If you’re wrestling with how to actually detect and interact with drawn shapes on a canvas, this deep-dive on pixel-to-object conversion is exactly what you need.
Canvas as raw pixels is basically working in assembly. The object model is just giving yourself a scene graph so hit detection isn’t a nightmare of manual bounding box math. Curious what they use for the object store though, is it just a flat array they iterate on every click, or something with spatial indexing? That matters a lot once you’ve got hundreds of shapes on screen.
