Testing programs with Cmds

elm-program-test allows you to test and simulate responses to Cmds produced by your Elm program. (The rest of this page walks through an example of how to do that.) Testing your program in this way gives a few notable benefits:

Your tests will more clearly describe how your app interacts with other systems. Steps in your test will read like "then the pending HTTP request to example.com fails with a network error" instead of "then the update function gets called with the FetchResponse (Err NetworkError) message".

Your tests will be more robust to refactoring of your implementation code. Because your tests don't need to directly refer to the internals of your program, your tests will still be valid even if you make drastic changes to the architecture of your code (like extracting new modules, renaming types or variants, or entirely reworking the flow of messages within your program).


elm-program-test's features for testing effects depend on being able to know what a given effect value represents. In the Elm architecture, effects are normally represented by the Cmd msg type. Unfortunately, Elm does not currently provide a way to inspect Cmd values to see which commands they represent. There's some extra work required to get around this current limitation, which the rest of this page will walk you through.

Hopefully you'll agree that the benefits (listed above) of testing your program in this way will be worth the extra effort! But be aware of this trade-off when considering whether this type of testing is right for your program.

Introducing the example program

The rest of this page will be working with an example program that interacts with a fictional API for controlling the lights in your house.

This diagram shows the architecture of the example application. The Elm code makes HTTP requests to two different endpoints: /v1/devices to get a list of all the lights that can be controlled, and /v1/devices/:id to change the setting of a specific light.

architecture diagram described in the preceding paragraph

Goal: the ideal test

The test we want to write represents the following scenario:

  1. The user loads the main page
  2. The page requests the list of lights
  3. The server responds with the list of lights and their current states
  4. The user clicks "Turn on" for the kitchen lights
  5. We expect the page to send a request to the server with the correct values

Here's what that test will look like in code:

import Expect
import ProgramTest
import Test exposing (test)

test "controlling a light" <|
    \() ->
            |> ProgramTest.simulateHttpOk
                """[{"id":"K001", "name":"Kitchen", "dimmable":false, "value":0}]"""
            |> ProgramTest.clickButton "Turn on"
            |> ProgramTest.expectHttpRequest
                (.body >> Expect.equal """{"value":1}""")

However, running this test produces the following failure:

↓ HomeAutomationExampleTest
✗ controlling a light

    TEST SETUP ERROR: In order to use simulateHttpResponse,
    you MUST use ProgramTest.withSimulatedEffects before calling ProgramTest.start


As the error explains, we must use ProgramTest.withSimulatedEffects before the test will work. But using it requires us to provide a function of type effect -> ProgramTest.SimulatedEffect. Because the effect type of our program (the type that init and update return as the second item in the tuple) is currently Cmd Msg (which Elm does not currently allow us to inspect the values of), we'll need to do the following before the test will work:

  1. Make a new type which can represent all the effects our program can produce.
  2. Implement the required effect -> ProgramTest.SimulatedEffect function.
  3. Use withSimulatedEffects to set up the test.
  4. ✅ Watch the test pass!

Making it work

Planning the change

Since Elm does not currently allow us to inspect Cmd values, the first thing we need to do is define a new type that we can inspect that can represent all the effects that our update and init functions want to produce. Our init and update functions will change to use this new type, and we'll also need to define a function that can turn this new type into Cmds for when our program is compiled for production.

So in our existing main module:

init : Flags -> ( Model, Cmd Msg )
update : Msg -> Model -> ( Model, Cmd Msg )

will change to:

type Effect

init : Flags -> ( Model, Effect )
update : Msg -> Model -> ( Model, Effect )

perform : Effect -> Cmd Msg

Defining an "Effect" type

Taking a look through the init and update functions, the following Cmds are being produced — the new type we make will need to have a variant for each one:

  • Cmd.none
  • loadDeviceList, which itself uses Http.get
  • changeLight, which itself uses Http.post

Each variant we create should contain just enough information to call the functions necessary to create the corresponding Cmd. Following that guideline, we'll end up with an Effect type like this and a corresponding perform function:

type Effect
    = NoEffect
    | GetDeviceList
        { url : String
        , decoder : Json.Decode.Decoder (List Light)
        , onResult : Result Http.Error (List Light) -> Msg
    | ChangeLight
        { url : String
        , body : Json.Encode.Value
        , decoder : Json.Decode.Decoder Light
        , onResult : Result Http.Error Light -> Msg

perform : Effect -> Cmd Msg
perform effect =
    case effect of
        NoEffect ->

        GetDeviceList { url, onResult, decoder } ->
                { url = url
                , expect = Http.expectJson onResult decoder

        ChangeLight { url, onResult, decoder, body } ->
                { url = url
                , body = Http.jsonBody body
                , expect = Http.expectJson onResult decoder

After changing update and init to return the corresponding Effect instead of a Cmd, the final change to make the program compile again is to update our main to make use of the perform function:


main : Program Flags Model Msg
main =
        { init =
            \flags ->
                init flags
                    |> Tuple.mapSecond perform
        , update =
            \msg model ->
                update msg model
                    |> Tuple.mapSecond perform
        , subscriptions = subscriptions
        , view = view

Simulating effects

Now that our update and init functions use an effect type that we can inspect the values of, we can write a function for our test suite that can convert our new Effect type into simulated effects that elm-program-test can understand:

import SimulatedEffect.Cmd
import SimulatedEffect.Http

simulateEffects : Main.Effect -> ProgramTest.SimulatedEffect Main.Msg
simulateEffects effect =
    case effect of
        Main.NoEffect ->

        Main.GetDeviceList { url, onResult, decoder } ->
                { url = url
                , expect = SimulatedEffect.Http.expectJson onResult decoder

        Main.ChangeLight { url, onResult, decoder, body } ->
                { url = url
                , body = SimulatedEffect.Http.jsonBody body
                , expect = SimulatedEffect.Http.expectJson onResult decoder

You may notice that this looks extremely similar to the perform function we wrote earlier. The SimulatedEffect.* modules are designed to parallel the core Elm modules that produce Cmds so that writing your simulateEffects function is as straightforward as possible.

One last step! Now that we have a function of the required type, we can use it to enable effect simulation in our tests:


start : ProgramTest Main.Model Main.Msg Main.Effect
start =
        { init = Main.init
        , update = Main.update
        , view = Main.view
        |> ProgramTest.withSimulatedEffects simulateEffects
        |> ProgramTest.start ()

Now that we've enabled effects simulation, our original test will successfully run!

You may have noticed that the test itself never directly refers to the effect values (GetDeviceList and ChangeLight) that we defined. elm-program-test handles that for you — when a ProgramTest receives an effect from your init or update functions, it will use the simulateEffects function we provided to understand what the effect means and update the state of the test accordingly. You don't need to worry about the details of how this works — just know that as long as you provide a simulateEffects function you'll be able to use the full API provided by ProgramTest for testing HTTP requests.

Try it out

You can play around with the example here: https://ellie-app.com/6wsbBjZ9z2Ma1 Try editing the tests or the implementation to make the tests fail and see what kind of failure messages you get.