Boxed errors

[rben01]

clippy warned that some Result types were quite large due to their massive error variant. For instance, interpreter::RuntimeError is 304 bytes, which means the following function returns a whopping 304 bytes (I'm assuming there's a niche somewhere; if not, then 305) despite its Ok variant having no payload!

numbat/numbat/src/bytecode_interpreter.rs

Lines 308 to 312 in b175c31

	fn compile_statement(
	&mut self,
	stmt: &Statement,
	dimension_registry: &DimensionRegistry,
	) -> Result<()> {

The same goes for TypeCheckError (264 bytes) and NumbatError (also 304 as it wraps RuntimeError). This seemed like a lot of data to allocate space to the error case.

I took clippy’s advice and boxed the errors so that they'd take 8 bytes at most. This obviously adds an indirection to the error path but uses much less stack space for the Ok path.

[sharkdp]

Thank you very much! It would be great if you could resolve the conflict (or let me know if you need help).

[rben01]

Thank you very much! It would be great if you could resolve the conflict (or let me know if you need help).

Done!

[sharkdp]

The wasm build is currently failing. There was an issue on master which I fixed, but now the failure seems to be related to this branch.

[rben01]

Should be good to go now. I was ignoring WASM build errors before but yeah, this one was legit. For whatever reason rust-analyzer can't analyze the numbat-wasm crate, maybe because it can't see through #[wasm_bindgen] (?), so I wasn't getting any error messages.

[sharkdp]

For whatever reason rust-analyzer can't analyze the numbat-wasm crate

The wasm package is not part of the workspace (there was a technical reason why this didn't work, but maybe it's worth revisiting). That's why rust analyzer doesn't see it. It works if you open the numbat-wasm project / folder.