How to oxc-parse -> AST -> oxc-minify in JS

[Fd929c2CE5fA]

We have WASM versions of both packages, but I don't see how to use them together?
https://www.npmjs.com/package/oxc-parser
https://www.npmjs.com/package/oxc-minify

oxc-parser seems to only provide one-way TEXT to AST parsing, and it appears not able to to regenerate code after editing.

Meanwhile, oxc-minify seems to only accept TEXT, not the already parsed AST, implying yet another redundant parsing?

How can I combine and utilize these packages in JS?

For example, a use case might be to parse the original JS/TS code, read or change something in the AST, and then regenerate the JS/TS code or directly minify it.

I don't understand Rust and Oxc's internals, but based on the examples I've read:
https://github.com/oxc-project/oxc/blob/main/crates/oxc_minifier/examples/minifier.rs
https://github.com/oxc-project/oxc/blob/main/crates/oxc_codegen/examples/codegen.rs

It seems there are a few issues here:

• The npm/wasm package does not provide the relevant APIs.
• The codegen has not published as a npm/wasm package.
  • Publish codegen as wasm / napi #3274

Are there any plans to provide these supports? Or any other suggestions?

I think rolldown is a bit heavy for this simple task, but should I try and look into rolldown-wasm instead?

• Browser Support rolldown/rolldown#4012

[camc314]

I would strongly consider using rolldown for your use case, as it sounds like it can do what you need, if you need to customize the transformation process, you can use rolldown's plugin hooks to mutate the ast, and use magic string to print the mutated ast.

To print an AST to a string, take a look at this issue #11495

[Fd929c2CE5fA]

Thank you, the issue #11495 is very helpful! So I now know some key information:

• oxc's Rust-AST and JS/TS-AST is not a same shape, so it cannot be returned to oxc-minify
• oxc-parse does not have condegen but we now have 3rd party esrap or prettier
• I'm not sure what the difference is between magic-string and them, maybe some methods

[Fd929c2CE5fA]

My use case was actually fairly simple, I just had to remove some nodes (exports / imports_ from the ast, instead of using regexp
I did not have any particular ast format need. so even if I had access to the rust ast in js and the codgen that would have been fine for my usecase.

Originally posted by @mizulu in #11495

Actually I have a similar simple use case, and I thought oxc-parser should be enough compared to the complexity of rolldown.

But I now even think that in the absence of built-in codegen in oxc-parser, manual processing may be a more straightforward approach.

What I mean is that AST has given us the indexes of the code block, and we can use these starting and ending indexes to manually splice the original string, which may be enough for simple needs. It's even cheaper than traversing whole AST do codegen.