytest.mark.asyncio
async def test_simple(aresponses):
aresponses.add("google.com", "/api/v1/", "GET", response="OK")
aresponses.add('foo.com', '/', 'get', aresponses.Response(text='error', status=500))
async with aiohttp.ClientSession() as session:
async with session.get("http://google.com/api/v1/") as response:
text = await response.text()
assert text == "OK"
async with session.get("https://foo.com") as response:
text = await response.text()
assert text == "error"
aresponses.assert_plan_strictly_followed()
In aresponses 1.x requests that didn't match a route stopped the event loop and thus forced an exception. In aresponses >2.x it's required to make assertions at the end of the test.
There are three assertion functions provided:
aresponses.assert_no_unused_routes
RaisesUnusedRouteError
if all the routes defined were not used up.aresponses.assert_called_in_order
- RaisesUnorderedRouteCallError
if the routes weren't called in the order they were defined.aresponses.assert_all_requests_matched
- RaisesNoRouteFoundError
if any requests were made that didn't match to a route. It's likely but not guaranteed that your code will throw an exception in this situation before the assertion is reached.
Instead of calling these individually, it's recommended to call
aresponses.assert_plan_strictly_followed()
at the end of each test as
it runs all three of the above assertions.
host_pattern
, path_pattern
, method_pattern
and body_pattern
may
be either strings (exact match) or regular expressions.
The repeat argument permits a route to be used multiple times.
If you want to just blanket mock a service, without concern for how many
times its called, you could set repeat to a large number and not call
aresponses.assert_plan_strictly_followed
or
arespones.assert_no_unused_routes
.
@pytest.mark.asyncio
async def test_regex_repetition(aresponses):
aresponses.add(re.compile(r".*\.?google\.com"), response="OK", repeat=2)
async with aiohttp.ClientSession() as session:
async with session.get("http://google.com") as response:
text = await response.text()
assert text == "OK"
async with session.get("http://api.google.com") as response:
text = await response.text()
assert text == "OK"
aresponses.assert_plan_strictly_followed()
As a convenience, if a dict or list is passed to response
then it will
create a json response. A aiohttp.web_response.json_response
object
can be used for more complex situations.
@pytest.mark.asyncio
async def test_json(aresponses):
aresponses.add("google.com", "/api/v1/", "GET", response={"status": "OK"})
async with aiohttp.ClientSession() as session:
async with session.get("http://google.com/api/v1/") as response:
assert {"status": "OK"} == await response.json()
aresponses.assert_plan_strictly_followed()
Custom functions can be used for whatever other complex logic is desired. In the example below, the handler is set to repeat infinitely and always return 500.
import math
@pytest.mark.asyncio
async def test_handler(aresponses):
def break_everything(request):
return aresponses.Response(status=500, text=str(request.url))
aresponses.add(response=break_everything, repeat=math.inf)
async with aiohttp.ClientSession() as session:
async with session.get("http://google.com/api/v1/") as response:
assert response.status == 500
Pass aresponses.passthrough
into the response argument to allow a
request to bypass mocking.
aresponses.add('httpstat.us', '/200', 'get', aresponses.passthrough)
History of calls can be inspected via aresponses.history
which returns
the namedTuple RoutingLog(request, route, response)
@pytest.mark.asyncio
async def test_history(aresponses):
aresponses.add(response=aresponses.Response(text="hi"), repeat=2)
async with aiohttp.ClientSession() as session:
async with session.get("http://foo.com/b") as response:
await response.text()
async with session.get("http://bar.com/a") as response:
await response.text()
assert len(aresponses.history) == 2
assert aresponses.history[0].request.host == "foo.com"
assert aresponses.history[1].request.host == "bar.com"
assert "Route(" in repr(aresponses.history[0].route)
aresponses.assert_plan_strictly_followed()
import aiohttp
import pytest
import aresponses
@pytest.mark.asyncio
async def test_foo(event_loop):
async with aresponses.ResponsesMockServer(loop=event_loop) as arsps:
arsps.add('foo.com', '/', 'get', 'hi there!!')
arsps.add(arsps.ANY, '/', 'get', arsps.Response(text='hey!'))
async with aiohttp.ClientSession(loop=event_loop) as session:
async with session.get('http://foo.com') as response:
text = await response.text()
assert text == 'hi'
async with session.get('https://google.com') as response:
text = await response.text()
assert text == 'hey!'
working with pytest-aiohttp
If you need to use aresponses together with pytest-aiohttp, you should re-initialize the main aresponses fixture with the loop
fixture
from aresponses import ResponsesMockServer
@pytest.fixture
async def aresponses(loop):
async with ResponsesMockServer(loop=loop) as server:
yield server
If you're trying to use the aiohttp_client
test fixture then you'll need to mock out the aiohttp loop
fixture
instead:
@pytest.fixture
def loop(event_loop):
"""replace aiohttp loop fixture with pytest-asyncio fixture"""
return event_loop
- install pyenv and pyenv-virtualenv - Makes it easy to install specific versions of python and switch between them. Make sure you install the virtualenv bash hook
git clone
the repo andcd
into it.make init
- installs proper version of python, creates the virtual environment, activates it and installs all the requirements
git checkout -b my-feature-branch
- make some cool changes
make autoformat
make test
make lint
- create pull request
make deploy
- fix: start using
asyncio.get_running_loop()
instead ofevent_loop
per the error:PytestDeprecationWarning: aresponses is asynchronous and explicitly requests the "event_loop" fixture. Asynchronous fixtures and test functions should use "asyncio.get_running_loop()" instead.
- drop support for python 3.6
- add comprehensive matrix testing of all supported python and aiohttp versions
- tighten up the setup.py requirements
- fix: incorrect pytest plugin entrypoint name (#72)
- support asyncio_mode = strict (#68)
- fix: don't assume utf8 request contents
- accidental no-op release
- documentation: add pypi documentation
- bugfix: RecursionError when aresponses is used in more than 1000 tests (#63)
- feature: add convenience method
add_local_passthrough
- bugfix: fix https subrequest mocks. support aiohttp_client compatibility
- bugfix: ensure request body is available in history
Warning! Breaking Changes!
- breaking change: require explicit assertions for test failures
- feature: autocomplete works in intellij/pycharm
- feature: can match on body of request
- feature: store calls made
- feature: repeated responses
- bugfix: no longer stops event loop
- feature: if dict or list is passed into
response
, a json response will be generated
- make passthrough feature work with binary data
- regex fix for Python 3.7.0
- Added passthrough option to permit live network calls
- Added example of using a callable as a response
- Added an optional
match_querystring
argument that lets you match on querystring as well
- Bryce Drennan, CircleUp [email protected]
- Marco Castelluccio, Mozilla [email protected]
- Jesse Vogt, CircleUp [email protected]
- Pavol Vargovcik, Kiwi.com [email protected]