Evaluate if we should use oxc instead of swc

[AbhiPrasad]

https://oxc.rs/

might give us some more performance wins, I think the DX for the rust types is also a little better.

There is also https://www.npmjs.com/package/oxc-parser

[BridgeAR]

That would indeed be great to get that extra performance boost! Would you mind working on that?

[AbhiPrasad]

Sure I can take a look! I'll wait till #12 to minimize merge conflicts cc @timfish

One thing that is not ideal is that oxc doesn't support custom plugins yet, so that might make plans for a bundler plugin much more difficult. Let me try pinging the folks at voidzero to see the timeline around that work.

[timfish]

My guess would be that using oxc-parser won't be any faster than #12 because oxc-parser calls back into JavaScript for every AST node. Whereas for #12 we pass the source and the AST walking is entirely in wasm. That's purely a guess though so we should probably try!

[AbhiPrasad]

because oxc-parser calls back into JavaScript

do you have a link to the source code or docs that shows this behaviour?

[timfish]

Sorry, oxc-parser returns the full estree compatible AST to JavaScript. So you walk and modify it in JavaScript?

import oxc from 'oxc-parser';

const code = 'const url: String = /* 🤨 */ import.meta.url;';

// File extension is used to determine which dialect to parse source as.
const filename = 'test.tsx';

const result = oxc.parseSync(filename, code);
// or `await oxc.parseAsync(filename, code)`

oxc-walker calls into JavaScript for every node.

import { parseAndWalk, walk } from 'oxc-walker'

parseAndWalk(code, filename, (node) => {
  // node is typed correctly
})

We could likely use the oxc crates in Rust and do something similar to what this repository already does.