Is there still a chance of kickstarting the project?

[baffalop]

I came here from https://zackoverflow.dev/writing/flappy-bird-in-type-level-typescript

This looks like a really worthwhile project. I love typescript's type-level power and recognise its compiler is crushingly slow compared to what's possible. I'm also really excited about the goal of expanding the language to a general purpose DSL/schema tool.

Is this officially dead, or is a push for funding still on the horizon? I'd love to chip in at a personal level. Or would you consider tyvm as a worthy successor?

[marcj]

Funding is definitely still on the horizon, but I have no idea how to kickstart it concretely. Do you have any? I think about it a lot (especially the theory behind improving it further like lazy evaluation, latent materialisation, and faster type comparison using probabilistic methods like bloom filter etc, on a academical level), but it's a project that needs a lot of time. If we would find a way to fund it, I'm definitely down to implement it and write about it (maybe even release a paper, since it seems it would be the first of its kind). I have already written Bun (@Jarred-Sumner) about sponsorship as I think it falls within his domain, but no response.

[baffalop]

Sorry I have no experience with OSS funding myself, but I'm glad to hear it's still something you're still pursuing. I don't know what your life situation is but if you're talking about academic innovation and publishing a paper, it seems almost like this could be a PhD topic?

I am curious what your thoughts are on https://github.com/zackradisic/tyvm/ since it appears to have the exact same goals?

[marcj]

it seems almost like this could be a PhD topic?

Absolutely, this is also something I'm interested in.

@zackradisic's TypeScript bytecode interpreter is very cool! I think it shares the same core ideas, e.g. using bytecode for TypeScript type computation and making TS available to other languages to replace inferior DSLs. Only time will tell if tyvm will solve more than that like actual type checking. It's experimental and more like a PoC just like TypeRunner.

That being said, type checking itself is much more complicated than just type computation. As the README of tyvm already says, TypeScript's language is pretty straightforward to compute, but actual type checking is the big task here: Doing inference of JS with all the branching/narrowing and comparing types (potentially very deeply nested complicated types) in a fast way with all sorts of optimisation like lazy evaluation/materialisation, better GC and memory layout.

[baffalop]

@marcj Right I see. Yeah I got my first window into how deep type inference goes with a recent talk at Typescript Congress, just talking through the inference logic for generics on functions. There's way more involved than you'd ever consider at first blush!

Cool that you're interested in the PhD. Good luck with that and the project overall!