#[var(set = function_name)] is not compatible with #[func(rename = gdscript_name)]

[Houtamelo]

Example:

#[derive(GodotClass)]
#[class(init, base = CharacterBody2D)]
pub struct Herbivore {
	#[var(get, set = gdscript_set_health)]
	pub(super) health: f64,
}

#[godot_api]
impl Herbivore {
	#[func(rename = set_health)]
	fn gdscript_set_health(&mut self, value: f64) { 
		self.set_health(value);
	}
}

This compiles in Rust but in Godot we get a runtime error when trying to invoke the property:

This happens because inside Godot its using the original name, not the renamed one.

Not sure if this is by design or a niche bug.

[Bromeon]

Interesting! Certainly not intended.

Might be a bit tricky because those are in two separate macros, and I don't know if the #[func(rename)] info is "stored" somewhere to be accessible by code generated from #[derive(Godot)]. It should be possible, but might require an extra indirection for each get/set function reference.

[Bromeon]

One idea to address this: #[godot_api] generates a compile-time lookup table with all #[func]s and their registered names. So

#[godot_api]
impl Herbivore {
	#[func]
	fn eat_herbs(&mut self, value: f64) {...}

	#[func(rename = set_health)]
	fn gdscript_set_health(&mut self, value: f64) {...}
}

generates something like

mod __funcs_Herbivore {
      pub const eat_herbs: &str = "eat_herbs";
      pub const gdscript_set_health: &str = "set_health";
}

Then, the custom setter

#[var(get, set = gdscript_set_health)]
pub(super) health: f64,

would, instead of registering the setter as "gdscript_set_health" in Godot, look up the name as __funcs_Herbivore::gdscript_set_health. This solves two problems:

1. It takes into account renamed methods.
2. If someone forgets #[func] above gdscript_set_health, this causes a compile-time error (at the moment: runtime error).

[0x53A]

That got kinda harder with my PR, where you can now have multiple #[godot_api].

Since you can't contribute to one mod from multiple places, the constants would have to live in the surrounding module.

#[godot_api]
impl Herbivore {
	#[func]
	fn eat_herbs(&mut self, value: f64) {...}
}

// generated
pub const __funcs_Herbivore_eat_herbs: &str = "eat_herbs";


#[godot_api]
impl Herbivore {
	#[func(rename = set_health)]
	fn gdscript_set_health(&mut self, value: f64) {...}
}

// generated
pub const __funcs_Herbivore_gdscript_set_health: &str  = "set_health";

[Bromeon]

Ugh... back to C namespacing then 😔

(It could be impl scoped in this case... but generally I don't like polluting neither the struct's nor the surrounding scope)

[0x53A]

would you merge a PR for this anyway? if yes, what pattern exactly?

/// Generated by the #[func] macro of gdext. Maps the name under which it is exported to godot to the rust name of the function.
#[doc(hidden)]
pub(self) const __func_{class}_{rust-fn-name}: &str  = "{godot-fn-name}";

(I could take a stab at implementing it next week or so)

[Bromeon]

Yes, a PR would still be appreciated! It would already fix the issue, and if we find another approach in the future, we can always change it 🙂

You could maybe use the __gdext_ prefix like here:

gdext/godot-core/src/init/mod.rs

Line 21 in d32d569

pub unsafe fn __gdext_load_library<E: ExtensionLibrary>(

Also, pub(self) is redundant , so it could be along:

#[doc(hidden)]
const __gdext_func_{class}_{rust-fn-name}: &str  = "{godot-fn-name}";