export_class alternative for napi runtime?

[CumpsD]

I am trying to create a very small test binding:

use neon::prelude::*;

#[neon::main]
fn main(mut cx: ModuleContext) -> NeonResult<()> {
    cx.export_class::<Blake3Hash>("Hasher");
    Ok(())
}

pub struct Blake3Hash {
    hasher: blake3::Hasher,
}

impl Blake3Hash {}

impl Default for Blake3Hash {
    fn default() -> Self {
        Self { hasher: blake3::Hasher::new() }
    }
}

impl Finalize for Blake3Hash { }

My cargo.toml is created by create neon:

[package]
name = "blakish"
version = "0.1.0"
license = "ISC"
edition = "2018"
exclude = ["index.node"]

[lib]
crate-type = ["cdylib"]

[dependencies]
blake3 = "1.2"

[dependencies.neon]
version = "0.9"
default-features = false
features = ["napi-6"]

However on cx.export_class::<Blake3Hash>("Hasher"); I am getting an error the method is not found because it is apparently only for the legacy-runtime.

    #[cfg(feature = "legacy-runtime")]
    /// Convenience method for exporting a Neon class constructor from a module.
    pub fn export_class<T: Class>(&mut self, key: &str) -> NeonResult<()> {
        let constructor = T::constructor(self)?;
        self.exports.set(self, key, constructor)?;
        Ok(())
    }

Is there still a way to export a class with napi? Or do we have to wrap it all in functions now with JsBox?

[kjvalencik]

Correct. Classes are only implemented in the legacy runtime. Their design is not compatible with with web workers and were not easily ported. There is proposal to add them (neon-bindings/rfcs#43), but it's not being currently worked on.

In the meantime, JsBox provides everything you need when combined with some JavaScript glue code. The idea is that store your Rust data in a JsBox and create a wrapper class in JavaScript that passes it to functions, either as the first argument or bound to this.

This example demonstrates how: https://github.com/neon-bindings/examples/blob/main/examples/async-sqlite/index.js

https://docs.rs/neon/latest/neon/types/struct.JsBox.html

[CumpsD]

Ok I will have to have a look. I am getting this with the legacy runtime: connor4312/blake3#38

Trying to move to the new one, but it needs to expose that class, will have to refactor I guess