Responses Integration¶

This library can be used to load http responses from files into the Responses package. While Responses has a native file loading capability in beta, Pytest-Scenario-Files can also load the responses from its own files.

Basic Usage¶

There are three steps to using the Responses integration:

Create the data files.
Pass the psf_responses fixture as a parameter to your test function.
Activate the Responses integration using the command line flags.

Data file format¶

Data to be loaded into responses should be put into fixtures whose names end with _response or _responses. For example, you might have a fixture named oauth2_response or a fixture named api_responses. These fixtures will not actually be created or parameterized for the test. Instead, each will be loaded into the psf_responses fixture and removed from the list of fixtures.

Each response should be structured as a dictionary that contains two required and five optional keys:

scenario_1:
  oauth2_response:
    method: GET|POST|PUT|etc. (required)
    url: https://www.example.com/rest/endpoint (required)
    status: 200 (optional)
    body: Text body of the http response (optional)
    json: (optional)
      key1: value1
      key2: value2
      key3: value3
    content_type: text/plain (optional)
    headers: (optional)
      header1: header-value-1
      header2: header-value-2

body and json are mutually exclusive and you should only use one of the two in a response.
If you are passing in json you should not pass in a content_type header as it should be set to application/json automatically.

The fixture may contain a list of responses in the same format:

scenario_2:
  api_responses:
  - method: GET|POST|PUT|etc.
    url: https://www.example.com/rest/endpoint
    body: Text body of the http response
  - method: GET|POST|PUT|etc.
    url: https://www.example.com/rest/endpoint2
    body: Text body of the http response2
  - method: GET|POST|PUT|etc.
    url: https://www.example.com/rest/endpoint3
    body: Text body of the http response3

All of the responses in the list will be loaded into a RequestsMock. While the response loading recognizes both _response and _responses, there is no actual difference in how they are handled. They underlying code checks to see whether the loaded value is a dict or a list and handles it accordingly. Having both suffixes is just to make reading the data files easier for humans.

The `psf-responses` fixture¶

Pytest-Scenario-Files provides a psf_responses fixture that is used to load the responses. It returns a currently active RequestsMock object that has all of the responses from the data files for the current test already loaded. If all of the responses your test needs are already loaded via the data files you can just leave it be. However, If you need to add additional responses or to change a response for this particular test you can use it as you would any standard RequestsMock.

def test_api_call(psf_responses):
    http_result = requests.get("https://www.example.com/rest/endpoint")
    assert http_result.status_code = 200

Command line flags¶

There are two command line flags for Pytest that are used for the Responses integration:

--psf-load-responses

This turns on the integration. Since the fixtures intended for use with Responses integration are marked by a special suffix, the integration should be explicitly triggered to avoid accidentally activating it for a developer who uses the suffix without realizing the special meaning.
--psf-fire-all-responses=[true|FALSE]

This allows you to turn on the flag assert_all_requests_are_fired for Responses. It defaults to false.

Advanced Usage¶

Overriding a response¶

You can use the psf_responses fixture to override a response for a particular test. Use the replace() or upsert() methods to do this. The replacement can be done in a separate fixture or in the test function itself. If you are doing this in a separate fixture the convention is to return the RequestsMock as the fixture value so that you can chain together multiple fixtures that add or alter the responses for a test.

@pytest.fixture
def error_response(psf_responses):
    psf_responses.replace(
        "GET",
        "https://www.example.com/rest/endpoint3",
        status=401
    )
    return psf_responses

def test_endpoint_3_error(error_response):
    http_result = requests.get("https://www.example.com/rest/endpoint3")
    assert http_result.status_code = 401

data_endpoint_3_error.yaml¶

api_call_scenario:
  api_responses:
  - method: GET
    url: https://www.example.com/rest/endpoint
    body: Text body of the http response
  - method: GET
    url: https://www.example.com/rest/endpoint2
    body: Text body of the http response2
  - method: GET
    url: https://www.example.com/rest/endpoint3
    body: Text body of the http response3

Use with `moto` when mocking AWS¶

If you are using the package moto to mock out AWS services, note that it uses Responses under the hood and sets its own RequestsMock. This will cause your own responses to not be found. You will need to call override_responses_real_send() as per the moto FAQ.

from moto.core.models import override_responses_real_send

def test_some_func(psf_responses):
    override_responses_real_send(psf_responses)
    ...

Usage with the `psf_expected_result` fixture¶

You can set up a data file with the generally expected response for a specific URL, then override the response to check error conditions. Here is an example using a file with the standard API response and a test that checks both a successful and an unsuccessful test of the API.

This first file contains the basic API responses, which are loaded by reference for each scenario:

all_api_responses.yaml¶

api_testing:
  api_responses:
  - url: https://www.example.com/rest/endpoint
    method: GET
    status: 200
    body: The call was successful.

The second file contains the scenarios, success and failure. The success scenario just runs through the call and contains no overrides. The failure scenario specifies that the call should return a 403 error and catch a responses.HTTPError exception:

data_api_check_full.yaml¶

success_scenario:
  api_responses: __all_api_responses.yaml:api_testing:api_responses
  psf_expected_result_indirect: The call was successful.
failure_scenario:
  api_responses: __all_api_responses.yaml:api_testing:api_responses
  response_override_indirect:
    url: https://www.example.com/rest/endpoint
    method: GET
    status: 403
    body: Access denied.
  psf_expected_result_indirect:
    expected_exception_type: requests.HTTPError

The third file is the Python unit tests. It has a fixture response_override() that will set up any overrides specified by the scenario. If the scenario has no overrides then it will just return the psf_responses fixture unchanged.

test_api.py¶

@pytest.fixture
def response_override(request, psf_responses):
    if hasattr(request, "param") and isinstance(request.param, dict):
        psf_responses.upsert(**request.param)
    return psf_responses

def test_api_check(response_override, psf_expected_result):
    with psf_expected_result as expected_result:
        api_call_result = requests.get("http://www.example.com/rest/endpoint")
        api_call_result.raise_for_status()
        assert api_call_result.body == "The call was successful."

When the test is run the first time (success_scenaro), Responses will return a 200 response with a body of “The call was successful.” — which is the expected value from the psf_expected_result fixture.

When the test is run the second time (failure_scenario), Responses will return a 403 response. raise_for_status() will then raise an exception requests.HTTPError, which will be caught by the context manager since the psf_expected_value fixture will return a pytest.raises(requests.HTTPError) context manager object. Any other kind of error or exception will cause the test to fail.

Detailed Example¶

Putting all this together is easiest to see using a detailed example. One system I work with (the NetBrain API) requires that you make four calls when you connect to it.

Authenticate and get an access token.
Get the list of available tenants and their tenant IDs.
Get the list of available domains and their domain IDs for the specified tenant.
Set the tenant and domain to be used for the current session.

In addition to checking for an HTTP error code of 4xx or 5xx, you also need to check the status code in the response JSON. 790200 generally means the API call succeeded while anything else means it failed.

The complete example (with the API connection code, test code, and data files) is contained in the source repository in the tests/Responses_example directory. Some highlights of this example are:

The common_test_data.yaml file. This holds a common set of responses that are used as a base by all of the tests.

common_scenario_data:
  common_responses:
    - method: POST
      url: https://netbrain-api.example.com/ServicesAPI/API/V1/Session
      status: 200
      json:
        statusCode: "790200"
        token: mock_token
    - method: GET
      url: https://netbrain-api.example.com/ServicesAPI/API/V1/CMDB/Tenants
      status: 200
      json:

Multiple scenarios (both success and failure) in each data file, covering both the happy (successful) path and any error paths through the code.
Each failure scenario uses a custom fixture url_response_override along with data from the file to give an error response.

url_response_override_indirect:
  method_or_response: GET
  url: https://netbrain-api.example.com/ServicesAPI/API/V1/CMDB/Domains
  status: 403
  json:
    statusCode: "795000"

Each failure scenario uses the psf_expected_result fixture with a dict containing a item with the key expected_exception_name to indicate the expected failure mode.

psf_expected_result_indirect:
  expected_exception_name: requests.HTTPError

Use of a regular, un-parameterized fixture that is used to prepare a NetbrainConnection object for each test.

@pytest.fixture
def netbrain_connection_obj() -> NetBrainConnection:
    return NetBrainConnection("username", "mock_password",
        "mock_tenant_name", "mock_domain_name")

Running all of the tests will give you complete coverage for the api_connection.py file.