This small C++ library contains my working code for performing natural cubic spline interpolation in one, two, and three dimensions.
dependencies: OpenMP (omp.h)
The CubicSpline class is fully functional. A test program is available in CubicSpineTest.cc, the results of which I have plotted in CubicSplineTest.png using Python.
BiCubicSpline and TriCubicSpline are still being developed. Documentation and usage will be provided in the future.
##Building
You should be able to just run 'make'. This compiles and links the programs. The user will either need to make this available on their standard include path or link to this directory.
I recommend something like:
export CSLDIR=/path/to/CubicSplineLib
export CSLINC="-I$CSLDIR -L$CSLDIR -lCubicSpline -fopenmp"
After which, programs can be compiled similar to:
g++ your_program.cc $CSLINC##Short Example Usage
// create spline object
CubicSpline<double> raw_data( x, y, 2 );
// resample 'x' data
std::vector<double> new_y = raw_data.interpolate( new_x );The user can construct the spline object with float, double, or long double types. The third argument in the constructor is optional and has a default value of one. This says what the maximum allowed threads should be for OpenMP. There is a fourth optional argument that takes either true or false. By default this implementation uses a quicksort algorithm to ensure the horizontal axis is in ascending order before building the spline polynomials. This behavior can be supressed by specifying false. Also, in the interpolation function, an optional second and third argument taking either true or false and a numeric type the same as the template respectively allow for returning fixed values outside the domain of the original data.
e.g.,
raw_data.interpolate( new_x, true, 0.0 );would return a std::vector containing 0.0 for all positions outside the original domain of 'x'.