SplineBoundaryCondition.NotAKnot or CubicSpline.InterpolateNotAKnot

[rummelsworth]

I'm porting some IMSL-based Fortran code to C#, including some calls to CSIEZ, which does cubic spline interpolation using the "not-a-knot" condition. I'm not terribly knowledgeable about splines, so I found this nice page that gives a concise visual and mathematical comparison of several boundary conditions, including "natural" and "not-a-knot".

For right now, CubicSpline's natural method seems like it might be close enough, closer anyway than Akima or parabolically terminated boundaries for my particular inputs.

[markusr]

You need to add the calculation of the left and right boundary of NotAKnot condition to CubicSpline.cs

See scipy implementation as a reference.

[ztmsimon]

Now in the MathNet.Numerics.Interpolation.CubicSpline,only implements the boundary-condition-type of 'natural', don't implements the boundary-condition-type of 'not-a-knot'.But, in other software, such as Python scipy.interpolate.CubicSpline or MATLAB spline(), the default boundary-condition-type is 'not-a-knot'. Are there any plans to implement the 'not-a-knot' type?

[xingtianxia]

I have the same problem, how do you solve this at last?