Submanifold index arrays should be typed

[chakravala]

Submanifold index arrays should be typed

[chakravala]

The type should not be explicitly declared, since it isn't always going to be Int. If you check with @code_warntype the method is already type stable, and sometimes it will not return an Int. For example, one might define the metric with Float64 instead of Int.

I will have to investigate this issue more, could you provide a minimum reproducible example>

[eschnett]

Found it... Was my mistake after all. The code below reproduces the problem:

using ComputedFieldTypes
using StaticArrays
using Grassmann

# @basis 1
V = v1
Vec(X) = Chain{V,1,X}
Mat(X) = Chain{V,1, fulltype(Vec(X))}
function Base.rand(::Type{<:Chain{V,1,T}}) where {V, T}
    D = ndims(V)
    Chain{V,1}(SVector{D,T}((rand(T) for i in 1:ndims(V))...))::Chain{V,1,<:T}
end
A = rand(Mat(Float64))
x = rand(Vec(Float64))

y = A ⋅ x

Of course, I didn't intend to define V in this way. If you remove this definition and use the @basis instead, the problem goes away.