The current implementation of lyap for complex scalar arguments is incorrect. We can see this by comparing the result obtained using 1x1 matrices vs scalars:

julia> A = (1.0+2.0im)*ones(1,1);

julia> C = (3.0+4.0im)*ones(1,1);

julia> lyap(A, C)  # expected
1×1 Matrix{ComplexF64}:
 -1.5 - 2.0im

julia> lyap(A[1], C[1])  # wrong! 
-1.1 + 0.2im

julia> -C[1] / 2real(A[1])  

Currently we've implemented a bunch of functions and constraints in a "lazy evaluation" fashion, i.e., whenever we need something for a Style program, we implement it in Penrose. We're getting to the point where we should really have a more complete/comprehensive list (e.g., that users can call at will).

How do we come up with this list? One natural idea is to simply pick some libraries, and

It should be noted that it currently applies to all of the existing packages.

Minor cleanup of nomenclature. In Type.hs we have:

vectorRef :: Reference
vectorRef = Reference.Builtin "Sequence"

vector :: Ord v => a -> Type v a
vector a = ref a vectorRef

We have to keep the Builtin "Sequence" as is, since otherwise that changes the identity of the type, but the names of vectorRef and vector should be changed to use the list nomenclature:

Environment

• SuperCollider version:
• 3.11.1
• Operating system:
• Windows 10 (?)

Steps to reproduce

// macOS:
"~".pathMatch; 
// -> [ "/Users/<myusername>"]
// windows: 
"~".pathMatch; 
// -> [ ];
// this fixes the behaviour, but is clumsy;
"~".standardizePath.pathMatch;  
// -> [ "C:\Users\<myusern

Should suppress --strict warnings.