Argument limited to Functions

[matthieugomez]

FiniteDifferences does not support objects that can be called without being a subtype of Function. It's a bit weird since Julia does not support multiple inheritances. Can the restriction <: Function be removed? If not it would be good to discuss the issue/workaround in the README.

using Interpolations
f = linear_interpolation(1:10, 1:10)
central_fdm(5, 1)(x -> f(x), 1.5)
# 1.000000000000018
central_fdm(5, 1)(f, 1.5)
#ERROR: MethodError: no method matching (::FiniteDifferences.AdaptedFiniteDifferenceMethod{5, 1, FiniteDifferences.UnadaptedFiniteDifferenceMethod{7, 5}})(::Interpolations.Extrapolation{Float64, 1, ScaledInterpolation{Float64, 1, Interpolations.BSplineInterpolation{Float64, 1, Vector{Float64}, BSpline{Linear{Throw{OnGrid}}}, Tuple{Base.OneTo{Int64}}}, BSpline{Linear{Throw{OnGrid}}}, Tuple{UnitRange{Int64}}}, BSpline{Linear{Throw{OnGrid}}}, Throw{Nothing}}, ::Float64)
#Closest candidates are:
#  (::FiniteDifferences.AdaptedFiniteDifferenceMethod)(::TF, ::Real) where TF<:Function
#   @ FiniteDifferences ~/.julia/packages/FiniteDifferences/zZFW0/src/methods.jl:191
#  (::FiniteDifferences.AdaptedFiniteDifferenceMethod{P, Q})(::TF, ::Real, ::Real) where {P, Q, TF<:Function}
#   @ FiniteDifferences ~/.julia/packages/FiniteDifferences/zZFW0/src/methods.jl:238

[oxinabox]

Easy work around is to wrap whatever you want to AD in a anon function.

I am trying to remember if there is a reason we don't just allow nonfunction subtypes.

[tpapp]

The purpose of this restriction is not apparent to me either. Yes, the workaround is trivial, but slightly tedious for unit tests etc.

Would you consider accepting a PR removing the restriction it if it can be made to work?

[matthieugomez]

Thanks!