play with wasm_bindgen for IDL

[chenyan-dfinity]

This works locally on my machine. Will try to build with nix.

Observations:

• Exported Rust function cannot have type variable
• Export C-like enum only
• Box<[JsValue]>, Result<T, JsValue>

JS code

    const bytes = js.js_encode([42,false, "哈哈",  { 'ok': 42, 123: 'test' }, [[1,2,3],[4,5]]]);
    console.log(bytes);
    const args = js.js_decode(bytes);
    console.log(args);

[hansl]

@chenyan-dfinity What's the status here?

[chenyan-dfinity]

What's our plan for wasm_bindgen?

It's working locally on my machine, but no nix support. We can keep this branch as a playground. Or I can absorb the code into the IDL rust library.

[hansl]

Do we need a plan for wasm_bindgen? What are the benefits of having support?

[chenyan-dfinity]

What are the benefits of having support?

• No need to maintain a JS Userlib.
• No language gaps: everything is in Rust.
• Assuming we trust wasm_bindgen, no specific testing for JS side.
• Serving both frontend and backend in WebAssembly.

Do we need a plan for wasm_bindgen?

I don't know. Given that we already have a JS userlib, it may not be a priority for SDK. But it's nice to have. Maybe when we open source, the community can build something similar for us.

[hansl]

No need to maintain a JS Userlib.

We would still have to maintain a userlib, just the IDL part would go away. And since we'll most likely have to maintain some kind of library for many languages, I don't see this as a big negative.

No language gaps: everything is in Rust.

Until we need to support another language.

Serving both frontend and backend in WebAssembly.

Apps will still have a JavaScript part.

[chenyan-dfinity]

We would still have to maintain a userlib, just the IDL part would go away.

We can apply wasm_bindgen to most of the dfx codebase, so we can replace all the userlib JS code, e.g. http_agent, request, etc.

And since we'll most likely have to maintain some kind of library for many languages, I don't see this as a big negative.

I think the idea is that with (hopefully) most language supporting WebAssembly, we don't need to maintain libraries for many languages.

No language gaps: everything is in Rust.

Until we need to support another language.

We already have gaps between JS and Rust. Every new feature/bug fix needs to be done twice, in both languages.

Serving both frontend and backend in WebAssembly.

Apps will still have a JavaScript part.

Not necessarily :)

[hansl]

I'm going to close this as it's at least 4 months old.