eigen(TensorValue) returning wrong values

[miguelmaso]

Consider the following example:

using Gridap
using LinearAlgebra
A = TensorValue{2,2,Float64}(1,2,3,4)

When trying to perform an eigen decomposition, it is giving wrong results:

julia> eigen(A)
Eigen{TensorValue{2, 2, Float64, 4}, TensorValue{2, 2, Float64, 4}, Matrix{TensorValue{2, 2, Float64, 4}}, Vector{TensorValue{2, 2, Float64, 4}}}
values:
1-element Vector{TensorValue{2, 2, Float64, 4}}:
 TensorValue{2, 2, Float64, 4}(1.0, 2.0, 3.0, 4.0)
vectors:
1×1 Matrix{TensorValue{2, 2, Float64, 4}}:
 TensorValue{2, 2, Float64, 4}(1.0, 0.0, 0.0, 1.0)

julia> eigen(get_array(A))
Eigen{Float64, Float64, StaticArraysCore.SMatrix{2, 2, Float64, 4}, StaticArraysCore.SVector{2, Float64}}
values:
2-element StaticArraysCore.SVector{2, Float64} with indices SOneTo(2):
 -0.3722813232690143
  5.372281323269014
vectors:
2×2 StaticArraysCore.SMatrix{2, 2, Float64, 4} with indices SOneTo(2)×SOneTo(2):
 -0.909377  -0.565767
  0.415974  -0.824565

It seems that TensorValue is missing an interface in order to be compatible with LinearAlgebra. Any clue how to fix this?

[JordiManyer]

I guess no-one has tried to do this before. I think it should dispatch to the StaticArrays interface anyways, like you do below. It's just a matter of adding the signature.

[zjwegert]

While we're at it, It might also be useful to add this for SymFourthOrderTensorValue. I believe this would require us to have a method for going back and forth between Mandel notation.

[JordiManyer]

Done!

[miguelmaso]

Done!

Thanks!