An Infinite Canvas Tutorial

What is an Infinite Canvas? The term "infinite" in infinitecanvas is described as follows:

• High scalability. Users can freely organize content structures in a non-linear fashion.
• Zooming. Emulates the "zoom in" to get an overview and "zoom out" to observe details as in the real world.
• Direct manipulation. Provides intuitive editing capabilities for basic shapes, including moving, grouping, and modifying styles.

The infinitecanvas showcases numerous examples ranging from design tools to creative boards, including some well-known products such as: Figma, Modyfi, rnote, tldraw, excalidraw and so on.

As a front-end developer, I am very interested in the rendering technologies involved. Although tldraw, excalidraw, and others generally use more user-friendly technologies like Canvas2D/SVG, there are also many editors and design tools in the JS and Rust ecosystems that use more low-level rendering technologies for 2D graphics with GPU acceleration to achieve better performance and experience:

• Figma uses a tile-based rendering engine written in C++, compiled into WASM and then calls WebGL for rendering.
• Modyfi uses wgpu from the Rust ecosystem, also compiled into WASM and then calls WebGL2 for rendering.
• Zed uses GPUI to render rectangles, shadows, text, images, and other UI elements.
• Vello and xilem experimentally use Compute Shader for 2D rendering.

Therefore, in this tutorial, I hope to implement the following features:

• Use @antv/g-device-api as a hardware abstraction layer, supporting WebGL1/2 and WebGPU.
• Referencing mapbox and Figma, attempt to use tile-based rendering.
• Use SDF (Signed Distance Field) rendering for circles, ellipses, rectangles, etc.
• GPU-accelerated text and Bezier curve rendering.
• Use rough.js to support hand-drawn styles.
• Use CRDT (Conflict-free Replicated Data Type) to support collaborative Yjs.

I hope to rewrite the rendering part of the canvas with Rust in the future, but the current project completion is still relatively low:

• wgpu is a very reliable hardware abstraction layer, which can even implement the backend for piet.
• Shaders can basically be reused.
• Hand-drawn styles can use rough-rs.
• y-crdt is the Rust implementation of Yjs.

Let's get started!

Getting Started

The course project uses pnpm workspace, so you need to install pnpm first.

pnpm i

After entering the course directory, run [vite]:

cd packages/lesson_001
pnpm run dev

Lesson 1 - Initialize canvas

• A hardware abstraction layer based on WebGL1/2 and WebGPU.
• Canvas API design.
• Implementing a simple plugin system.
• Implementing a rendering plugin based on the hardware abstraction layer.

Lesson 2 - Draw a circle

• Adding shapes to the canvas.
• Drawing a circle using SDF.
• Anti Aliasing.
• Dirty flag design pattern.

Lesson 3 - Scene graph and transform

• Transformations. Make shapes support pan, zoom, rotate, and skew transformations.
• Scene graph.

Lesson 4 - Camera

• What is a Camera?
• Projection transformation.
• Camera transformation.
• Camera animation. Using Landmark transition between different camera states.

Lesson 5 - Grid

• Drawing straight lines using Line Geometry or screen-space techniques.
• Drawing dots grid.

Lesson 6 - Event System

• Implement an event system compatible with DOM Event API.
• How to pick a circle.
• Implement a drag-and-drop plugin based on our event system.
• Support for pinch zoom gestures.

Lesson 7 - Web UI

• Developing Web UI with Lit and Shoelace
• Implementing a canvas component
• Implementing a zoom toolbar component

Lesson 8 - Optimize performance

• What is a draw call
• Reducing draw calls with culling
• Reducing draw calls by combining batches
• Using spatial indexing to improve pickup efficiency