prelude.new_array type params differ between declaration and signature computation

[aborgna-q]

The type_args are declared as [length, elem_type]

hugr/hugr-core/src/extension/prelude/array.rs

Lines 249 to 250 in 6bf6c82

	new_array | pop_left | pop_right | get | set | swap => {
	vec![TypeArg::BoundedNat { n: self.size }, ty_arg]

But the binary signature computation only uses the element type

hugr/hugr-core/src/extension/prelude/array.rs

Lines 61 to 67 in 6bf6c82

	let params = vec![TypeBound::Any.into()];
	let poly_func_ty = match self {
	ArrayOpDef::new_array => PolyFuncTypeRV::new(
	params,
	FuncValueType::new(vec![elem_ty_var.clone(); n as usize], array_ty),
	),
	ArrayOpDef::pop_left | ArrayOpDef::pop_right => {

Which one is correct?

[ss2165]

Not an issue: the computation is two stage - first the length argument is used to compute a function type polymorphic over the element type, which is later initialised by the second type argument

[aborgna-q]

It'd be nice to declare matching args/params on both places to avoid confusion.
Guppy used the ones in the signature definition, and that caused validation to fail.

It looks like using [length, ty] passes validation, so this is not an urgent issue.

[ss2165]

agreed, it is confusing to say the least
Will defer to fallout from #1433 - we may yet get to some simplifications to the polymorphism that help

[doug-q]

Closing as NOTABUG, reopen if you disagree