Vectors not working in structs

[David-OConnor]

I'm receiving the following error:
the traitstd::marker::Copyis not implemented forstd::vec::Vec

for this struct:

#[wasm_bindgen]
pub struct Mesh{
    pub tris: Vec<u32>,
}

If I change tris' type to u32, the error goes away.

This page lists Vectors and slices of supported integer types as supported.

[alexcrichton]

Thanks for the report! Currently though public struct fields are a special case in that they only work with Copy types. This should definitely be better documented!

[David-OConnor]

Thanks; removing pub on the field fixed it

[David-OConnor]

Related Q: How do you extract a value from a struct passed via bindgen to JS? The object I see when printing to console.log() contains a ptr field of a large integer, and a very deep recursive structure containing values like , free, constructor, bind, call etc. Eg, how can I make mesh.tris work, eg returning a number[] ?

This page in the official guide shows example struct passing from Rust to JS, but doesn't show accessing a value contained in one.

edit: Getter functions, like in the example on that guide appear to work. Is there a way around this? Perhaps auto-generating getter funcs for all struct fields?

[alexcrichton]

@David-OConnor ah that's got its own separate bug for defining JS getters in Rust

[David-OConnor]

@alexcrichton Perhaps the solution, assuming there's an obstacle to exposing fields directly, is to serialize the struct, pass it to JS, then deserialize into an object. This assumes no methods.

[alexcrichton]

Oh currently the main obstacle in a sense is that it's not clear what to do if the field isn't Copy because JS gets just a snapshot and mutating that isn't clear you're not mutating the original Rust value. For example in JS if you did mesh.tris.push(3) it wouldn't necessarily be reflected back in the tris field in Rust.

[David-OConnor]

That makes sense. One way to handle this is to send one-way structs to JS, that don't call back to the Rust code. I got it working like this, using serde:

Rust:

#[derive(Serialize)]
#[wasm_bindgen]
pub struct MyStruct {
    // All fields here are are serializable by Serde, or are structs that derive Serialize themselves.
}

#[wasm_bindgen]
pub fn get_mystruct() -> String {
    let mystruct = MyStruct {..};

    serde_json::to_string(&mystruct).unwrap()
}

JS:

const rust = import("./from_rust");
rust.then(
    r =>
    {
        let myStruct = JSON.parse(get_mystruct());
    })

No methods, but can be fed into a constructor for a JS object that has methods, or just used as-is, optionally by declaring it with a Typescript interface.

@alexcrichton Do you think this would be useful in a more streamlined, official way, eg that handles the serialization/deserialization automatically? This seems like it could be a common use-case. Eg make Rust code that does something complicated or interesting, then call a Rust func from JS, which passes the result in whichever data structures are convenient... At least when I heard of WASM and bindgen, this was the first use-case that came to mind.

[alexcrichton]

@David-OConnor oh we already have those apis actually!

wasm-bindgen/src/lib.rs

Lines 153 to 206 in 1a84901

	/// Creates a new `JsValue` from the JSON serialization of the object `t`
	/// provided.
	///
	/// This function will serialize the provided value `t` to a JSON string,
	/// send the JSON string to JS, parse it into a JS object, and then return
	/// a handle to the JS object. This is unlikely to be super speedy so it's
	/// not recommended for large payloads, but it's a nice to have in some
	/// situations!
	///
	/// Usage of this API requires activating the `serde-serialize` feature of
	/// the `wasm-bindgen` crate.
	///
	/// # Errors
	///
	/// Returns any error encountered when serializing `T` into JSON.
	#[cfg(feature = "serde-serialize")]
	pub fn from_serde<T>(t: &T) -> serde_json::Result<JsValue>
	where
	T: serde::ser::Serialize + ?Sized,
	{
	let s = serde_json::to_string(t)?;
	unsafe {
	Ok(JsValue {
	idx: __wbindgen_json_parse(s.as_ptr(), s.len()),
	})
	}
	}
	/// Invokes `JSON.stringify` on this value and then parses the resulting
	/// JSON into an arbitrary Rust value.
	///
	/// This function will first call `JSON.stringify` on the `JsValue` itself.
	/// The resulting string is then passed into Rust which then parses it as
	/// JSON into the resulting value.
	///
	/// Usage of this API requires activating the `serde-serialize` feature of
	/// the `wasm-bindgen` crate.
	///
	/// # Errors
	///
	/// Returns any error encountered when parsing the JSON into a `T`.
	#[cfg(feature = "serde-serialize")]
	pub fn into_serde<T>(&self) -> serde_json::Result<T>
	where
	T: for<'a> serde::de::Deserialize<'a>,
	{
	unsafe {
	let mut ptr = ptr::null_mut();
	let len = __wbindgen_json_serialize(self.idx, &mut ptr);
	let s = Vec::from_raw_parts(ptr, len, len);
	let s = String::from_utf8_unchecked(s);
	serde_json::from_str(&s)
	}
	}

[David-OConnor]

@alexcrichton Sick!

[eminence]

You also cannot store a pub String in a struct, presumably for the same reason outlined in this issue. Strings are pretty fundamental types. Could we expand the documentation to mention this limitation? https://rustwasm.github.io/docs/wasm-bindgen/reference/types/string.html

[benma]

This ByteString struct can't be returned from a function because the Vec is not Copy:

#[wasm_bindgen]
pub struct ByteString {
    pub bytes: Vec<u8>,
}

But a clone directly in a getter works:

#[wasm_bindgen]
impl ByteString {
    pub fn get(&self) -> Vec<u8> {
        self.bytes.clone()
    }
}

This adds a seemingly useless indrection - now every Vec that is returned anywhere needs to be wrapped in such a struct and accessed via a getter. Is there a reason why this could not work directly? Would be very appreciated.

Same for String.

[daxpedda]

#439 (comment)

Oh currently the main obstacle in a sense is that it's not clear what to do if the field isn't Copy because JS gets just a snapshot and mutating that isn't clear you're not mutating the original Rust value. For example in JS if you did mesh.tris.push(3) it wouldn't necessarily be reflected back in the tris field in Rust.

This would require some design work and figuring out what to do here.

[brainstorm]

@daxpedda Would generating getters for all struct members as @David-OConnor suggested be a good initial compromise (usability for perf) or were you thinking on other (better) design patterns from the get-go?

I'm willing to give either a try after fixing my cargo issues with wasm-bindgen.

[daxpedda]

It would have to be done with the help of an attribute, because doing it by default is probably a bad idea.

Personally I'm not much in favor of doing this, as this can be done outside of wasm-bindgen.
I'm generally very hesitant to keep increasing the scope of this project.