Conversion of basis to input type can fail

[gdalle]

The reason is that I use zero in DI.basis to construct basis vectors, and zero(x) doesn't necessarily have the same type as x. Subsequently, when I convert dx for Enzyme to ensure that Duplicated gets two arguments of the same type, I get an error.

julia> x = view(ones(2), 1:2)
2-element view(::Vector{Float64}, 1:2) with eltype Float64:
 1.0
 1.0

julia> zero(x)
2-element Vector{Float64}:
 0.0
 0.0

julia> convert(typeof(x), zero(x))
ERROR: MethodError: Cannot `convert` an object of type Vector{Float64} to an object of type SubArray{Float64, 1, Vector{Float64}, Tuple{UnitRange{Int64}}, true}

Detected in SciML/NonlinearSolve.jl#476 (comment)

Enzyme.onehot uses similar so it has the same problem.

Related:

• [Sugar] handle view inputs for Enzyme.onehot and therefore Enzyme.gradient / Enzyme.jacobian EnzymeAD/Enzyme.jl#1950