No method for isless(::Complex128,::Int64)

[jebej]

It seems that there is no method for comparing complex numbers to integers, which seems to be required for (some) matrix exponentials, and for all complex matrix exponentials.

i.e., the following happens intermitently, depending on the random matrix

julia> rand(3,3)^2.3
ERROR: MethodError: no method matching isless(::Complex{Float64}, ::Int64)
Closest candidates are:
  isless(::AbstractFloat, ::Real) at operators.jl:98
  isless(::Real, ::Real) at operators.jl:266
Stacktrace:
 [1] (::Base.LinAlg.##17#18)(::Complex{Float64}) at .\<missing>:0
 [2] broadcast_t(::Function, ::Type{Any}, ::Tuple{Base.OneTo{Int64}}, ::Cartesia
nRange{CartesianIndex{1}}, ::Array{Complex{Float64},1}) at .\broadcast.jl:255
 [3] broadcast_c at .\broadcast.jl:298 [inlined]
 [4] broadcast(::Function, ::Array{Complex{Float64},1}) at .\broadcast.jl:415
 [5] ^(::Array{Float64,2}, ::Float64) at .\linalg\dense.jl:320

but this always throws an error:

julia> rand(Complex128,3,3)^2.3
ERROR: MethodError: no method matching isless(::Complex{Float64}, ::Int64)
...

[tkelman]

how would you compare complex numbers to real integers?

it's a bug in the matrix exponential, being worked on in #21184, that it's trying to do this

[jebej]

It seems that defining

Base.isless(a::Complex,b::Real) = isless(norm(a),b)

gives the right matrix exponential, and I think this is how MATLAB defines < as well.

[jebej]

Actually scratch that, it seems MATLAB compares the real part only.

[JeffBezanson]

Complex numbers don't have a well-defined total order. We should just fix matrix exponential.

[jebej]

Yes. Does #21184 also take care of complex matrix exponentials? If so we can close this issue.

[ararslan]

Does #21184 also take care of complex matrix exponentials?

Yes