Skip to main content

· 8 min read
Justin Young

I have been in need of an AI system that can execute a simulation that I desire to run. In my research, I have come across Goal-Oriented Action Planning. This technique can give me the flexibility I need to run my simulation, let's dive into the implementation a bit.

Link to GOAP Demo

GOAP, what is it

Goal-Oriented Action Planning, or GOAP, is a flexible AI technique that enables the developer build up a set of actions and objectives, and allows the NPC (agent) itself determine what the best objective is, and how to accomplish said objective.

GOAP includes the use of Agents, Goals, Actions, and State, to plan out its next series of decisions and activities. This is a useful system for Non-Playable Characters(NPCs) or enemy AI logic.

Quick History

GOAP was developed by Jeff Orkin in the early 2000's while working on the AI system for F.E.A.R.

The desire was to generate automated planning sequences for Enemies and NPCs to create a more immersive game experience.

GOAP can be considered an alternative to classic behavioral trees, which was more standard at that time.

Theory of operations

There are 5 aspects of GOAP that interact to create the magic: State, Agents, Goals, Actions, and the Planner.

flow of GOAP

First, let's talk about State.

State

State is the data set conditions that describes the world in which an agent exists. For my implementation, an established set of key/value pair data was used to fuel the simulation. A simple example of a world state:

ts
world = {
trees: 3;
bears: 1;
playerPosition: {x: 100, y:200};
};
ts
world = {
trees: 3;
bears: 1;
playerPosition: {x: 100, y:200};
};

This is the data that gets used, not only as a starting point, but gets cloned and mutated over the course of the algorithm processing the plan.

Goals

Next, let us review the goals or objectives that are intended to be accomplished. The goal defines the target state that the algorithm evaluates against to determine if the objectives are met.

The goal assessment will take a copy of the mutated state and compare it against the target state defined for the goal, and if it matches, let's the algorithm know that is done with that branch of evaluation.

The goal also contains a method of assessing its own priority conditions, which takes in the world state and returns a defined factor of prioritization. For example, a floating-point value from 0.0 to 1.0, where 1.0 is the highest priority.

Agents

Agencts are the entities (enemies or other NPCs) that get the planning system attached to it. If the entity is not currently executing a plan, it can call the planning process to assess what it should do next.

One aspect of the agents that is important to remember, is including the ability to cancel the plan, and reassess, even if the sequence isn't complete.

Think about if the environment in which the current plan was created, no longer is viable, you need to be able to change your mind. i.e. a power up is no longer available, or a targeted enemy is dead, etc...

Actions

Actions are very discrete units of activity:

  • Move to spot
  • Pick up Item
  • Fire weapon
  • Duck

These actions should have a cost component, time or energy is common, and the actions will be linked together to form a sequence of actions that constitutes 'plan'.

What is unique about components of an action beyond cost, is the precondition and effect components. These are super important.

The precondition component is what the planner evaluates if the action is viable under the current condition. The current condition is the cloned, mutated state that is considered for that sequence of the plan.

If the conditions are true for the precondition, then the action is considered an available choice for the next step.

The effect component of an action is the defined changes to state that occur when that action is executed. This is used by the planner to clone and mutate the state as it churns through the different possible options.

Planner

The planner is the algorithm which generates the plan, and it has several tasks. To use the Planner, you pass the current world state, all available actions for the agent, all available goals for the agent.

The planner's first task is to assess all available goals for the agent to determine which is the highest priority.

Then, with that goal selected and the current world state, find a list of actions that can be executed.

With your state, your goal, and your available actions, you can start building a graph network, a branching tree of available plans.

When the graph is constructed, the planner can assess the best course of action based on any number of custom factors, and returns the sequence of actions as the plan.

The algorithm

There are two aspects of the algorithm that should be discussed. The graph network and the assessment.

The graph network

Building the Graph

The graph network is built with a recursion that forms a tree structure, and branching is based on the new available list of actions that meet the mutated state condition, for that branch.

As you walk through each branch, the actions taken at each node will mutate the state. That mutated stated then gets checked against the goal provided, to see if you are done.

If the goal passes, an endnode is created. If not, then that newly mutated state is used to generate the new list of available actions and the recursion continues.

The recursion ends when a branch's mutated state cannot create further list of actions, or the goal is met.

Picking a plan

Once the graph is assembled, you can filter out any branches that do not end in a completed goal, then the Planner can assess which path makes most sense.

This is where you can have different style planners. The planner i created simply creates a 'cheapest cost' plan based on the the aggregate cost of each plan created.

I use a dijkstra's algorithm to calculate, based on each actions 'cost', the cheapest path to execute.

But there is flexibility here as well, including using different costing structures, maybe you want to balance energy and time both? Then you could construct a planner that favors one over the other based on conditions.

The test

Demo Test

Link to Demo

I spent a couple weeks building a simulation of my GOAP library that I created. It is a simple "actor feeds fire with wood, while avoiding bears" simulation.

The Actor has two goals, "keep fire aive", and "avoid bear"

If the actor is currently without a plan to execute, it passes its worldstate into the planner. The world state looks vaguely like this:

ts
export const world = {
tree: 500,
tree2: 500,
tree3: 500,
campfire: 0,
player: 0,
playerState: playerState.idle,
bearDistance: 300,
};
ts
export const world = {
tree: 500,
tree2: 500,
tree3: 500,
campfire: 0,
player: 0,
playerState: playerState.idle,
bearDistance: 300,
};

The actions available are:

ts
player.goapActions = [
feedFireAction,
collectWoodAction,
moveToTreeAction,
moveToFireAction,
moveToTree2Action,
collectWood2Action,
moveToTree3Action,
collectWood3Action,
runAwayAction,
relaxAction,
];
ts
player.goapActions = [
feedFireAction,
collectWoodAction,
moveToTreeAction,
moveToFireAction,
moveToTree2Action,
collectWood2Action,
moveToTree3Action,
collectWood3Action,
runAwayAction,
relaxAction,
];

When the planner is fed these components, it assesses the priority of each action based on their weighting, is the fire getting low? or is the bear close by?

With the goal selected, it uses the state data to determine which actions to take, for the fire building, the first round of actions usually are moving to trees. That is unless the player is holding some wood, then it will decide to just go to the fire directly.

If the player moves to a tree, it then collects its wood, then it moves to fire, and feeds the fire, and it waits till the fire gets lower before going to collect more wood.

I mentioned earlier that agents have to be able to cancel their plans. If the bear comes close to the player, it triggers a cancelPlan() method and the player is forced to generate a new plan.

Since the bear is close, it picks "avoid bear" plan, and then the process starts again with that new goal.

Conclusion

We have covered GOAP, some history of it, what the components of a GOAP system are, and how to implement them.

What I have learned in this process is that GOAP is very powerful and flexible. That does not imply that GOAP is easy, I would consider implementing a GOAP system at the intermediate level.

When trying to connect different actions and insuring they chain together to form a complete plan, there are many chances in implementation to create issues. But when dialed in, GOAP can provide a foundation for a very flexible AI system that can lead to enriching gameplay.

· 3 min read
Jae Edeen

The Excalibur team has built a number of games over the years. During our last few game jams, we started paper prototyping soon after our brainstorming process. It’s been working great and we highly recommend giving it a try!

More flexibility

Changing rules or mechanics that only exist on paper is a lot faster than having to adjust any code you may have written for those changes. It’s possible to alter your game without much worry, because you’re operating primarily in the realm of imagination, rather than within the constraints of your software development environment.

Paper prototyping can also help avoid the “sunk-cost fallacy”, which encourages you to stick with whatever you’ve spent a lot of time on just because you’ve spent a lot of time on it. Instead, you can change as much or as little as you wish without having to worry about deleting a bunch of code that you’ve already written.

Identify problems early

You also have the opportunity to fix game design problems before you've devoted time to implementing them in the actual software. While we were paper prototyping Sweep Stacks, we uncovered a game design complication with how the board filled up over time. Without having to write any code, we were able to determine a solution for the issue and implement it directly when we started programming the game.

Easier once you start writing code

If you’ve spent time prototyping, you'll have a more concrete idea of what you want your game to be. When you actually start writing code, you can begin with a more specific idea of what you want to accomplish. We’ve found it’s much easier to visualize and architect our code when we have a clear idea of how the rules and systems of a game will work together.

Virtual paper

While it's called “paper prototyping”, this process doesn't literally have to be done with paper, or any physical components at all. Virtual paper prototyping can be just as effective, and allows you to collaborate more easily with remote teammates. There are plenty of wireframing and “virtual tabletop” web apps out there that you can use to put together a digital prototype for your game (we usually use Excalidraw).

Give paper prototyping a try

The next time you’re working on a game, try doing some prototyping before you write any code. Adjust your rules, modify your designs, and dream of what you want to build.

· 6 min read
Erik Onarheim

Updated to the latest Excalibur, Capacitor.js & Parcel!

In this post we put a web canvas game built in Excalibur into an Android (or iOS) app with Capacitor.js!

In the past I would have used something like Cordova, but this new thing from the folks at Ionic has TypeScript support out of the box for their native APIs and support for using any Cordova plugins you might miss.

TLDR show me the code

Capacitor Setup

The capacitor project setup is pretty straightforward from their docs, it can drop in place in an existing project or create a brand new project from scratch.

I opted for the brand new project:

> npm init @capacitor/app
> npm init @capacitor/app

Then follow their wizard and instructions to configure.

After that step add the platforms you're interested in, in this case Android

> npx cap add android
> npx cap add android

I recommend reading the capacitor documentation on workflow with a hybrid native app. The gist is this

  1. Run npx cap sync to copy your web project into capacitor
  2. Run npx cap run android to start the project on android (or start in the Android SDK)

Android Setup

Before you try to run the project

  1. Download Android Studio Android Studio
  2. Open it up and check for updates if needed (first time initialization takes some time)
  3. Accept your SDK package licenses, the easiest way I've found to do this is with the SDK command line tools with Powershell on W.
    1. Find the SDK Manager Android Studio SDK Manager
    2. In SDK Tools, check Android SDK Command-line Tools SDK Tools Command-line
  4. Next we need to accept licenses. Android SDK location
    • In powershell, navigate to the Android SDK Location for command line tools C:\Users\<username>\AppData\Local\Android\Sdk\cmdline-tools\latest\bin
    • Set your java home temporarily $env:JAVA_HOME = 'C:\Program Files\Android\Android Studio\jre'
    • Run .\sdkmanager.bat --licenses and select y for each

Starting the App

Now that we have Android all setup we can start the app with the capacitor command line.

The gist is that it copies the final compiled html/css/js assets from your favorite frontend frameworks and build tools into the native container

> npx cap sync
> npx cap sync

After that we can open it in Android Studio with the capacitor commandline

> npx cap open android
> npx cap open android

Building the project and running the first time can take some time, so be patient after hitting the big green play button.

Android Studio Start Bar with Green Play Triangle Button

ProTipTM The Emulator is MEGA slow to start so once you get it on, leave it on. You can redeploy the app to a running emulator with the "re-run" hightlighted below.

Android Studio Restart Activity Button

If your Android emulator crashes on the first try like mine did with something like The emulator process for AVD Pixel_3a_API_30_x86 was killed, this youtube video was super helpful. For me the problem was disk space, the AVD needs 7GBs of disk space to start so I had to clean out some junk on the laptop 😅

Building Your Canvas Game

The dev cycle is pretty slick, run npm cap copy android to move your built JS living in the www to the right android folder. The default app looks like this after running it in the android emulator.

Default Capacitor screen on Android emulator

Setting Up Your JS Build

First let's setup our TypeScript by installing and creating an empty tsconfig.json

> npm install typescript --save-dev --save-exact
> npx tsc --init`
> npm install typescript --save-dev --save-exact
> npx tsc --init`

Recently I've been a big fan of parcel(v1) for quick and easy project setup, and it works great with excalibur also webpack is cool too if you need more direct control of your js bundling.

> npm install parcel --save-dev --save-exact
> npm install parcel --save-dev --save-exact

I copied the generated manifest.json, index.html, and css/ folder out of the original generated www/ and put it into game/.

Folder structure of capacitor frontend project

We need to setup our development and final build script in the package.json. The npm "start" script tells parcel to run a dev server and use game/index.html as our entry point to the app and follow the links and build them (notice the magic inline <script type="module" src="./main.ts"></script>) ✨

html
<!DOCTYPE html>
<html lang="en" dir="ltr">
<head>
<meta charset="UTF-8">
<title>Game Test</title>
<meta name="viewport" content="viewport-fit=cover, width=device-width, initial-scale=1.0, minimum-scale=1.0, maximum-scale=1.0, user-scalable=no">
<meta name="format-detection" content="telephone=no">
<meta name="msapplication-tap-highlight" content="no">
<link rel="manifest" href="./manifest.json">
<link rel="stylesheet" href="./css/style.css">
</head>
<body>
<script type="module" src="./main.ts"></script>
</body>
</html>
html
<!DOCTYPE html>
<html lang="en" dir="ltr">
<head>
<meta charset="UTF-8">
<title>Game Test</title>
<meta name="viewport" content="viewport-fit=cover, width=device-width, initial-scale=1.0, minimum-scale=1.0, maximum-scale=1.0, user-scalable=no">
<meta name="format-detection" content="telephone=no">
<meta name="msapplication-tap-highlight" content="no">
<link rel="manifest" href="./manifest.json">
<link rel="stylesheet" href="./css/style.css">
</head>
<body>
<script type="module" src="./main.ts"></script>
</body>
</html>

In this setup I'm sending all my built output with --dist-dir into the www directory, which is what capacitor will copy to android. I went ahead and deleted the provided default app in the www directory.

json
/* package.json */
{
"name": "my-cool-game",
"scripts": {
"start": "parcel game/index.html --dist-dir www",
"typecheck": "tsc -p . --noEmit",
"build": "parcel build game/index.html --dist-dir www"
}
...
}
json
/* package.json */
{
"name": "my-cool-game",
"scripts": {
"start": "parcel game/index.html --dist-dir www",
"typecheck": "tsc -p . --noEmit",
"build": "parcel build game/index.html --dist-dir www"
}
...
}

Vanilla Canvas code

To start with I have a really awesome game that shows the fps and a red square. This shows how get started from scratch with the HTML Canvas.

typescript
// main.ts
const canvas = document.createElement('canvas') as HTMLCanvasElement;
const ctx = canvas.getContext('2d') as CanvasRenderingContext2D;
canvas.height = window.innerHeight;
canvas.width = window.innerWidth;
document.body.appendChild(canvas);
let lastTime = performance.now();
const mainloop: FrameRequestCallback = (now) => {
const delta = (now - lastTime)/1000;
lastTime = now;
ctx.fillStyle = 'blue';
ctx.fillRect(0, 0, canvas.width, canvas.height);
ctx.font = '50px sans-serif';
ctx.fillStyle = 'lime';
ctx.fillText((1/delta).toFixed(1), 20, 100);
ctx.fillStyle = 'red';
ctx.fillRect(canvas.width/2, canvas.height/2, 40, 40);
requestAnimationFrame(mainloop);
}
mainloop(performance.now());
typescript
// main.ts
const canvas = document.createElement('canvas') as HTMLCanvasElement;
const ctx = canvas.getContext('2d') as CanvasRenderingContext2D;
canvas.height = window.innerHeight;
canvas.width = window.innerWidth;
document.body.appendChild(canvas);
let lastTime = performance.now();
const mainloop: FrameRequestCallback = (now) => {
const delta = (now - lastTime)/1000;
lastTime = now;
ctx.fillStyle = 'blue';
ctx.fillRect(0, 0, canvas.width, canvas.height);
ctx.font = '50px sans-serif';
ctx.fillStyle = 'lime';
ctx.fillText((1/delta).toFixed(1), 20, 100);
ctx.fillStyle = 'red';
ctx.fillRect(canvas.width/2, canvas.height/2, 40, 40);
requestAnimationFrame(mainloop);
}
mainloop(performance.now());

Vanilla js game running in Android emulator

Using Excalibur🗡

Using the Excalibur engine with capacitor and parcel will be a breeze! Really any web based game engine could be substituted here if you want. Here is the source on github!

> npm install excalibur --save-exact
> npm install excalibur --save-exact

Update the main.ts with some Excalibur

typescript
import { Actor, DisplayMode, Engine, Input, Loader, ImageSource } from "excalibur";
const game = new Engine({
displayMode: DisplayMode.FillScreen,
pointerScope: Input.PointerScope.Canvas
});
const sword = new ImageSource('assets/sword.png');
const loader = new Loader([sword]);
game.start(loader).then(() => {
game.input.pointers.primary.on('move', event => {
const delta = event.worldPos.sub(actor.pos);
actor.vel = delta;
// Original asset is at a 45 degree angle need to adjust
actor.rotation = delta.toAngle() + Math.PI/4;
});
const actor = new Actor({
x: game.halfDrawWidth,
y: game.halfDrawHeight,
width: 40,
height: 40
});
actor.graphics.use(sword.toSprite());
game.add(actor);
});
typescript
import { Actor, DisplayMode, Engine, Input, Loader, ImageSource } from "excalibur";
const game = new Engine({
displayMode: DisplayMode.FillScreen,
pointerScope: Input.PointerScope.Canvas
});
const sword = new ImageSource('assets/sword.png');
const loader = new Loader([sword]);
game.start(loader).then(() => {
game.input.pointers.primary.on('move', event => {
const delta = event.worldPos.sub(actor.pos);
actor.vel = delta;
// Original asset is at a 45 degree angle need to adjust
actor.rotation = delta.toAngle() + Math.PI/4;
});
const actor = new Actor({
x: game.halfDrawWidth,
y: game.halfDrawHeight,
width: 40,
height: 40
});
actor.graphics.use(sword.toSprite());
game.add(actor);
});

Note, depending on your emulator settings you may need to tweak it's graphics settings and restart Android Studio for it to build and run (This works out of the box fine on real hardware tested in BrowserStack, for some reason the emulator graphics can be confused)

Update graphics support

Tada! 🎉

Animated gif of excalibur sword running with Capacitor in Android

Hope this helps you web game devs out there!

-Erik

· 6 min read
Erik Onarheim

After the winter break, the team has released [email protected] with a lot of improvements to the core engine and plugins! Check the roadmap for our current plans.

Check out the new version on npm!

> npm install [email protected]
> npm install [email protected]

"Winter holiday, when developers work on their side projects" - Anonymous Coworker

Dev tools

> npm install @excaliburjs/dev-tools
> npm install @excaliburjs/dev-tools

We've built a new tool to help debug Excalibur games! This tool lets you see information about the Excalibur engine, scenes, actors, clocks, and more!

Debugging why things aren't working has historically been pretty difficult. This plugin will greatly assist in the game development cycle. Nearly everything is available and configurable.

It's pretty low effort to install into your game:

typescript
import { DevTool } from '@excaliburjs/dev-tools';
const game = new ex.Engine({...});
const devtool = new DevTool(game);
typescript
import { DevTool } from '@excaliburjs/dev-tools';
const game = new ex.Engine({...});
const devtool = new DevTool(game);

Excalibur dev tools, a sidebar of sliders and graphs on the right, with bounding boxes around all Actors in the game

Tiled updates

> npm install @excaliburjs/plugin-tiled
> npm install @excaliburjs/plugin-tiled

The Tiled plugin now implicitly adds a z-index to each layer (which can be overridden) which means things will look as you expect in Excalibur as they do in the Tiled editor.

Set the starting layer z (defaults to -1) and get gaming!

typescript
const map = new TiledMapResource('path/to/map.tmx', { firstLayerZIndex: -2 });
typescript
const map = new TiledMapResource('path/to/map.tmx', { firstLayerZIndex: -2 });

A side-by-side: on the left, a game with a blue square traveling along city roads. the right side is the level in a tiled map editor

Renderer performance improvements

The performance gains were achieved through some core renderer refactors and identifying places where expensive calculations could be cached!

This is huge, we stay above 30fps in the 4000 actor benchmark, and we have dramatic improvement in average fps in both cases!

Excalibur v0.25.0 vs v0.25.2 benchmarks showing that v0.25.2 has much more consistent average FPS

This benchmark was performed in Chrome on a Surface Book 2 with the power plugged in.

  • Processor: Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz, 2112 Mhz, 4 Core(s), 8 Logical Processor(s)
  • Physical Memory: (RAM) 16.0 GB
  • Graphics: NVIDIA GeForce GTX 1060

A number of improvements were made to the Excalibur graphics systems to get to this performance. The big factors to this improvement were:

  1. Avoiding recalculation of graphics transforms and other expensive operations when they can be cached
  2. Refactoring the renderer to be simpler and to use index buffers to share geometry vertices.
  3. Rendering batches at the actual maximum for the batch renderer
  4. Avoid recreating Matrix types, they are somewhat expensive to create then garbage collect

Post processing

The postprocessor improvements allow custom WebGL shaders, which can produce some cool effects! (Minimally modified from this ShaderToy)

Excalibur example running with a CRT television postprocessor

To produce the above effect, Excalibur has a new built in ScreenShader type for doing quick shaders meant for the whole screen.

typescript
class CrtPostProcessor implements ex.PostProcessor {
private _shader: ex.ScreenShader;
initialize(gl: WebGLRenderingContext): void {
const crtEffectSource = document.getElementById("modified-crt-shader-source").innerText;
this._shader = new ex.ScreenShader(crtEffectSource);
}
getLayout(): ex.VertexLayout {
return this._shader.getLayout();
}
getShader(): ex.Shader {
return this._shader.getShader();
}
}
game.graphicsContext.addPostProcessor(new CrtPostProcessor());
typescript
class CrtPostProcessor implements ex.PostProcessor {
private _shader: ex.ScreenShader;
initialize(gl: WebGLRenderingContext): void {
const crtEffectSource = document.getElementById("modified-crt-shader-source").innerText;
this._shader = new ex.ScreenShader(crtEffectSource);
}
getLayout(): ex.VertexLayout {
return this._shader.getLayout();
}
getShader(): ex.Shader {
return this._shader.getShader();
}
}
game.graphicsContext.addPostProcessor(new CrtPostProcessor());

Renderer improvements

Renderer design

When v0.25.0 was released, it was a "monolithic" renderer design, meaning everything Excalibur could possibly draw was built into a single renderer and shader program. This became onerous fairly quickly. And as the old adage goes: "you don't know how to build something until you've built it twice".

With v0.25.2, each type of drawing is split internally into separate renderer plugins. While this does come with some overhead when switching shader programs, it's worth it for the the simplicity, maintainability, and extensibility benefits.

Image filtering

Excalibur now allows you the ability to control the WebGL image filtering mode both implicitly and explicitly. Really this means Excalibur will try to pick a smart default, but allows overrides

Explicitly when loading ex.ImageSource:

typescript
const myImage = new ex.ImageSource('path/to/image', false, ex.ImageFiltering.Pixel);
typescript
const myImage = new ex.ImageSource('path/to/image', false, ex.ImageFiltering.Pixel);
  • ex.ImageFiltering.Blended - Blended is useful when you have high resolution artwork and would like it blended and smoothed

    Example of blended mode, where the edges of pixels are smoother

  • ex.ImageFiltering.Pixel - Pixel is useful for pixel art when you do not want smoothing aka antialiasing applied to your graphics.

    Example of pixel mode, where the pixels remain jagged

Implicitly if the ex.EngineOption antialiasing property is set:

  • antialiasing: true, then the blend mode defaults to ex.ImageFiltering.Blended
  • antialiasing: false, then the blend mode defaults to ex.ImageFiltering.Pixel

Custom renderer plugins

Excalibur knows how to draw many types of graphics to the screen by default comes with those pre-installed into the ExcaliburGraphicsContext. However, you may have a unique requirement to provide custom WebGL commands into Excalibur, this can be done with a custom renderer plugin.

A custom renderer can be registered with Excalibur and draw in any draw routine! Read more in the docs about custom rendere plugins

typescript
const game = new ex.Engine({...});
export class MyCustomRenderer extends ex.RendererPlugin {
public readonly type = 'myrenderer';
...
}
game.start().then(() => {
// register
game.graphicsContext.register(new MyCustomRenderer());
});
// call from a graphics callback or event
const actor = new ex.Actor({...});
actor.graphics.onPostDraw = (graphicsContext) => {
graphicsContext.draw<MyCustomRenderer>('myrenderer', ...);
}
typescript
const game = new ex.Engine({...});
export class MyCustomRenderer extends ex.RendererPlugin {
public readonly type = 'myrenderer';
...
}
game.start().then(() => {
// register
game.graphicsContext.register(new MyCustomRenderer());
});
// call from a graphics callback or event
const actor = new ex.Actor({...});
actor.graphics.onPostDraw = (graphicsContext) => {
graphicsContext.draw<MyCustomRenderer>('myrenderer', ...);
}

Community

We've had a lot of community engagement this iteration, especially through the issues and github discussions. Lots of good issues, and lots of cool things in the show and tell

Big thanks to everyone who helped with this release:

  • @ivasilov
  • @luttje
  • @tsanyqudsi
  • @lampewebdev
  • @joshuadoan
  • @berkayyildiz
  • @simon-jaeger
  • @YJDoc2
  • @JumpLink

The future

We are progressing at full speed toward the v1 vision, there is still a lot to do but the end is in sight. Here are a few things that I'm personally really excited for:

  • Event system redo
  • Particle system refactor
  • Final deprecation of old drawing api
  • New TileMap enhancements for hexagonal and isometric maps

This was a point release, but despite that a lot of exciting things were added! Looking forward to v0.26.0!

-Erik

· 11 min read
Erik Onarheim

After a year of work, a lot of great additions and improvements have made it into Excalibur, and we are making good progress towards our v1.0 release! Check the development roadmap for our current plans. It's hard to believe how different things are now since the first commit of Excalibur (back when it was called GameTS)!

Excalibur started as a tech demo in a presentation to show how powerful TypeScript can be. The engine has come so far since then, it's really amazing!

We are really excited to share this release with you! This release contains over 30 bug fixes and 50 new features! It's been a labor of love over the last year by many people, and we have some big features to share.

Check out the official release!

npm install [email protected]

Performance

There is a combination of features (mentioned below) that resulted in big performance gains. Across the board, there's been a dramatic increase in what Excalibur can do in v0.25.0 vs v0.24.5.

In the gif below, we demonstrate the graphics performance specifically.

4000 small robot Actors (no collisions) exploding outwards from the center of the screen

There is much better performance across the board with a higher baseline FPS in v0.25.0 for the same number of actors. You'll notice that FPS improves over time as more actors are offscreen in v0.25.0 compared to v0.24.5.

graphs showing an average improvement of 8 FPS for 1000 Actors, 24.75 FPS for 2000 Actors, and 21.27 FPS for 3000 Actors

This benchmark was performed in the Chrome browser on a Surface Book 2 with the power plugged in.

  • Processor: Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz, 2112 Mhz, 4 Core(s), 8 Logical Processor(s)
  • Physical Memory: (RAM) 16.0 GB
  • Graphics: NVIDIA GeForce GTX 1060

New plugin versioning strategy

We are adopting a similar versioning strategy to Angular, during pre-1.0. All plugins compatible with the core library will share the same prefix through the minor version. For example, if core Excalibur is [email protected], then the plugins that support that version are formatted like @excaliburjs/[email protected].

DisplayMode updates

Excalibur DisplayModes have been refactored and renamed to clarify their utility.

  • FillContainer - Fill the game viewport to take up as much of the immediate parent as possible
  • FillScreen - Fill the game viewport to take up as much of the screen as possible
  • FitContainer - Fit the game maintaining aspect ratio into the immediate parent
  • FitScreen - Fit the game maintaining aspect ration into the screen
  • Fixed - Specify a static size for the game width/height

Refactor to Entity Component System (ECS) based architecture

The core plumbing of Excalibur has been refactored to use an ECS style architecture. However, developers using Excalibur do not need to know or care about the this underlying change to ECS if they don't want to.

What does ECS mean for Excalibur? At a high level, ECS architecture breaks down into three things:

  • Components contain data needed for various systems.
  • Systems implement the "behavior" by looping over entities that match a list of components.
    • For example, the graphics system processes all entities with a TransformComponent and a GraphicsComponent
  • Entities are the "holders" of components

Actor, Scene, and Engine remain as the familiar interface to build games; they're only implemented differently under-the-hood. The reason for the change was to break down ever-growing and complex logic that had accumulated in the Actor and Scene implementations into Components and Systems for maintainability. This change increases the flexibility of Excalibur, and allows you to add new novel behavior directly into the core loop with custom components ones if you desire.

Excalibur does not have the purest implementation of an ECS by design; our built-in components are more than just data. The built-in components do provide behavior, convenience features, and helper functions to maintain our core mission of keeping Excalibur easy to use. The Excalibur core team goal with ECS is flexibility and maintainability, not performance. If you wish, you can read more about our goals for ECS.

Here's A quick example of using the new ECS features:

typescript
class SearchComponent extends ex.Component<'search'> {
public readonly type = 'search'
constructor(public target: ex.Vector) {
super();
}
}
class SearchSystem extends ex.System<ex.TransformComponent | SearchComponent> {
// Types need to be listed as a const literal
public readonly types = ['ex.transform', 'search'] as const;
// Lower numbers mean higher priority
// 99 is low priority
public priority = 99;
// Run this system in the "update" phase
public systemType = ex.SystemType.Update
private _searchSpeed = 10 // pixels/sec
public update(entities: ex.Entity[], delta: number) {
for (let entity of entities) {
const target = entity.get(SearchComponent)!.target;
// ex.TransformComponent is a built in type
const transform = entity.get(ex.TransformComponent) as ex.TransformComponent;
const direction = target.sub(transform.pos);
const motion = direction.normalize().scale(this._searchSpeed);
// Moves these entities towards the target at 10 pixels per second
transform.pos = transform.pos.add(motion.scale(delta / 1000))
}
}
}
// Actors come with batteries included built in features
const actor = new ex.Actor({
pos: ex.vec(100, 100),
width: 30,
height: 30,
color: ex.Color.Red
});
actor.addComponent(new SearchComponent(ex.vec(400, 400)));
// Create a scene with your new system
const scene = new ex.Scene();
scene.world.add(new SearchSystem());
scene.add(actor);
typescript
class SearchComponent extends ex.Component<'search'> {
public readonly type = 'search'
constructor(public target: ex.Vector) {
super();
}
}
class SearchSystem extends ex.System<ex.TransformComponent | SearchComponent> {
// Types need to be listed as a const literal
public readonly types = ['ex.transform', 'search'] as const;
// Lower numbers mean higher priority
// 99 is low priority
public priority = 99;
// Run this system in the "update" phase
public systemType = ex.SystemType.Update
private _searchSpeed = 10 // pixels/sec
public update(entities: ex.Entity[], delta: number) {
for (let entity of entities) {
const target = entity.get(SearchComponent)!.target;
// ex.TransformComponent is a built in type
const transform = entity.get(ex.TransformComponent) as ex.TransformComponent;
const direction = target.sub(transform.pos);
const motion = direction.normalize().scale(this._searchSpeed);
// Moves these entities towards the target at 10 pixels per second
transform.pos = transform.pos.add(motion.scale(delta / 1000))
}
}
}
// Actors come with batteries included built in features
const actor = new ex.Actor({
pos: ex.vec(100, 100),
width: 30,
height: 30,
color: ex.Color.Red
});
actor.addComponent(new SearchComponent(ex.vec(400, 400)));
// Create a scene with your new system
const scene = new ex.Scene();
scene.world.add(new SearchSystem());
scene.add(actor);

Collision system improvements

The collision system has been significantly overhauled to improve the quality of the simulation and the stability of collisions. The core simulation loop "solver" has been redone to use an iterative impulse constraint solver, which provides a robust method of computing resolution that has improved performance and stability.

Collision intersection logic has now also been refactored to report multiple contact points at once. Multiple contacts improves the stability of stacks of colliders over single contact collisions (which can result in oscillations of boxes back and forth).

variously-sized rectangles being stacked one at a time on top of each other and not falling over (like they usually would without multiple contact point collisions)

Colliding bodies can now optionally go to sleep. This relieves some of the pressure on the collision solver and improves the stability of the simulation by not moving these objects if they don't need to move. Colliders can be started asleep before a player in a game might interact with them

a sleeping collisions demo, where a horizontal rectangle is dropped onto two parallel vertical rectangles; the wobbling ceases quickly, and the structure remains stable because the collisions went to sleep

New CompositeColliders now make it possible to combine Excalibur Collider primitives (PolygonCollider, CircleCollider, and EdgeCollider) to make any arbitrary collision geometry. These new composite colliders power the new TileMap cell collisions and also power the new ex.Shape.Capsule(width, height) collider.

a grid of red bricks with composite collider lines drawn around groups of multiple bricks at a time

The Capsule collider is a useful geometry tool for making games with ramps or slightly jagged floors you want a character to glide over without getting stuck. This collider also helps with any "ghost collisions" that you might run into under certain conditions in your game.

an image of two green circles connected on each outer side by two green lines. the structure is standing on a platform

CollisionGroups allow more granular control over what collides above and beyond collision type. Collsion groups allow you to create named groups of colliders like "player", "npc", or "enemy". With these groups, you can specify that players and enemies collide, player and npcs don't collide, and that npcs and enemies don't collide without needing to implement that logic in a collision event handler.

typescript
// Create a group for each distinct category of "collidable" in your game
const playerGroup = ex.CollisionGroupManager.create('player');
const npcGroup = ex.CollisionGroupManager.create('npcGroup');
const floorGroup = ex.CollisionGroupManager.create('floorGroup');
const enemyGroup = ex.CollisionGroupManager.create('enemyGroup');
// Define your rules
const playersCanCollideWith = ex.CollisionGroup.collidesWith([
playersGroup, // collide with other players
floorGroup, // collide with the floor
enemyGroup // collide with enemies
]);
const player = new ex.Actor({
collisionGroup: playersCanCollideWith
});
typescript
// Create a group for each distinct category of "collidable" in your game
const playerGroup = ex.CollisionGroupManager.create('player');
const npcGroup = ex.CollisionGroupManager.create('npcGroup');
const floorGroup = ex.CollisionGroupManager.create('floorGroup');
const enemyGroup = ex.CollisionGroupManager.create('enemyGroup');
// Define your rules
const playersCanCollideWith = ex.CollisionGroup.collidesWith([
playersGroup, // collide with other players
floorGroup, // collide with the floor
enemyGroup // collide with enemies
]);
const player = new ex.Actor({
collisionGroup: playersCanCollideWith
});

New graphics system

The new Excalibur graphics system has been rebuilt from the ground up with speed in mind. It is now built on a WebGL foundation with a built-in batch renderer. This means that Excalibur will batch up draw commands and submit the minimum amount of draw calls to the machine when the screen is updated. This dramatically improves the draw performance and also the number of things wec can display on screen (as noted in the benchmarks earlier).

For drawing hooks the ExcaliburGraphicsContext is replacing the browser CanvasRenderingContext2D. If you still need to do some custom drawing using the CanvasRenderingContext2D the new Canvas graphic can help you out.

typescript
const canvas = new ex.Canvas({
cache: true, // If true draw once until flagged dirty again, otherwise draw every time
draw: (ctx: CanvasRenderingContext2D) => {
ctx.fillStyle = 'red';
ctx.fillRect(0, 0, 200, 200);
}
});
actor.graphics.use(canvas);
typescript
const canvas = new ex.Canvas({
cache: true, // If true draw once until flagged dirty again, otherwise draw every time
draw: (ctx: CanvasRenderingContext2D) => {
ctx.fillStyle = 'red';
ctx.fillRect(0, 0, 200, 200);
}
});
actor.graphics.use(canvas);

TileMap and Tiled updates

Tiled is easily one of the best tools out there for building and designing levels for your game. It has certainly been a valuable tool in our toolbox. We have doubled down on our efforts to provide a first class Tiled integration with Excalibur via the excaliburjs/plugin-tiled. This work also involved a few improvements to the TileMap to improve it's graphics API and collision performance.

Check out the Tiled Excalibur Plugin!

  • Full support for the Tiled object model
  • Full support for all Tiled file types
  • Excalibur built ins
  • Not yet supported
    • Tiled Group Layers
    • Custom Tile colliders
    • Isometric/Hexagonal maps
    • Parallax

a blue square moving around a pixelated cityscape build in the Tiled map editor

Documentation

A lot of time was spent reviewing and improving our documentation. Part of this work was ensuring that the snippets don't go stale over time by building them in GitHub Actions.

Please check out the new and shiny doc site with new code examples at excaliburjs.com!

Testing

The Excalibur core repo now has WallabyJS enabled to improve the VS Code test development and debugging experience. Wallaby is a paid tool; because of that Excalibur will always also support the Karma based testing framework for official tests.

A useful update to excalibur-jasmine allows async matchers, which greatly simplifies checking image diffs in Jasmine unit tests.

typescript
it('should match images', async () => {
let engine = new ex.Engine({width: 100, height: 100});
await expectAsync(engine.canvas).toEqualImage('images/expectedcanvas.png', .99);
});
typescript
it('should match images', async () => {
let engine = new ex.Engine({width: 100, height: 100});
await expectAsync(engine.canvas).toEqualImage('images/expectedcanvas.png', .99);
});

A brand new integration test utility has been created called @excaliburjs/testing, which provides a quick way to drive Excalibur games with Puppeteer and do image-based snapshot testing.

typescript
// excalibur testing
test('An integration test', async (page) => {
// Check for the excalibur loaded page
await expectLoaded();
// Compare game to expected an expected image
await expectPage('Can check a page', './images/actual-page.png').toBe('./images/expected-page.png');
// Use puppeteer page object to interact
await page.evaluate(() => {
var actor = ((window as any).actor);
actor.pos.x = 400;
actor.pos.y = 400;
});
// Compare game to a new expected image
await expectPage('Can move an actor and check', './images/actual-page-2.png').toBe('./images/expected-page-2.png');
});
typescript
// excalibur testing
test('An integration test', async (page) => {
// Check for the excalibur loaded page
await expectLoaded();
// Compare game to expected an expected image
await expectPage('Can check a page', './images/actual-page.png').toBe('./images/expected-page.png');
// Use puppeteer page object to interact
await page.evaluate(() => {
var actor = ((window as any).actor);
actor.pos.x = 400;
actor.pos.y = 400;
});
// Compare game to a new expected image
await expectPage('Can move an actor and check', './images/actual-page-2.png').toBe('./images/expected-page-2.png');
});

running an interactive image integration test and showing the ability to update the expected image snapshot

Templates

There are a lot of different ways to build web apps; we've created repo templates for some of the popular ones:

Samples

Community

We've had tons of community contributions since the last release. Heartfelt thanks to everyone in the discussions, issues and pull requests!

Contributors:

  • @jedeen
  • @kamranayub
  • @alanag13
  • @DaVince
  • @DrSensor
  • @djcsdy
  • @catrielmuller
  • @AndrewCraswell
  • @miqh
  • @rledford
  • @SirPedr
  • @helloausrine
  • @dpayne5
  • @herobank110
  • @didii
  • @Charkui
  • @muirch
  • @rumansaleem
  • @mogoh
  • @kala2
  • @MrBartusek
  • @josh-greenlaw
  • @LokiMidgard
  • @romaintailhurat
  • @EduardoHidalgo
  • @jaredegan

Breaking changes

There are some breaking changes in v0.25.0 from v0.24.5; see the changelog and release notes for more specifics, but they generally fall into the categories below. See the migration guide for guidance.

  • New APIs replacements
    • Graphics API
    • Actor drawing functions moved to graphics component
  • API renames for clarity
  • Bug fixed necessitated change
  • Extracted behavior to a plugin
    • Perlin noise is now offered as a plugin and is no longer included in the core library @excaliburjs/plugin-perlin
  • Big plugin changes
    • The Tiled plugin is now published under @excaliburjs/plugin-tiled and will start with version v0.25.0

Looking towards "version 1"

  • Pointer events plumbing refactor; the current system is hard to follow and debug/enhance
  • Particle system refactor
  • Graphics enhancements to support advanced postprocessing/shaders
  • ExcaliburGraphicsContext enhancements to grant more flexibility
  • Event system redo
  • Better Scene management and granular asset loading
  • Expand and enhance TileMap
  • Progressive WebAssembly enhancements in the physics simulation
  • Potential new plugins on the horizon
  • AI patterns and plugins like A* search
  • API finalization

I want to thank everyone who helped make this version of Excalibur possible. A lot of effort went into it and I'm really proud of what we achieved.

- Erik

· 2 min read
Erik Onarheim

This is a big release for Excalibur on our journey to 1.0.0. If you’d like to follow along, we now have a tentative roadmap available! The goal for this release was to simplify our collision infrastructure and utilities.

Thanks to our community contributors for all of their help! (see the full release notes)

Notable highlights

  • Collision groups have been re-implemented to be more in line with industry practice. They allow you to determine which colliders collide with others.
  • Collision behavior and properties are now contained within the new type ex.Collider
    • Collision types are now sourced from ex.Collider
    • Collision groups now live on ex.Collider
    • Collision shapes dictate collision geometry live on ex.Collider
    • Collision pixel offset allows shifting of colliders by a pixel amount
    • Properties like mass, torque, friction, inertia, bounciness are now all part of ex.Collider instead of ex.Body
  • Decoupling Actor from the collision system
    • ex.CollisionPair now works on a pair of Colliders instead of a pair of Actors to represent a potential collision
    • ex.CollisionContact now works on a pair of Colliders instead of a pair of Actors to represent an actual collision
  • New helpful methods for colliders
    • Find the closest line between 2 colliders or shapes
    • ex.Actor.within now works based on the surface of the geometry, not the center of the object

animated gif demonstrating finding the closest lines between several shapes

  • Actions moveBy, rotateBy, and scaleBy have been changed to move an actor relative to the current position
    • This change makes implementing patrolling behavior moving 400 pixels left and right forever as easy as: actor.actions.moveBy(-400, 0, 50).moveBy(400, 0, 50).repeatForever();

repeated patrolling behavior demo for the above Actions code example, showing the Actor moving back and forth along a platform

  • Many name refactorings and deprecations to improve usability (see the full release notes)

New sample game

We have a new sample game to illustrate best practices when developing with Excalibur.

sample platformer animation, showing the player, a patrolling NPC, and patrolling enemies

Look forward to many more updates in the months ahead!

· 3 min read
Erik Onarheim
Jae Edeen
Kamran Ayub
Alan Grgic
Sean Igo

Play the Ludum Dare version of ‘Office Daydream’

screenshot showing the game: the lower half is various office mini games while the upper &quot;daydrem&quot; is a motorcycle platformer

What went well

Shorter workdays

We had the fortunate opportunity to get out of the city for a bit and take a vacation preceding the game jam. We wanted to take advantage of our time away from home, so we instituted regular work days (8 hours), rather than the 10-12 hour days we usually fall into the trap of doing . And surprise, it went great! Everyone was more relaxed, and we delivered a quality game in less time! It was easier for the team to focus on getting stuff done. We also managed to finish early, just ahead of the submission hour.

Art pipeline

the arm that indicates your clicking actions, shown in various skin tones

We did most of our work in Aseprite, with a little bit of Photoshop thrown in. A lot of the art in this game is simple shapes, and we recycled many of the backgrounds across the mini-games.

Brainstorming

the copy machine minigame, where the player matches the lit-up buttons to get the copy machine to work

The theme was announced at 8:00 pm in our time zone. We spent the evening hours brainstorming, picking a few to develop a little more, and then deciding on our favorite choice.

Scope pruning

office posters on a wall that say &quot;you should be working&quot; and &quot;synergy&quot;

We were ruthless at eliminating extras that tried to sneak in to the game. We had less time than we usually do, so we had to work efficiently. The resulting game is about half of the scope it had the potential to become, and we kept it under control.

What could have gone better

the stapling minigame, where you have to click on the various pages of a report in the correct order to staple and complete them

We had to spend some extra time ahead of the jam updating our game template and tools. We could have done this at any earlier point. It wasn’t too bad, but it did cut a bit into our relaxation time for the days preceding the jam.

We also struggled again with managing state in Excalibur. We’re working on incorporating this into the Excalibur engine to improve the development process.

timelapse of us working on our computers in a well-lit room with wood trim

Overall, this Ludum Dare went great. We look forward to playing all of the cool games that we’ve seen so far for LD41!

· 2 min read
Erik Onarheim

Five years in review

The first commit to Excalibur.js was published on January 5th, 2013. Since then, we’ve been working to build a game engine that’s easy to develop with and fun to use. Along the way, we put together a release pipeline, constructed a test suite, wrote a lot of documentation, and created a number of extensions and samples. Here’s a quick rundown of some of the numbers:

  • 103068 npm downloads
  • 1275 commits to core repository
  • 447 closed issues
  • 35 contributors
  • 17 releases

Future plans

There are some pretty big improvements coming up, and we’re looking forward to sharing those changes with you over the next year. We’re also working on a larger collection of samples and games to help new developers and showcase Excalibur’s capabilities. For a more detailed look, check out the roadmap.

Thank you

We want to extend our sincere thanks to everyone who has written code, opened an issue, posted in our forum, or made a game and let us know about it. This is a very fulfilling project, and seeing others contribute to and use Excalibur means a lot to us. Thanks for your support!

-The Excalibur.js team

· 6 min read
Erik Onarheim
Jae Edeen
Kamran Ayub
Alan Grgic
Sean Igo

screenshot of the game, showing a top-down 2d grocery store filled with customers

Play the LD version of ‘I Just Wanted Groceries’

This is our fifth time back for Ludum Dare. We had a full house again, and the experience was a lot of fun!

What went well

Preparation

screenshot of some of our Trello cards to keep track of tasks and info

In the few days before Ludum Dare, we made sure everything was ready. We set up version control, automatic deploys, and scripted tasks to help us build and develop the game. We ran into some problems during this setup, which is exactly why we do this early. These steps have become necessary for us before every game jam.

Brainstorming

The theme was announced at 8:00 P.M. our time, and we spent the entire rest of the evening brainstorming. We made sure not to settle on anything too quickly, and came up with as many ideas as possible. From those fifty or so possibilities we picked a handful that seemed interesting enough to build. An hour or two later, we’d thrown all but one of those out, and settled on “avoid talking to people someplace – grocery store”.

Prototyping

a whiteboard with a grocery store layout and various pathing arrows drawn on it

We built the first version of the game level on a dry erase board. This let us iterate on how a player would traverse it, as well as devise methods for pathfinding and item generation. Every time we’ve taken the time to do this step, we saved ourselves a load of heartache. Clearing up design issues is much simpler before the code has been written.

Art

the various food and supply items: canned tomatoes, chips, cereal, frozen pizza, steak, toilet paper, bread, bananas, and carrots

an example spritesheet for one of the character models

the background tileset for the grocery store

We had three team members working on art at different points in the weekend. We built the level, the characters, and the food items from scratch. It can be difficult to maintain a visual consistency when multiple people are drawing things; we mitigated a portion of this by standardizing on the “x11” palette built in to Aseprite, so at least all of our colors matched.

Sound

an audio waveform

We didn’t make our own music this time, but we did design most of the sound effects using littleBits and a guitar. We feel like we did a good job of unifying the soundscape and setting a cohesive mood with the audio.

Simple construction

a short timelapse of various stages of the game&#39;s development, where it&#39;s mostly just squares and lines moving around

One of the most useful things we do is enforce the restriction of building a “playable” game as soon as possible. It doesn’t need any fancy extras, it just needs to let the player interact with and experience it. We managed to get to that point by Sunday morning, which left us two full days to add all those cool extras. It also allowed us to play the game a lot and polish the rough edges.

Recycled code

We started our game with a Yeoman template we built to structure the basics for us. We also reused a number of code snippets that we had left over from other games, including animation code and player input logic. Every little bit of time we saved helped us build this game better than we could have otherwise.

Excalibur

We hardly encountered any actual bugs in Excalibur this time around. We did put together a long list of potential improvements, though, and look forward to incorporating those into the engine.

Using Tiled

a screenshot of our level loaded in the Tiled Editor

If you’re making a tile or grid-based game, Tiled is a great editor to build your levels in. We used it to define zones for our different grocery items to spawn in, as well as waypoints to define the shoppers’ movements.

Custom analytics

a heatmap of aggregated player movement throughout the grocery store

We’ve tracked analytics in our games before, but we wanted a little more granularity this time. We configured custom analytics with Azure functions, and were able to track whatever game properties we wanted.

Eating and sleeping

a deep dish pizza ready to be cooked in the oven

We ended each day after about 10:00 P.M. Ending early meant we could start early, and sleep is the best medicine for tired minds. We also cooked several meals instead of just getting fast food all the time. These two things were marked improvements from previous game jams.

Results

results: 77th overall, 88th in &quot;Fun&quot; category, 298th in &quot;Innovation&quot; category, 230th in &quot;Theme&quot; category, 328th in &quot;Graphics&quot; category, 38th in &quot;Humor&quot; category, 202nd in &quot;Mood&quot; category

Our results were pretty good! We scored highest in humor and fun, which was what we were aiming for. We’re glad people liked the game.

What didn’t go so well

Tiled plugin

We experienced a small complication with our custom Tiled plugin, due to a versioning issue with Excalibur. It wasn’t too difficult to fix, but it did slow us down a bit.

Bugs in Excalibur

While the bug count this jam was low, it still wasn’t zero. One day, perhaps, but not this time. If anything, it’s better we run into these things before other users do, in order to prevent frustration with the development experience.

Difficulty in pathfinding

It was a bit difficult to get exactly what we were looking for in so short a time span. We managed to achieve most of the desired results, but there are still a few rough edges. We’ll be looking at adding pathfinding support directly into Excalibur in the future.

Conclusion

Every time we do a restrospective, the “what went well” section takes more and more space away from the “what didn’t go so well” section. It’s really encouraging to see this become a more interesting and rewarding process as we make new games. Ludum Dare was a lot of fun, and we hope to participate again someday.

Thanks for reading!

· 2 min read
Erik Onarheim

Excalibur version 0.8.0 is now available! We have several new features in this release.

Fast body collision checking

Actors can now move much faster without fear of unexpectedly passing through other collision bodies.

visualization fo fast body collision checking: a ball heads towards a wall, but a line ahead of the ball detects that the ball may collide soon with the wall, and prevents it from accidentally passing through it

demo of fast body collision checking: projectiles are thrown around inside a box at high speeds and do not escape the box

New vector and line functionality

We’ve added a few helpful things to Line and Vector, including determining points, calculating distance, and a vector magnitude alias.

Debug statistics

We now have a utility from which Excalibur will provide useful statistics to help you debug your game. For now the stats are focused on Actors and specific frames; look for more helpful stats in future releases!

PhantomJS testing structure

Behind the scenes, we have new testing tools that will allow us to visually test complicated interactions on the canvas.

There were quite a few commits from the Excalibur community in this release. Thanks to FerociousQuasar and hogart for your contributions, and check out the full release notes for all of the details for this release.