Fix convert(::Type{<:StaticArray}, ::UniformScaling) (#759)

Author: c42f

src/linalg.jl

@@ -102,6 +102,7 @@ end
 # because we do not have `SArray{Tuple{2,3}} <: StaticMatrix`.
 (::Type{SM})(I::UniformScaling) where {SM<:(StaticArray{Tuple{N,M}} where {N,M})} =
     _scalar_matrix(Size(SM), SM, I.λ)
+convert(T::Type{<:StaticArrayLike}, I::UniformScaling) = T(I)
 
 # Construct a matrix with the scalar λ on the diagonal and zeros off the
 # diagonal. The matrix can be non-square.

test/linalg.jl

@@ -75,6 +75,13 @@ using LinearAlgebra: checksquare
         @test @inferred(SDiagonal{3}(2.0I))::SDiagonal{3,Float64} == 2I3x3
         @test @inferred(SDiagonal{3,Float64}(I))::SDiagonal{3,Float64} == I3x3
         @test @inferred(SDiagonal{3,Float32}(2.0I))::SDiagonal{3,Float32} == 2I3x3
+
+        # convert from UniformScaling falls back to constructors
+        @test @inferred(convert(SMatrix{2,2,Float64}, I)) === SA_F64[1 0; 0 1]
+        @test @inferred(convert(MMatrix{2,2,Float64}, I))::MMatrix{2,2,Float64} == SA[1 0; 0 1]
+        @test @inferred(convert(SMatrix{2,2}, I)) === SA{Bool}[1 0; 0 1]
+        @test @inferred(convert(SArray{Tuple{2,2}}, I)) === SA{Bool}[1 0; 0 1]
+        @test @inferred(convert(SDiagonal{2}, I)) === Diagonal{Bool,SArray{Tuple{2},Bool,1,2}}((1,1))
     end
 
     @testset "diagm()" begin