Operations on type paramteres

General Usage
1

Is there is an easy way to do something like this

struct TensorField{T,ND,NC} <: AbstractArray{T,ND+NC}
end

I need to define a scaler/vector/tensor field in 1D/2D/3D and I don’t want to define a type for each dimension.

2

My guess is you are trying to indicate the number of covariant and contravariant indices/dimensions with ND and NC, and they must add up to the total number of array dimensions N. You can’t add numeric type variables in type definitions, but you can add their concrete values inside type constructors. Thus, something like the following should work.

struct TensorField{T,ND,NC,N} <: AbstractArray{T,N}
   ...
   function TensorField{T,ND,NC} where {T,ND,NC}
       ...
       new{T, ND, NC, ND+NC}(...)
   end
end

This should essentially enforce the invariant that ND+NC == N for any instance of TensorField. You could alternatively use an outer constructor and include a check in the inner constructor that ND+NC == N or else raise an exception.

3 Likes
3

You can’t do this yet, but there have been discussions of supporting it someday: Proposal: Defer calculation of field types until type parameters are known · Issue #18466 · JuliaLang/julia · GitHub