Wonderland Screenshot Testing

Tests runner helping to prevent visual regressions in Wonderland Engine projects.

Learn more about Wonderland Engine at https://wonderlandengine.com.

Usage

Install via npm or yarn:

npm i --save-dev @wonderlandengine/screenshot-testing
# or:
yarn add @wonderlandengine/screenshot-testing --D

Add a test script in the package.json file:

{
    "scripts": {
        "test": "wle-screenshot-testing path/to/config.screenshot.json"
    }
}

Alternatively, you can reference the binary using:

./node_modules/.bin/wle-fidely path/to/config.screenshot.json

For more information about the CLI, have a look at the CLI Arguments section.

By default, the CLI looks for a file named config.screenshot.json in the working directory.

Basic Example

A simple setup contains a reference to compare to (reference.png), and a configuration file (config.screenshot.json):

MyWonderland.wlp
config.screenshot.json
deploy/
js/
test/
    reference.png

config.screenshot.json

{
    "scenarios": [{
        "readyEvent": "MyWonderland.bin",
        "reference": "./test/reference.png"
    }]
}

The readyEvent is used to take the screenshot once the scene MyWonderland.bin is loaded. The screenshot is then compared to the file ./test/reference.png.

When running the CLI for the first time, you might want to specify the screenshot size using:

wle-screenshot-testing config.screenshot.json --width 1920 --height 1080 --save

This will save all references with a size of 1920x1080. If no references are found and no size is specified, the runner will use a default size of 480x270.

For more information about the configuration, have a look at the Configuration File section.

Configuration File

Every project must have a configuration file:

{
    // If this timeout is reached, the test suite will fail.
    "timeout": 60000,

    // Width of the screenshot capture. Must match the image reference width.
    "width": 480,
    // Height of the screenshot capture. Must match the image reference height.
    "height": 270,

    "scenarios": [
        {
            // Default loading event: Wait for `MyScene.bin` to load
            "readyEvent": "MyScene.bin",
            // Reference image to compare against
            "reference": "./scene-loaded.png",
            // RMSE tolerance for the entire image
            "tolerance": 4,
            // Maximum authorized error per-pixel
            "perPixelTolerance": 16
        },
        {
            // Custom event sent from the application
            "event": "on-shoot",
            "reference": "./shoot.png"
        }
    ]
}

The test suite is made of multiple scenarios, associating a screenshot event to a reference image:

• readyEvent: Event sent after a scene load
• event: Programmatic custom event coming from the project. For more information, have a look at the Project section
• reference: Path to the reference file, i.e., the ground truth image to compare against

Custom Events

Screenshots are taken in an event-based fashion. The runner works in three steps:

1. Load and run the project
2. Listen for events coming from the project
3. Compare screenshots captured upon event to their reference

You can send events in your running application using:

Application

/* The argument represents the event id.
 * The id must match the `event` field in the configuration. */

player.shoot();
await window.testScreenshot('on-shoot');

game.showGameOver();
await window.testScreenshot('gameover');

The testScreenshot method returns a Promise. You must wait until the promise resolves before taking another screenshot.

Test Entry Point

In order to avoid mixing production & test code, it's advised to use a custom entry point (index.js) with the test runner.

This custom entry point can then reference components for the sole purpose of testing, e.g.,

if (isPlayerShooting()) {
    await window.testScreenshot('on-shoot');
    console.log('Test screenshot captured!');
}

Debugging

You can start the runner in watch mode:

wle-screenshot-testing config.screenshot.json -w

In watch mode, the browser will automatically open and kept alive even if tests are failing.

CLI Arguments

Argument	Type	Description
--save-on-failure	Flag	Overwrites failed reference(s) with the test(s) screenshot
--save	Flag	Save every screenshot
-o, --output	Path	Output folder for saved screenshots. References overwritten by default
-w, --watch	Flag	Watch, i.e., open the browser for each project
--logs	Path	Path to store the browser logs. If not provided, logs will be discarded