Interpolation not type stable?

[BeastyBlacksmith]

@code_warntype spline( 0.5 )

yields Any for the return type, when I tried this today.
Is it me or is it type unstable?

[kbarbary]

I don't see this for Spline1D. What kind of spline are you using?

[BeastyBlacksmith]

ParametricSpline

[xanfus]

using Dierckx

julia> spline = Spline1D([1,2],[1,2],k=1)
Spline1D(knots=[1.0, 2.0], k=1, extrapolation="nearest", residual=0.0)

julia> @code_warntype 1.0*spline(1.0)
Variables
  #self#::Core.Compiler.Const(*, false)
  x::Float64
  y::Float64

Body::Float64
1 \u2500 %1 = Base.mul_float(x, y)::Float64
\u2514\u2500\u2500      return %1

julia> f(x) = 1.0*spline(x)
f (generic function with 1 method)

julia> @code_warntype f(1.0)
Variables
  #self#::Core.Compiler.Const(f, false)
  x::Float64

Body::Any
1 \u2500 %1 = Main.spline(x)::Any
\u2502   %2 = (1.0 * %1)::Any
\u2514\u2500\u2500      return %2

[xanfus]

Spline has to be defined right in scope of function, which calls the spline. Access to dataframe (S0) by index isn't type-stable, so that leaf-type arrays have to be sent to input of Spline1D.

begin
    S0 = CSV.read("S0.csv",header=false)#Rn222
    S0x = S0[:,2];S0zzi = S0[:,1];
    # S0sp = Spline1D(S0x,S0zzi,k=1)
    # FetchActivity(zzi) = 4.0*S0sp(zzi)
    function FetchActivity(zzi)
        # Dierckx.evaluate(S0sp,zzi)
        S0sp = Spline1D(S0x,S0zzi,k=1)
        # S0sp = Spline1D(S0[:,2],S0[:,1],k=1)
        # x = S0sp(zzi)
        return 4*S0sp(zzi)
    end
end