A utility library for mocking out the requests Python library.

pip install responses

Here you will find a list of deprecated functionality and a migration path for each. Please ensure to update your code according to the guidance.

Below you can find a list of BETA features. Although we will try to keep the API backwards compatible with released version, we reserve the right to change these APIs before they are considered stable. Please share your feedback via GitHub Issues.

You can perform real requests to the server and responses will automatically record the output to the file. Recorded data is stored in YAML format.

Apply @responses._recorder.record(file_path="out.yaml") decorator to any function where you perform requests to record responses to out.yaml file.

Following code

import requests
from responses import _recorder


def another():
    rsp = requests.get("https://httpstat.us/500")
    rsp = requests.get("https://httpstat.us/202")


@_recorder.record(file_path="out.yaml")
def test_recorder():
    rsp = requests.get("https://httpstat.us/404")
    rsp = requests.get("https://httpbin.org/status/wrong")
    another()

will produce next output:

responses:
- response:
    auto_calculate_content_length: false
    body: 404 Not Found
    content_type: text/plain
    method: GET
    status: 404
    url: https://httpstat.us/404
- response:
    auto_calculate_content_length: false
    body: Invalid status code
    content_type: text/plain
    method: GET
    status: 400
    url: https://httpbin.org/status/wrong
- response:
    auto_calculate_content_length: false
    body: 500 Internal Server Error
    content_type: text/plain
    method: GET
    status: 500
    url: https://httpstat.us/500
- response:
    auto_calculate_content_length: false
    body: 202 Accepted
    content_type: text/plain
    method: GET
    status: 202
    url: https://httpstat.us/202

You can populate your active registry from a yaml file with recorded responses. (See Record Responses to files to understand how to obtain a file). To do that you need to execute responses._add_from_file(file_path="out.yaml") within an activated decorator or a context manager.

The following code example registers a patch response, then all responses present in out.yaml file and a post response at the end.

import responses


@responses.activate
def run():
    responses.patch("http://httpbin.org")
    responses._add_from_file(file_path="out.yaml")
    responses.post("http://httpbin.org/form")


run()

The core of responses comes from registering mock responses and covering test function with responses.activate decorator. responses provides similar interface as requests.

• responses.add(Response or Response args) - allows either to register Response object or directly provide arguments of Response object. See Response Parameters

import responses
import requests


@responses.activate
def test_simple():
    # Register via 'Response' object
    rsp1 = responses.Response(
        method="PUT",
        url="http://example.com",
    )
    responses.add(rsp1)
    # register via direct arguments
    responses.add(
        responses.GET,
        "http://twitter.com/api/1/foobar",
        json={"error": "not found"},
        status=404,
    )

    resp = requests.get("http://twitter.com/api/1/foobar")
    resp2 = requests.put("http://example.com")

    assert resp.json() == {"error": "not found"}
    assert resp.status_code == 404

    assert resp2.status_code == 200
    assert resp2.request.method == "PUT"

If you attempt to fetch a url which doesn't hit a match, responses will raise a ConnectionError:

import responses
import requests

from requests.exceptions import ConnectionError


@responses.activate
def test_simple():
    with pytest.raises(ConnectionError):
        requests.get("http://twitter.com/api/1/foobar")

Shortcuts provide a shorten version of responses.add() where method argument is prefilled

• responses.delete(Response args) - register DELETE response
• responses.get(Response args) - register GET response
• responses.head(Response args) - register HEAD response
• responses.options(Response args) - register OPTIONS response
• responses.patch(Response args) - register PATCH response
• responses.post(Response args) - register POST response
• responses.put(Response args) - register PUT response

import responses
import requests


@responses.activate
def test_simple():
    responses.get(
        "http://twitter.com/api/1/foobar",
        json={"type": "get"},
    )

    responses.post(
        "http://twitter.com/api/1/foobar",
        json={"type": "post"},
    )

    responses.patch(
        "http://twitter.com/api/1/foobar",
        json={"type": "patch"},
    )

    resp_get = requests.get("http://twitter.com/api/1/foobar")
    resp_post = requests.post("http://twitter.com/api/1/foobar")
    resp_patch = requests.patch("http://twitter.com/api/1/foobar")

    assert resp_get.json() == {"type": "get"}
    assert resp_post.json() == {"type": "post"}
    assert resp_patch.json() == {"type": "patch"}

Instead of wrapping the whole function with decorator you can use a context manager.

import responses
import requests


def test_my_api():
    with responses.RequestsMock() as rsps:
        rsps.add(
            responses.GET,
            "http://twitter.com/api/1/foobar",
            body="{}",
            status=200,
            content_type="application/json",
        )
        resp = requests.get("http://twitter.com/api/1/foobar")

        assert resp.status_code == 200

    # outside the context manager requests will hit the remote server
    resp = requests.get("http://twitter.com/api/1/foobar")
    resp.status_code == 404

The following attributes can be passed to a Response mock:

method (str)

The HTTP method (GET, POST, etc).

url (str or compiled regular expression)

The full resource URL.

match_querystring (bool)

DEPRECATED: Use responses.matchers.query_param_matcher or responses.matchers.query_string_matcher

Include the query string when matching requests. Enabled by default if the response URL contains a query string, disabled if it doesn't or the URL is a regular expression.

body (str or BufferedReader or Exception)

The response body. Read more Exception as Response body

json

A Python object representing the JSON response body. Automatically configures the appropriate Content-Type.

status (int)

The HTTP status code.

content_type (content_type)

Defaults to text/plain.

headers (dict)

Response headers.

stream (bool)

DEPRECATED: use stream argument in request directly

auto_calculate_content_length (bool)

Disabled by default. Automatically calculates the length of a supplied string or JSON body.

match (tuple)

An iterable (tuple is recommended) of callbacks to match requests based on request attributes. Current module provides multiple matchers that you can use to match:

• body contents in JSON format
• body contents in URL encoded data format
• request query parameters
• request query string (similar to query parameters but takes string as input)
• kwargs provided to request e.g. stream, verify
• 'multipart/form-data' content and headers in request
• request headers
• request fragment identifier

Alternatively user can create custom matcher. Read more Matching Requests

You can pass an Exception as the body to trigger an error on the request:

import responses
import requests


@responses.activate
def test_simple():
    responses.get("http://twitter.com/api/1/foobar", body=Exception("..."))
    with pytest.raises(Exception):
        requests.get("http://twitter.com/api/1/foobar")

When adding responses for endpoints that are sent request data you can add matchers to ensure your code is sending the right parameters and provide different responses based on the request body contents. responses provides matchers for JSON and URL-encoded request bodies.

import responses
import requests
from responses import matchers


@responses.activate
def test_calc_api():
    responses.post(
        url="http://calc.com/sum",
        body="4",
        match=[matchers.urlencoded_params_matcher({"left": "1", "right": "3"})],
    )
    requests.post("http://calc.com/sum", data={"left": 1, "right": 3})

Matching JSON encoded data can be done with matchers.json_params_matcher().

import responses
import requests
from responses import matchers


@responses.activate
def test_calc_api():
    responses.post(
        url="http://example.com/",
        body="one",
        match=[
            matchers.json_params_matcher({"page": {"name": "first", "type": "json"}})
        ],
    )
    resp = requests.request(
        "POST",
        "http://example.com/",
        headers={"Content-Type": "application/json"},
        json={"page": {"name": "first", "type": "json"}},
    )

You can use the matchers.query_param_matcher function to match against the params request parameter. Just use the same dictionary as you will use in params argument in request.

Note, do not use query parameters as part of the URL. Avoid using match_querystring deprecated argument.

import responses
import requests
from responses import matchers


@responses.activate
def test_calc_api():
    url = "http://example.com/test"
    params = {"hello": "world", "I am": "a big test"}
    responses.get(
        url=url,
        body="test",
        match=[matchers.query_param_matcher(params)],
    )

    resp = requests.get(url, params=params)

    constructed_url = r"http://example.com/test?I+am=a+big+test&hello=world"
    assert resp.url == constructed_url
    assert resp.request.url == constructed_url
    assert resp.request.params == params

By default, matcher will validate that all parameters match strictly. To validate that only parameters specified in the matcher are present in original request use strict_match=False.

As alternative, you can use query string value in matchers.query_string_matcher to match query parameters in your request

import requests
import responses
from responses import matchers


@responses.activate
def my_func():
    responses.get(
        "https://httpbin.org/get",
        match=[matchers.query_string_matcher("didi=pro&test=1")],
    )
    resp = requests.get("https://httpbin.org/get", params={"test": 1, "didi": "pro"})


my_func()

To validate request arguments use the matchers.request_kwargs_matcher function to match against the request kwargs.

Only following arguments are supported: timeout, verify, proxies, stream, cert.

Note, only arguments provided to matchers.request_kwargs_matcher will be validated.

import responses
import requests
from responses import matchers

with responses.RequestsMock(assert_all_requests_are_fired=False) as rsps:
    req_kwargs = {
        "stream": True,
        "verify": False,
    }
    rsps.add(
        "GET",
        "http://111.com",
        match=[matchers.request_kwargs_matcher(req_kwargs)],
    )

    requests.get("http://111.com", stream=True)

    # >>>  Arguments don't match: {stream: True, verify: True} doesn't match {stream: True, verify: False}

To validate request body and headers for multipart/form-data data you can use matchers.multipart_matcher. The data, and files parameters provided will be compared to the request:

import requests
import responses
from responses.matchers import multipart_matcher


@responses.activate
def my_func():
    req_data = {"some": "other", "data": "fields"}
    req_files = {"file_name": b"Old World!"}
    responses.post(
        url="http://httpbin.org/post",
        match=[multipart_matcher(req_files, data=req_data)],
    )
    resp = requests.post("http://httpbin.org/post", files={"file_name": b"New World!"})


my_func()
# >>> raises ConnectionError: multipart/form-data doesn't match. Request body differs.

To validate request URL fragment identifier you can use matchers.fragment_identifier_matcher. The matcher takes fragment string (everything after # sign) as input for comparison:

import requests
import responses
from responses.matchers import fragment_identifier_matcher


@responses.activate
def run():
    url = "http://example.com?ab=xy&zed=qwe#test=1&foo=bar"
    responses.get(
        url,
        match=[fragment_identifier_matcher("test=1&foo=bar")],
        body=b"test",
    )

    # two requests to check reversed order of fragment identifier
    resp = requests.get("http://example.com?ab=xy&zed=qwe#test=1&foo=bar")
    resp = requests.get("http://example.com?zed=qwe&ab=xy#foo=bar&test=1")


run()

When adding responses you can specify matchers to ensure that your code is sending the right headers and provide different responses based on the request headers.

import responses
import requests
from responses import matchers


@responses.activate
def test_content_type():
    responses.get(
        url="http://example.com/",
        body="hello world",
        match=[matchers.header_matcher({"Accept": "text/plain"})],
    )

    responses.get(
        url="http://example.com/",
        json={"content": "hello world"},
        match=[matchers.header_matcher({"Accept": "application/json"})],
    )

    # request in reverse order to how they were added!
    resp = requests.get("http://example.com/", headers={"Accept": "application/json"})
    assert resp.json() == {"content": "hello world"}

    resp = requests.get("http://example.com/", headers={"Accept": "text/plain"})
    assert resp.text == "hello world"

Because requests will send several standard headers in addition to what was specified by your code, request headers that are additional to the ones passed to the matcher are ignored by default. You can change this behaviour by passing strict_match=True to the matcher to ensure that only the headers that you're expecting are sent and no others. Note that you will probably have to use a PreparedRequest in your code to ensure that requests doesn't include any additional headers.

import responses
import requests
from responses import matchers


@responses.activate
def test_content_type():
    responses.get(
        url="http://example.com/",
        body="hello world",
        match=[matchers.header_matcher({"Accept": "text/plain"}, strict_match=True)],
    )

    # this will fail because requests adds its own headers
    with pytest.raises(ConnectionError):
        requests.get("http://example.com/", headers={"Accept": "text/plain"})

    # a prepared request where you overwrite the headers before sending will work
    session = requests.Session()
    prepped = session.prepare_request(
        requests.Request(
            method="GET",
            url="http://example.com/",
        )
    )
    prepped.headers = {"Accept": "text/plain"}

    resp = session.send(prepped)
    assert resp.text == "hello world"

If your application requires other encodings or different data validation you can build your own matcher that returns Tuple[matches: bool, reason: str]. Where boolean represents True or False if the request parameters match and the string is a reason in case of match failure. Your matcher can expect a PreparedRequest parameter to be provided by responses.

Note, PreparedRequest is customized and has additional attributes params and req_kwargs.

By default, responses will search all registered Response objects and return a match. If only one Response is registered, the registry is kept unchanged. However, if multiple matches are found for the same request, then first match is returned and removed from registry.

In some scenarios it is important to preserve the order of the requests and responses. You can use registries.OrderedRegistry to force all Response objects to be dependent on the insertion order and invocation index. In following example we add multiple Response objects that target the same URL. However, you can see, that status code will depend on the invocation order.

import requests

import responses
from responses.registries import OrderedRegistry


@responses.activate(registry=OrderedRegistry)
def test_invocation_index():
    responses.get(
        "http://twitter.com/api/1/foobar",
        json={"msg": "not found"},
        status=404,
    )
    responses.get(
        "http://twitter.com/api/1/foobar",
        json={"msg": "OK"},
        status=200,
    )
    responses.get(
        "http://twitter.com/api/1/foobar",
        json={"msg": "OK"},
        status=200,
    )
    responses.get(
        "http://twitter.com/api/1/foobar",
        json={"msg": "not found"},
        status=404,
    )

    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 404
    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 200
    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 200
    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 404

Built-in registries are suitable for most of use cases, but to handle special conditions, you can implement custom registry which must follow interface of registries.FirstMatchRegistry. Redefining the find method will allow you to create custom search logic and return appropriate Response

Example that shows how to set custom registry

import responses
from responses import registries


class CustomRegistry(registries.FirstMatchRegistry):
    pass


print("Before tests:", responses.mock.get_registry())
""" Before tests: <responses.registries.FirstMatchRegistry object> """


# using function decorator
@responses.activate(registry=CustomRegistry)
def run():
    print("Within test:", responses.mock.get_registry())
    """ Within test: <__main__.CustomRegistry object> """


run()

print("After test:", responses.mock.get_registry())
""" After test: <responses.registries.FirstMatchRegistry object> """

# using context manager
with responses.RequestsMock(registry=CustomRegistry) as rsps:
    print("In context manager:", rsps.get_registry())
    """ In context manager: <__main__.CustomRegistry object> """

print("After exit from context manager:", responses.mock.get_registry())
"""
After exit from context manager: <responses.registries.FirstMatchRegistry object>
"""

You can utilize callbacks to provide dynamic responses. The callback must return a tuple of (status, headers, body).

import json

import responses
import requests


@responses.activate
def test_calc_api():
    def request_callback(request):
        payload = json.loads(request.body)
        resp_body = {"value": sum(payload["numbers"])}
        headers = {"request-id": "728d329e-0e86-11e4-a748-0c84dc037c13"}
        return (200, headers, json.dumps(resp_body))

    responses.add_callback(
        responses.POST,
        "http://calc.com/sum",
        callback=request_callback,
        content_type="application/json",
    )

    resp = requests.post(
        "http://calc.com/sum",
        json.dumps({"numbers": [1, 2, 3]}),
        headers={"content-type": "application/json"},
    )

    assert resp.json() == {"value": 6}

    assert len(responses.calls) == 1
    assert responses.calls[0].request.url == "http://calc.com/sum"
    assert responses.calls[0].response.text == '{"value": 6}'
    assert (
        responses.calls[0].response.headers["request-id"]
        == "728d329e-0e86-11e4-a748-0c84dc037c13"
    )

You can also pass a compiled regex to add_callback to match multiple urls:

import re, json

from functools import reduce

import responses
import requests

operators = {
    "sum": lambda x, y: x + y,
    "prod": lambda x, y: x * y,
    "pow": lambda x, y: x**y,
}


@responses.activate
def test_regex_url():
    def request_callback(request):
        payload = json.loads(request.body)
        operator_name = request.path_url[1:]

        operator = operators[operator_name]

        resp_body = {"value": reduce(operator, payload["numbers"])}
        headers = {"request-id": "728d329e-0e86-11e4-a748-0c84dc037c13"}
        return (200, headers, json.dumps(resp_body))

    responses.add_callback(
        responses.POST,
        re.compile("http://calc.com/(sum|prod|pow|unsupported)"),
        callback=request_callback,
        content_type="application/json",
    )

    resp = requests.post(
        "http://calc.com/prod",
        json.dumps({"numbers": [2, 3, 4]}),
        headers={"content-type": "application/json"},
    )
    assert resp.json() == {"value": 24}


test_regex_url()

If you want to pass extra keyword arguments to the callback function, for example when reusing a callback function to give a slightly different result, you can use functools.partial:

from functools import partial


def request_callback(request, id=None):
    payload = json.loads(request.body)
    resp_body = {"value": sum(payload["numbers"])}
    headers = {"request-id": id}
    return (200, headers, json.dumps(resp_body))


responses.add_callback(
    responses.POST,
    "http://calc.com/sum",
    callback=partial(request_callback, id="728d329e-0e86-11e4-a748-0c84dc037c13"),
    content_type="application/json",
)

@pytest.fixture
def mocked_responses():
    with responses.RequestsMock() as rsps:
        yield rsps


def test_api(mocked_responses):
    mocked_responses.get(
        "http://twitter.com/api/1/foobar",
        body="{}",
        status=200,
        content_type="application/json",
    )
    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 200

When run with unittest tests, this can be used to set up some generic class-level responses, that may be complemented by each test. Similar interface could be applied in pytest framework.

class TestMyApi(unittest.TestCase):
    def setUp(self):
        responses.get("https://example.com", body="within setup")
        # here go other self.responses.add(...)

    @responses.activate
    def test_my_func(self):
        responses.get(
            "https://httpbin.org/get",
            match=[matchers.query_param_matcher({"test": "1", "didi": "pro"})],
            body="within test",
        )
        resp = requests.get("https://example.com")
        resp2 = requests.get(
            "https://httpbin.org/get", params={"test": "1", "didi": "pro"}
        )
        print(resp.text)
        # >>> within setup
        print(resp2.text)
        # >>> within test

responses has start, stop, reset methods very analogous to unittest.mock.patch. These make it simpler to do requests mocking in setup methods or where you want to do multiple patches without nesting decorators or with statements.

class TestUnitTestPatchSetup:
    def setup(self):
        """Creates ``RequestsMock`` instance and starts it."""
        self.r_mock = responses.RequestsMock(assert_all_requests_are_fired=True)
        self.r_mock.start()

        # optionally some default responses could be registered
        self.r_mock.get("https://example.com", status=505)
        self.r_mock.put("https://example.com", status=506)

    def teardown(self):
        """Stops and resets RequestsMock instance.

        If ``assert_all_requests_are_fired`` is set to ``True``, will raise an error
        if some requests were not processed.
        """
        self.r_mock.stop()
        self.r_mock.reset()

    def test_function(self):
        resp = requests.get("https://example.com")
        assert resp.status_code == 505

        resp = requests.put("https://example.com")
        assert resp.status_code == 506

When used as a context manager, Responses will, by default, raise an assertion error if a url was registered but not accessed. This can be disabled by passing the assert_all_requests_are_fired value:

import responses
import requests


def test_my_api():
    with responses.RequestsMock(assert_all_requests_are_fired=False) as rsps:
        rsps.add(
            responses.GET,
            "http://twitter.com/api/1/foobar",
            body="{}",
            status=200,
            content_type="application/json",
        )

Each Response object has call_count attribute that could be inspected to check how many times each request was matched.

@responses.activate
def test_call_count_with_matcher():
    rsp = responses.get(
        "http://www.example.com",
        match=(matchers.query_param_matcher({}),),
    )
    rsp2 = responses.get(
        "http://www.example.com",
        match=(matchers.query_param_matcher({"hello": "world"}),),
        status=777,
    )
    requests.get("http://www.example.com")
    resp1 = requests.get("http://www.example.com")
    requests.get("http://www.example.com?hello=world")
    resp2 = requests.get("http://www.example.com?hello=world")

    assert resp1.status_code == 200
    assert resp2.status_code == 777

    assert rsp.call_count == 2
    assert rsp2.call_count == 2

Assert that the request was called exactly n times.

import responses
import requests


@responses.activate
def test_assert_call_count():
    responses.get("http://example.com")

    requests.get("http://example.com")
    assert responses.assert_call_count("http://example.com", 1) is True

    requests.get("http://example.com")
    with pytest.raises(AssertionError) as excinfo:
        responses.assert_call_count("http://example.com", 1)
    assert (
        "Expected URL 'http://example.com' to be called 1 times. Called 2 times."
        in str(excinfo.value)
    )


@responses.activate
def test_assert_call_count_always_match_qs():
    responses.get("http://www.example.com")
    requests.get("http://www.example.com")
    requests.get("http://www.example.com?hello=world")

    # One call on each url, querystring is matched by default
    responses.assert_call_count("http://www.example.com", 1) is True
    responses.assert_call_count("http://www.example.com?hello=world", 1) is True

Request object has calls list which elements correspond to Call objects in the global list of Registry. This can be useful when the order of requests is not guaranteed, but you need to check their correctness, for example in multithreaded applications.

import concurrent.futures
import responses
import requests


@responses.activate
def test_assert_calls_on_resp():
    rsp1 = responses.patch("http://www.foo.bar/1/", status=200)
    rsp2 = responses.patch("http://www.foo.bar/2/", status=400)
    rsp3 = responses.patch("http://www.foo.bar/3/", status=200)

    def update_user(uid, is_active):
        url = f"http://www.foo.bar/{uid}/"
        response = requests.patch(url, json={"is_active": is_active})
        return response

    with concurrent.futures.ThreadPoolExecutor(max_workers=3) as executor:
        future_to_uid = {
            executor.submit(update_user, uid, is_active): uid
            for (uid, is_active) in [("3", True), ("2", True), ("1", False)]
        }
        for future in concurrent.futures.as_completed(future_to_uid):
            uid = future_to_uid[future]
            response = future.result()
            print(f"{uid} updated with {response.status_code} status code")

    assert len(responses.calls) == 3  # total calls count

    assert rsp1.call_count == 1
    assert rsp1.calls[0] in responses.calls
    assert rsp1.calls[0].response.status_code == 200
    assert json.loads(rsp1.calls[0].request.body) == {"is_active": False}

    assert rsp2.call_count == 1
    assert rsp2.calls[0] in responses.calls
    assert rsp2.calls[0].response.status_code == 400
    assert json.loads(rsp2.calls[0].request.body) == {"is_active": True}

    assert rsp3.call_count == 1
    assert rsp3.calls[0] in responses.calls
    assert rsp3.calls[0].response.status_code == 200
    assert json.loads(rsp3.calls[0].request.body) == {"is_active": True}

You can also add multiple responses for the same url:

import responses
import requests


@responses.activate
def test_my_api():
    responses.get("http://twitter.com/api/1/foobar", status=500)
    responses.get(
        "http://twitter.com/api/1/foobar",
        body="{}",
        status=200,
        content_type="application/json",
    )

    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 500
    resp = requests.get("http://twitter.com/api/1/foobar")
    assert resp.status_code == 200

In the following example you can see how to create a redirection chain and add custom exception that will be raised in the execution chain and contain the history of redirects.

A -> 301 redirect -> B
B -> 301 redirect -> C
C -> connection issue

import pytest
import requests

import responses


@responses.activate
def test_redirect():
    # create multiple Response objects where first two contain redirect headers
    rsp1 = responses.Response(
        responses.GET,
        "http://example.com/1",
        status=301,
        headers={"Location": "http://example.com/2"},
    )
    rsp2 = responses.Response(
        responses.GET,
        "http://example.com/2",
        status=301,
        headers={"Location": "http://example.com/3"},
    )
    rsp3 = responses.Response(responses.GET, "http://example.com/3", status=200)

    # register above generated Responses in ``response`` module
    responses.add(rsp1)
    responses.add(rsp2)
    responses.add(rsp3)

    # do the first request in order to generate genuine ``requests`` response
    # this object will contain genuine attributes of the response, like ``history``
    rsp = requests.get("http://example.com/1")
    responses.calls.reset()

    # customize exception with ``response`` attribute
    my_error = requests.ConnectionError("custom error")
    my_error.response = rsp

    # update body of the 3rd response with Exception, this will be raised during execution
    rsp3.body = my_error

    with pytest.raises(requests.ConnectionError) as exc_info:
        requests.get("http://example.com/1")

    assert exc_info.value.args[0] == "custom error"
    assert rsp1.url in exc_info.value.response.history[0].url
    assert rsp2.url in exc_info.value.response.history[1].url

If you are using the Retry features of urllib3 and want to cover scenarios that test your retry limits, you can test those scenarios with responses as well. The best approach will be to use an Ordered Registry

import requests

import responses
from responses import registries
from urllib3.util import Retry


@responses.activate(registry=registries.OrderedRegistry)
def test_max_retries():
    url = "https://example.com"
    rsp1 = responses.get(url, body="Error", status=500)
    rsp2 = responses.get(url, body="Error", status=500)
    rsp3 = responses.get(url, body="Error", status=500)
    rsp4 = responses.get(url, body="OK", status=200)

    session = requests.Session()

    adapter = requests.adapters.HTTPAdapter(
        max_retries=Retry(
            total=4,
            backoff_factor=0.1,
            status_forcelist=[500],
            method_whitelist=["GET", "POST", "PATCH"],
        )
    )
    session.mount("https://", adapter)

    resp = session.get(url)

    assert resp.status_code == 200
    assert rsp1.call_count == 1
    assert rsp2.call_count == 1
    assert rsp3.call_count == 1
    assert rsp4.call_count == 1

If you use customized processing in requests via subclassing/mixins, or if you have library tools that interact with requests at a low level, you may need to add extended processing to the mocked Response object to fully simulate the environment for your tests. A response_callback can be used, which will be wrapped by the library before being returned to the caller. The callback accepts a response as it's single argument, and is expected to return a single response object.

import responses
import requests


def response_callback(resp):
    resp.callback_processed = True
    return resp


with responses.RequestsMock(response_callback=response_callback) as m:
    m.add(responses.GET, "http://example.com", body=b"test")
    resp = requests.get("http://example.com")
    assert resp.text == "test"
    assert hasattr(resp, "callback_processed")
    assert resp.callback_processed is True

In some cases you may wish to allow for certain requests to pass through responses and hit a real server. This can be done with the add_passthru methods:

import responses


@responses.activate
def test_my_api():
    responses.add_passthru("https://percy.io")

This will allow any requests matching that prefix, that is otherwise not registered as a mock response, to passthru using the standard behavior.

Pass through endpoints can be configured with regex patterns if you need to allow an entire domain or path subtree to send requests:

responses.add_passthru(re.compile("https://percy.io/\\w+"))

Lastly, you can use the passthrough argument of the Response object to force a response to behave as a pass through.

# Enable passthrough for a single response
response = Response(
    responses.GET,
    "http://example.com",
    body="not used",
    passthrough=True,
)
responses.add(response)

# Use PassthroughResponse
response = PassthroughResponse(responses.GET, "http://example.com")
responses.add(response)

Registered responses are available as a public method of the RequestMock instance. It is sometimes useful for debugging purposes to view the stack of registered responses which can be accessed via responses.registered().

The replace function allows a previously registered response to be changed. The method signature is identical to add. response s are identified using method and url. Only the first matched response is replaced.

import responses
import requests


@responses.activate
def test_replace():
    responses.get("http://example.org", json={"data": 1})
    responses.replace(responses.GET, "http://example.org", json={"data": 2})

    resp = requests.get("http://example.org")

    assert resp.json() == {"data": 2}

The upsert function allows a previously registered response to be changed like replace. If the response is registered, the upsert function will registered it like add.

remove takes a method and url argument and will remove all matched responses from the registered list.

Finally, reset will reset all registered responses.

responses supports both Coroutines and Multithreading out of the box. Note, responses locks threading on RequestMock object allowing only single thread to access it.

async def test_async_calls():
    @responses.activate
    async def run():
        responses.get(
            "http://twitter.com/api/1/foobar",
            json={"error": "not found"},
            status=404,
        )

        resp = requests.get("http://twitter.com/api/1/foobar")
        assert resp.json() == {"error": "not found"}
        assert responses.calls[0].request.url == "http://twitter.com/api/1/foobar"

    await run()

Responses uses several linting and autoformatting utilities, so it's important that when submitting patches you use the appropriate toolchain:

Clone the repository:

git clone https://github.com/getsentry/responses.git

Create an environment (e.g. with virtualenv):

virtualenv .env && source .env/bin/activate

Configure development requirements:

make develop

The easiest way to validate your code is to run tests via tox. Current tox configuration runs the same checks that are used in GitHub Actions CI/CD pipeline.

Please execute the following command line from the project root to validate your code against:

• Unit tests in all Python versions that are supported by this project
• Type validation via mypy
• All pre-commit hooks

tox

Alternatively, you can always run a single test. See documentation below.

Responses uses Pytest for testing. You can run all tests by:

tox -e py37
tox -e py310

OR manually activate required version of Python and run

pytest

And run a single test by:

pytest -k '<test_function_name>'

To verify type compliance, run mypy linter:

tox -e mypy

OR

mypy --config-file=./mypy.ini -p responses

To check code style and reformat it run:

tox -e precom

OR

pre-commit run --all-files