Mox is a library for defining concurrent mocks in Elixir.

The library follows the principles outlined in "Mocks and explicit contracts", summarized below:

1. No ad-hoc mocks. You can only create mocks based on behaviours

2. No dynamic generation of modules during tests. Mocks are preferably defined in your test_helper.exs or in a setup_all block and not per test

3. Concurrency support. Tests using the same mock can still use async: true

4. Rely on pattern matching and function clauses for asserting on the input instead of complex expectation rules

Example

As an example, imagine that your project defines a WeatherAPI behaviour:

defmodule MyApp.WeatherAPI do
  @callback temp(MyApp.LatLong.t()) :: {:ok, integer()}
  @callback humidity(MyApp.LatLong.t()) :: {:ok, integer()}
end

If you want to mock the WeatherAPI behaviour during tests, the first step is to define the mock with defmock/2, usually in your test_helper.exs:

Mox.defmock(MyApp.MockWeatherAPI, for: MyApp.WeatherAPI)

Now in your tests, you can define expectations with expect/4 and verify them via verify_on_exit!/1:

defmodule MyApp.HumanizedWeatherTest do
  use ExUnit.Case, async: true

  import Mox

  # Make sure mocks are verified when the test exits
  setup :verify_on_exit!

  test "gets and formats temperature and humidity" do
    MyApp.MockWeatherAPI
    |> expect(:temp, fn {_lat, _long} -> {:ok, 30} end)
    |> expect(:humidity, fn {_lat, _long} -> {:ok, 60} end)

    assert MyApp.HumanizedWeather.temp({50.06, 19.94}) ==
      "Current temperature is 30 degrees"
    assert MyApp.HumanizedWeather.humidity({50.06, 19.94}) ==
      "Current humidity is 60%"
  end
end

In practice, you will have to ensure that during tests, the code you're testing uses the mock rather than the actual implementation. If the system under test relies on application configuration, you should set it before the test suite starts so that you can use async: true. You can do this in your config files:

config :my_app, :weather_api, MyApp.MockWeatherAPI

Or in your test_helper.exs:

Application.put_env(:my_app, :weather_api, MyApp.MockWeatherAPI)

All expectations are defined based on the current process. This means multiple tests using the same mock can still run concurrently unless the Mox is set to global mode. See the "Multi-process collaboration" section.

Especially if you put the mock into the config/application environment you might want the implementation to fall back to a stub (or actual) implementation when no expectations are defined. stub_with/2 is just what you need! You can use it in your test (or in setup) as follows:

Mox.stub_with(MyApp.MockWeatherAPI, MyApp.StubWeatherAPI)

Now, if no expectations are defined it will call the implementation in MyApp.StubWeatherAPI.

Multiple behaviours

Mox supports defining mocks for multiple behaviours.

Suppose your library also defines a behaviour for getting past weather:

defmodule MyApp.PastWeatherAPI do
  @callback past_temp(MyApp.LatLong.t(), DateTime.t()) :: {:ok, integer()}
end

You can mock both the weather and past weather behaviour:

Mox.defmock(MyApp.MockWeatherAPI, for: [MyApp.WeatherAPI, MyApp.PastWeatherAPI])

Compile-time requirements

If the mock needs to be available during the project compilation, for instance because you get undefined function warnings, then instead of defining the mock in your test_helper.exs, you should instead define it under test/support/mocks.ex:

Mox.defmock(MyApp.MockWeatherAPI, for: MyApp.WeatherAPI)

Then you need to make sure that files in test/support get compiled with the rest of the project. Edit your mix.exs file to add the test/support directory to compilation paths:

def project do
  [
    ...
    elixirc_paths: elixirc_paths(Mix.env),
    ...
  ]
end

defp elixirc_paths(:test), do: ["test/support", "lib"]
defp elixirc_paths(_),     do: ["lib"]

Multi-process collaboration

Mox supports multi-process collaboration via two mechanisms:

1. explicit allowances
2. global mode

The allowance mechanism can still run tests concurrently while the global one doesn't. We explore both next.

Explicit allowances

An allowance permits a child process to use the expectations and stubs defined in the parent process while still being safe for async tests.

test "invokes add and mult from a task" do
  MyApp.MockWeatherAPI
  |> expect(:temp, fn _loc -> {:ok, 30} end)
  |> expect(:humidity, fn _loc -> {:ok, 60} end)

  parent_pid = self()

  Task.async(fn ->
    MyApp.MockWeatherAPI |> allow(parent_pid, self())

    assert MyApp.HumanizedWeather.temp({50.06, 19.94}) ==
      "Current temperature is 30 degrees"
    assert MyApp.HumanizedWeather.humidity({50.06, 19.94}) ==
      "Current humidity is 60%"
  end)
  |> Task.await
end

Note: if you're running on Elixir 1.8.0 or greater and your concurrency comes from a Task then you don't need to add explicit allowances. Instead $callers is used to determine the process that actually defined the expectations.

Global mode

Mox supports global mode, where any process can consume mocks and stubs defined in your tests. set_mox_from_context/0 automatically calls set_mox_global/1 but only if the test context doesn't include async: true.

By default the mode is :private.

setup :set_mox_from_context
setup :verify_on_exit!

test "invokes add and mult from a task" do
  MyApp.MockWeatherAPI
  |> expect(:temp, fn _loc -> {:ok, 30} end)
  |> expect(:humidity, fn _loc -> {:ok, 60} end)

  Task.async(fn ->
    assert MyApp.HumanizedWeather.temp({50.06, 19.94}) ==
      "Current temperature is 30 degrees"
    assert MyApp.HumanizedWeather.humidity({50.06, 19.94}) ==
      "Current humidity is 60%"
  end)
  |> Task.await
end

Blocking on expectations

If your mock is called in a different process than the test process, in some cases there is a chance that the test will finish executing before it has a chance to call the mock and meet the expectations. Imagine this:

test "calling a mock from a different process" do
  expect(MyApp.MockWeatherAPI, :temp, fn _loc -> {:ok, 30} end)

  spawn(fn -> MyApp.HumanizedWeather.temp({50.06, 19.94}) end)

  verify!()
end

The test above has a race condition because there is a chance that the verify!/0 call will happen before the spawned process calls the mock. In most cases, you don't control the spawning of the process so you can't simply monitor the process to know when it dies in order to avoid this race condition. In those cases, the way to go is to "sync up" with the process that calls the mock by sending a message to the test process from the expectation and using that to know when the expectation has been called.

test "calling a mock from a different process" do
  parent = self()
  ref = make_ref()

  expect(MyApp.MockWeatherAPI, :temp, fn _loc ->
    send(parent, {ref, :temp})
    {:ok, 30}
  end)

  spawn(fn -> MyApp.HumanizedWeather.temp({50.06, 19.94}) end)

  assert_receive {^ref, :temp}

  verify!()
end

This way, we'll wait until the expectation is called before calling verify!/0.