kaeedo/Scrutiny

Randomly test state machines (such as your UI) by randomly navigating through transitions

F# and C# library for testing state machines by randomly choosing available states and valid transitions. Designed for usage with UI tests

Description

Describe your UI as a state machine, and then use Scrutiny to simulate a "User" that randomly clicks around on your site. Scrutiny will attempt to create a Directed Adjacency Graph of your states, and then randomly choose an unvisited state to navigate to. It will repeat this process until all states have been visited. During each state, Scrutiny will attempt to run any defined actions within that state. Once all states have been visited, if an exit action has been defined it will then navigate there and quit. Scrutiny will then also generate an HTML file which visualizes the State Machine as a graph.

Scrutiny was designed to run UI tests, but using e.g. CanopyUI or Selenium is only an implementation detail. In theory, any state machine can be tested with Scrutiny.

---

There are several usage example projects in the usageExamples directory, implemented using different technologies. The first two are implemented in F#, and the third one in C#.

• Canopy UsageExample for a sample test implemented with CanopyUI
• Playwright UsageExample for a sample test implemented with PlaywrightSharp
• C# UsageExample for a sample test implementation also using Playwright, but this time with C#

A tiny sample site exists in the Usage Example directory. This is the website that the usage examples are testing. It features three pages, a home page, comment page, and a sign in page. A user can only leave a comment if they are signed in. The usage examples showcase a certain approach a developer can take as to how to model their web site as a state machine. In this case, the home and comment page are each listed twice, once as logged out, and once as logged in. This is only one way to handle this case, and the developer could choose to model it in any other way.

Scrutiny will also draw a diagram representing the system under test as has been modeled by the various pages. The Sample Web site looks like this:

Documentation

Click for F# documentation

Define one page object for each state in your UI. A state can be anything from a page, or an individual modal, or the same page as a different state, but altered, for example a logged in user.

The possible custom operations are:

• name: Name of the state. Required
• onEnter: Function to run when entering this page. Only one allowed
• onExit: Function to run when exiting this page. Only one allowed
• transition: Possible transition. Define how to transition to the next state, as well as which state to navigate to. Any number of transitions allowed
• action: Possible action. Define function to run while in this page state. Any number of actions allowed

A page looks like this:

let loggedInComment = fun (globalState: GlobalState) ->
    page {
        name "Logged In Comment"

        onEnter (fun ls ->
            printfn "Checking on page comment"
            "#header" == "Comments"
        )

        onExit (fun _ ->
            printfn "Exiting comment"
        )

        transition {
            via (fun ls -> click ls.HomeLink)
            destination home
        }
        transition {
            via (fun _ -> click "#signin")
            destination signIn
        }

        action {
            fn (fun _ -> () (*do something on the page*))
        }
        action {
            fn (fun _ -> () (*do something else on the page*))
        }
        action {
            isExit
            fn (fun _ -> () (*final action to perform before exiting the test*))
        }
    }

The name must be unique. Any number of transitions and any number of actions can be defined. The onEnter function will be executed everytime scrutiny transitions to this state, and onExit will execute everytime scrutiny leaves this state. name, onEnter, and onExit must be defined before any transitions and actions.

Any action can be be marked as isExit, and multiple pages can have an action that is the exit action. If multiple are defined, Scrutiny will randomly choose one to perform. The GlobalState in the example is any type defined in your test that you can use to pass data between states, e.g. Username or IsLoggedIn

actions are defined as follow within a page CE:

page {
    name "something"

    action {
        name "Name of action"
        dependantActions [ "Other action" ]
        isExit
        fn (fun _ -> (*This is the function that gets run*))
    }
}

The name defines a name for this action. Optional. This is how this action is reffered to when another action or transition depends on it The dependantActions list defines any actions that will be run before this action is run. Optional The isExit marks this action as a potential exit action. Optional The fn is the actual function to run as this action. Required

transitions are defined as follows within a page CE:

page {
    name "something"

    transition {
        dependantActions [ "Other action" ]
        via (fun _ -> (*how to transition to the next state*))
        destination otherPage
    }
}

The dependantActions list defines any actions that will be run before this action is run. Optional The via function is executed that will actually transition the state machine to the next state. Required The destination is the state that will be transitioned to. Required

Configuration

Some things can be configured via ScrutinyConfig. The default config is:

{ ScrutinyConfig.Seed = Environment.TickCount
  MapOnly = false
  ComprehensiveActions = true
  ComprehensiveStates = true
  ScrutinyResultFilePath = Directory.GetCurrentDirectory() + "/ScrutinyResult.html"
  Logger = printfn "%s" }

Seed is printed during each test to be able to recreate a specific test run. MapOnly won't run the test at all, but only generate the HTML Graph report. ComprehensiveActions will run ALL defined actions anytime it enters a state with actions defined. If false, it will run a random subset of actions. ComprehensiveStates will visit ALL states in the state machine. If this is false, then it will visit at least half of all states before randomly quitting. ScrutinyResultFilePath is the directory and specified file name that the generated HTML report will be saved in Logger is how individual messages from scrutiny will be logged. The signature is string -> unit. This is useful for things like XUnit that bring their own console logging mechanism, or if you wanted to integrate a larger logging framework.

To actually run the test, call the scrutinize function with your entry state, config, and global state object. e.g.

// Sample Global State. This can be anything, and all page states will receive the same instance
type GlobalState() =
    member val IsSignedIn = false with get, set
    member val Username = "MyUsername" with get, set
    member val Number = 42

[<EntryPoint>]
let main argv =
    let options = FirefoxOptions()
    do options.AddAdditionalCapability("acceptInsecureCerts", true, true)

    use ff = new FirefoxDriver(options)
    let currentDirectory = DirectoryInfo(Directory.GetCurrentDirectory())

    let config =
        { ScrutinyConfig.Default with
              Seed = 553931187
              MapOnly = false
              ComprehensiveActions = true
              ComprehensiveStates = true
              ScrutinyResultFilePath = currentDirectory.Parent.Parent.Parent.FullName + "/myResult.html" }

    // Start tests. In this case we're using CanopyUI, but can be any test runner e.g. XUnit or Expecto
    // Start CanopyUI tests
    "Scrutiny" &&& fun _ ->
        printfn "opening url"
        url "https://localhost:5001/home"

        let gs = GlobalState()

        // The call to start Scrutiny, and construct a graph and "click" through all states
        scrutinize config gs home
        // or
        // scrutinizeWithDefaultConfig gs home

    switchTo ff
    pin canopy.types.direction.Right

    run()
    quit ff

    0

At the end of the run, Scrutiny will return an object which contains the generated adjacency graph, as well as a list of individual steps taken, along with the actions performed in each state.

Important note for F# users

As the transitions ultimately depict a cyclic graph, it is necessary to declare module or namespace as recursive so that pages defined later can be referenced by pages earlier. Note the usage of the rec keyword. e.g.:

module rec MyPages =
    let firstPage = fun (globalState: GlobalState) ->
        page {
            name "First Page"
            transition {
                via (fun _ -> click "#second")
                destination secondPage
            }
        }

    let secondPage = fun (globalState: GlobalState) ->
        page {
            name "Second Page"
            transition {
                via (fun _ -> click "#first")
                destination firstPage
            }
        }

Migration v1 to v2

• Within a page computation expression, ensure that name is first, and that any onEnter and onExit functions are defined before any transitions and actions.
• transitions are now defined using a transition computation expression:
  • Before: transition ((fun _ -> click "#signin") ==> signIn)
  • After:

    transition {
        via (fun _ -> click "#signin")
        destination signIn
    }
    

• actions are now defined using an action computation expression:
  • Before: action (fun _ -> () /*do something on the page*/)
  • After:

    action {
        fn (fun _ -> () /*do something on the page*/)
    }
    

• exitActions are now defined as a regular action, but with the isExit property set:
  • Before: exitAction (fun _ -> () /*final action to perform before exiting the test*/)
  • After:

    action {
        isExit
        fn (fun _ -> () /*final action to perform before exiting the test*/)
    }
    

---

Click for C# documentation

Define one class for each state in your UI, and decorate it with the PageState attribute. A state can be anything from a page, or an individual modal, or the same page as a different state, but altered, for example a logged in user.

The possible attributes are:

• PageState: Define a class as a Page state.
• OnEnter: Function to run when entering this page. Only one allowed
• OnExit: Function to run when exiting this page. Only one allowed
• TransitionTo: Possible transition. Define how to transition to the next state, as well as which state to navigate to. Any number of transitions allowed
• Action: Possible action. Define function to run while in this page state. Any number of actions allowed. Optionally can be configured to be an exit action via the property IsExit
• DependantAction: Takes a string as a parameter. Only valid on Transitions and Actions. References an action that should be run before this action/transition. Multiple dependant actions can be referenced per action/transition

A PageState could look like this:

using Scrutiny.CSharp;

[PageState]
public class LoggedInComment
{
    private readonly GlobalState globalState;

    public LoggedInComment(GlobalState globalState)
    {
        // Construct anything necessary.
        // The constructor is called everytime Scrutiny navigates to this state
    }

    [OnEnter]
    public void OnEnter()
    {
        // Do something when scrutiny enters this state
        // Can optionally be async/await
        // Can only define one
    }

    [Action]
    public async Task WriteComments()
    {
        // Do something on the page
        // Can optionally be non-async
        // Define any number of these
    }

    [Action(IsExit = true)]
    public async Task ExitAction()
    {
        // One exit actions amongst all page states is chosen
        // Define any number of these
        // Can optionally be non-async
    }

    [ExitAction]
    public async Task ExitAction()
    {

    }

    [TransitionTo(nameof(AnotherState))]
    [DependantAction(nameof(WriteComments))] // Optioanlly run the WriteComments action before executing this transition
    public void TransitionToAnotherState()
    {
        // Code to perform state transition
        // Define any number of these
        // Can optionally be async/await
    }
}

Configuration

Some things can be configured via the Scrutiny.CSharp.Configuration.Configuration POCO. The default config is:

Seed = Environment.TickCount
MapOnly = false
ComprehensiveActions = true
ComprehensiveStates = true
ScrutinyResultFilePath = Directory.GetCurrentDirectory() + "/ScrutinyResult.html"
Logger = (Action<string>)((s) => Console.WriteLine(s))

Seed is printed during each test to be able to recreate a specific test run. MapOnly won't run the test at all, but only generate the HTML Graph report. ComprehensiveActions will run ALL defined actions anytime it enters a state with actions defined. If false, it will run a random subset of actions. ComprehensiveStates will visit ALL states in the state machine. If this is false, then it will visit at least half of all states before randomly quitting. ScrutinyResultFilePath is the directory and specified file name that the generated HTML report will be saved in Logger is how individual messages from scrutiny will be logged. This is useful for things like XUnit that bring their own console logging mechanism, or if you wanted to integrate a larger logging framework.

To actually run the test, call the Scrutiny.CSharp.Scrutinize.Start<Home>(gs, config) method. It takes your entry state as a generic type argument, and a constructed global state object as well as your config as parameters.

using Scrutiny.CSharp;

[Fact]
public async Task WithAttrs()
{
    var browser = await playwright.Firefox.LaunchAsync(headless: false);
    var context = await browser.NewContextAsync(ignoreHTTPSErrors: true);
    var page = await context.NewPageAsync();

    await page.GoToAsync("https://127.0.0.1:5001/home");

    var config = new Configuration
    {
        Seed = 553931187,
        MapOnly = false,
        ComprehensiveActions = true,
        ComprehensiveStates = true
    };

    var gs = new GlobalState(page, outputHelper);
    var result = Scrutinize.Start<Home>(gs, config);

    Assert.Equal(7, result.Steps.Count());
    Assert.Equal(5, result.Graph.Count());
}

The global state can be any class you want it to be. Scrutiny will pass the instance that is passed into the start around to each PageState it visits. At the end of the run, Scrutiny will return an object which contains the generated adjacency graph, as well as a list of individual steps taken, along with the actions performed in each state.

Migration v1 to v2

• [ExitAction] attribute removed. Set the IsExit property on an Action isntead"
  • Before:

    [ExitAction]
    public async Task ExitAction()
    

  • After:

    [Action(IsExit = true)]
    public async Task ExitAction()
    

---

Development

To run the usage examples, you must start the web project.

The HTML report is a single file with all javascript written inline

---

Sponsor

Thank you to Valora for sponsoring this project:

---

Donations

Donations are greatly appreciated, but not needed at all. Please only donate if you are in a position to be able to afford it, and only if you truly believe in the gift of giving.

Liberapay:

Click for cryptocurrency links

Ethereum: 0x05f231D19c19A2111fe03c923F26813Bad43B57f

Cardano ADA: addr1qx35nmy62dfp3n5tqgga92gxcnq5vkvflw963yg7fm5e5my68x9frc2qq0r8nstjtnjcrcnpmtpzwvp0sqz46y4ykrmqrd4dg9