Giving your branch a spin now to see if I can find the issue you mentioned on discord.

After getting everything to build I've gotten to the same issue you ran into. I even tried defining the Symbol.for(Jupyter.display) earlier in the code and referencing that directly.

Well I got it to work and it must be a TypeScript or Deno quirk. If you bind it in the constructor then the Deno kernel picks it up:

  constructor(width: number, height: number, gpu = false) {
    this[$display] = this[$display].bind(this);

👀 Something weird with Deno/Rust picking up the available function on this class @bartlomieju. We gotta find a minimal version that has this odd behavior.

Could you point me which one is not working here? Is it canvas.ts or svgcanvas.ts?

Could you point me which one is not working here? Is it canvas.ts or svgcanvas.ts?

@bartlomieju it's canvas.ts

Thanks, I'll try to take a look ASAP.

I just tried this and I got:

I think this is related to denoland/deno#20528

So the underlying issue is that this error is raised:

TypeError: Cannot read properties of undefined (reading 'encode')\n    at [Jupyter.display] (file:///Users/ib/dev/skia_canvas/src/canvas.ts:215:38)\n    at <anonymous>:4:33

However I don't yet understand why that's the case. The transpiled output looks correct and AFAIK this should be the class intance in this case. I will consult with the team and get back to you.

The bug causing it to not render properly has been fixed: denoland/deno#20789

I had to jump to denoland/deno@cbddf57 to make it work though because deno main can't load this module right now (#61).

Awesome fix @rgbkrk - I'll cut a release once that fix makes it to stable!

The fix for ffi is on main.

1.37.2 is out with the fix, feel free to ship 🚢

I've released 0.5.5: