Functionnal lib

Simple header-only set of classes for efficient Polynomials and Bezier curves evaluations, in arbitrary dimensions. This library is made to stay simple and easy to maintain.

Compilation

Use Cmake to build the sources. For the root folder:

mkdir build
cd build
cmake ..
make

Usage

Checkout the demos/ folder for concrete examples.

Types of curves

All the curves are implemented using the struct template < typename EvalBase > struct Functionnal, and provides the following API:

OutputVectorType eval(const InputVectorType& x) const;  // evaluate the input function f(x)
Derivative derivative() const;                          // construct a new object f'(x)
operator << (Stream s) const;                           // print to input stream s: f(x) = 3x^2 + 1

This API is implemented by the EvalBase template class.

In addition to Functionnal, we also provides the classes FunctionnalMap and ConstFunctionnalMap to use the same API with objects storing the functionnal coefficients as Eigen::Map. This is very useful if you need to use the library with raw buffers for storage and synchronisation. Both FunctionnalMap and ConstFunctionnalMap constructors take a pointer to the coefficient memory as input. They also provide with a asFunctionnal() method to create duplicate objects. For instance:

BezierCurve bezier ( {
                        0.20, 1.0 ,
                        0.35, 0.25,
                        0.45, 0.25,
                        0.80, 1.0
                      } );
BezierMap bezierMap ( bezier.coeffs.data() );       // changes in bezier will affect bezierMap
BezierCurve cBezierCopy = bezierMap.asFunctionnal(); // duplicate memory: objects are independants

Bezier

Example with cubic bezier curves in 3D

typedef double Scalar;
enum{Dim=2};
enum{Degree=3};
typedef functionnal::Bezier<Scalar, Degree, Dim> BezierCurve;
BezierCurve bezier ( {
                        0.20, 1.0 ,
                        0.35, 0.25,
                        0.45, 0.25,
                        0.80, 1.0
                      } );
const int nbSample = 100;
for (int t = 0; t != nbSample; ++t){
    BezierCurve::InputVectorType input;
    input << Scalar(t)/Scalar(nbSample-1);
    std::cout << bezier.eval( input ).transpose() << std::endl;
}

N-Dimensionnal Quadric

Defined as in https://en.wikipedia.org/wiki/Quadric. Example of use in 2D:

enum{Dim = 2};
using Scalar  = float;
using Quadric = functionnal::Quadric<Scalar,  Dim>;

Quadric q;

// set f(x,y) = x^2 + y^2
Quadric::EvalBase::getQMap( q.coeffs ) = Quadric::EvalBase::QType::Identity();
Quadric::EvalBase::getPMap( q.coeffs ) = Quadric::EvalBase::PType::Zero();

// evaluate quadric:
Scalar res0 = q.eval( {{ 1., 1. }} );
Scalar res1 = q.eval( Quadric::InputVectorType::Random() );

Coefficient-wise polynomial of arbitrary degree

Evaluate a 1D polynomial function of arbitrary degree on a set of scalar values.

using Curve = functionnal::CWisePolynomial<Scalar, Degree, NbElemt>;
Curve c2 ( { 0.5, 0, -2, 0.7 } );        // 0.5 - 2x^2 + 0.7x^3

typename Curve::InputVectorType  input  = Curve::InputVectorType::Random();

// Process all the input samples at once
typename Curve::OutputVectorType o2 = c2.eval( input );

for (int i = 0; i < NbElemt; ++i)
        std::cout << input(i) << " " << o2(i) << "\n";
    std::cout << std::flush;

yunxiaoshan / Functionnal