#include "PolyMath.h"

#include "CoolPropTools.h"
#include "Exceptions.h"
#include "MatrixMath.h"

#include <vector>
#include <string>
//#include <sstream>
//#include <numeric>
#include <math.h>
#include <iostream>

#include "Solvers.h"

#include "unsupported/Eigen/Polynomials"

namespace CoolProp{

/// Basic checks for coefficient vectors.
/** Starts with only the first coefficient dimension
 *  and checks the matrix size against the parameters rows and columns.
 */
/// @param coefficients matrix containing the ordered coefficients
/// @param rows unsigned integer value that represents the desired degree of the polynomial in the 1st dimension
/// @param columns unsigned integer value that represents the desired degree of the polynomial in the 2nd dimension
bool Polynomial2D::checkCoefficients(const Eigen::MatrixXd &coefficients, const unsigned int rows, const unsigned int columns){
	if (coefficients.rows() == rows) {
		if (coefficients.cols() == columns) {
			return true;
		} else {
			throw ValueError(format("%s (%d): The number of columns %d does not match with %d. ",__FILE__,__LINE__,coefficients.cols(),columns));
		}
	} else {
		throw ValueError(format("%s (%d): The number of rows %d does not match with %d. ",__FILE__,__LINE__,coefficients.rows(),rows));
	}
	return false;
}


/// Integration functions
/** Integrating coefficients for polynomials is done by dividing the
 *  original coefficients by (i+1) and elevating the order by 1
 *  through adding a zero as first coefficient.
 *  Some reslicing needs to be applied to integrate along the x-axis.
 *  In the brine/solution equations, reordering of the parameters
 *  avoids this expensive operation. However, it is included for the
 *  sake of completeness.
 */
/// @param coefficients matrix containing the ordered coefficients
/// @param axis integer value that represents the desired direction of integration
/// @param times integer value that represents the desired order of integration
Eigen::MatrixXd Polynomial2D::integrateCoeffs(const Eigen::MatrixXd &coefficients, const int &axis = -1, const int &times = 1){
	if (times < 0) throw ValueError(format("%s (%d): You have to provide a positive order for integration, %d is not valid. ",__FILE__,__LINE__,times));
	if (times == 0) return Eigen::MatrixXd(coefficients);
	Eigen::MatrixXd oldCoefficients;
	Eigen::MatrixXd newCoefficients(coefficients);

	switch (axis) {
	case 0:
		newCoefficients = Eigen::MatrixXd(coefficients);
		break;
	case 1:
		newCoefficients = Eigen::MatrixXd(coefficients.transpose());
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	std::size_t r, c;
	for (int k = 0; k < times; k++){
		oldCoefficients = Eigen::MatrixXd(newCoefficients);
		r = oldCoefficients.rows(), c = oldCoefficients.cols();
		newCoefficients = Eigen::MatrixXd::Zero(r+1,c);
		newCoefficients.block(1,0,r,c) = oldCoefficients.block(0,0,r,c);
		for (size_t i = 0; i < r; ++i) {
			for (size_t j = 0; j < c; ++j) newCoefficients(i+1,j) /= (i+1.);
		}
	}

	switch (axis) {
	case 0:
		break;
	case 1:
		newCoefficients.transposeInPlace();
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	return newCoefficients;
}


/// Derivative coefficients calculation
/** Deriving coefficients for polynomials is done by multiplying the
 *  original coefficients with i and lowering the order by 1.
 */
/// @param coefficients matrix containing the ordered coefficients
/// @param axis integer value that represents the desired direction of derivation
/// @param times integer value that represents the desired order of integration
Eigen::MatrixXd Polynomial2D::deriveCoeffs(const Eigen::MatrixXd &coefficients, const int &axis, const int &times){
	if (times < 0) throw ValueError(format("%s (%d): You have to provide a positive order for derivation, %d is not valid. ",__FILE__,__LINE__,times));
	if (times == 0) return Eigen::MatrixXd(coefficients);
	// Recursion is also possible, but not recommended
	//Eigen::MatrixXd newCoefficients;
	//if (times > 1) newCoefficients = deriveCoeffs(coefficients, axis, times-1);
	//else newCoefficients = Eigen::MatrixXd(coefficients);
	Eigen::MatrixXd newCoefficients;

	switch (axis) {
	case 0:
		newCoefficients = Eigen::MatrixXd(coefficients);
		break;
	case 1:
		newCoefficients = Eigen::MatrixXd(coefficients.transpose());
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	std::size_t r, c, i, j;
	for (int k = 0; k < times; k++){
		r = newCoefficients.rows(), c = newCoefficients.cols();
		for (i = 1; i < r; ++i) {
			for (j = 0; j < c; ++j) newCoefficients(i,j) *= i;
		}
		removeRow(newCoefficients,0);
	}

	switch (axis) {
	case 0:
		break;
	case 1:
		newCoefficients.transposeInPlace();
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	return newCoefficients;
}

/// The core functions to evaluate the polynomial
/** It is here we implement the different special
 *  functions that allow us to specify certain
 *  types of polynomials.
 *  The derivative might bee needed during the
 *  solution process of the solver. It could also
 *  be a protected function...
 */
/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input
double Polynomial2D::evaluate(const Eigen::MatrixXd &coefficients, const double &x_in){
	double result = Eigen::poly_eval( makeVector(coefficients), x_in );
	if (this->do_debug()) std::cout << "Running      1D evaluate(" << mat_to_string(coefficients) << ", x_in:" << vec_to_string(x_in) << "): " << result << std::endl;
	return result;
}
/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param y_in double value that represents the current input in the 2nd dimension
double Polynomial2D::evaluate(const Eigen::MatrixXd &coefficients, const double &x_in, const double &y_in){
	size_t r = coefficients.rows();
	double result = evaluate(coefficients.row(r-1), y_in);
	for(int i=r-2; i>=0; i--) {
		result *= x_in;
		result += evaluate(coefficients.row(i), y_in);
	}
	if (this->do_debug()) std::cout << "Running      2D evaluate(" << mat_to_string(coefficients) << ", x_in:" << vec_to_string(x_in) << ", y_in:" << vec_to_string(y_in) << "): " << result << std::endl;
	return result;
}


/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param y_in double value that represents the current input in the 2nd dimension
/// @param axis unsigned integer value that represents the axis to derive for (0=x, 1=y)
double Polynomial2D::derivative(const Eigen::MatrixXd &coefficients, const double &x_in, const double &y_in, const int &axis = -1){
	return this->evaluate(this->deriveCoeffs(coefficients, axis, 1), x_in, y_in);
}


/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param y_in double value that represents the current input in the 2nd dimension
/// @param axis unsigned integer value that represents the axis to integrate for (0=x, 1=y)
double Polynomial2D::integral(const Eigen::MatrixXd &coefficients, const double &x_in, const double &y_in, const int &axis = -1){
	return this->evaluate(this->integrateCoeffs(coefficients, axis, 1), x_in,y_in);
}

/// Uses the Brent solver to find the roots of p(x_in,y_in)-z_in
/// @param res Poly2DResidual object to calculate residuals and derivatives
/// @param min double value that represents the minimum value
/// @param max double value that represents the maximum value
double Polynomial2D::solve_limits(Poly2DResidual* res, const double &min, const double &max){
	if (do_debug()) std::cout << format("Called solve_limits with: min=%f and max=%f", min, max) << std::endl;
	std::string errstring;
	double macheps = DBL_EPSILON;
	double tol     = DBL_EPSILON*1e3;
	int    maxiter = 10;
	double result = Brent(res, min, max, macheps, tol, maxiter, errstring);
	if (this->do_debug()) std::cout << "Brent solver message: " << errstring << std::endl;
	return result;
}

/// Uses the Newton solver to find the roots of p(x_in,y_in)-z_in
/// @param res Poly2DResidual object to calculate residuals and derivatives
/// @param guess double value that represents the start value
double Polynomial2D::solve_guess(Poly2DResidual* res, const double &guess){
	if (do_debug()) std::cout << format("Called solve_guess with: guess=%f ", guess) << std::endl;
	std::string errstring;
	//set_debug_level(1000);
	double tol     = DBL_EPSILON*1e3;
	int    maxiter = 10;
	double result = Newton(res, guess, tol, maxiter, errstring);
	if (this->do_debug()) std::cout << "Newton solver message: " << errstring << std::endl;
	return result;
}


/// @param coefficients vector containing the ordered coefficients
/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param axis unsigned integer value that represents the axis to solve for (0=x, 1=y)
Eigen::VectorXd Polynomial2D::solve(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const int &axis = -1){
	std::size_t r = coefficients.rows(), c = coefficients.cols();
	Eigen::VectorXd tmpCoefficients;
	switch (axis) {
	case 0:
		tmpCoefficients = Eigen::VectorXd::Zero(r);
		for(size_t i=0; i<r; i++) {
			tmpCoefficients(i,0) = evaluate(coefficients.row(i), in);
		}
		break;
	case 1:
		tmpCoefficients = Eigen::VectorXd::Zero(c);
		for(size_t i=0; i<c; i++) {
			tmpCoefficients(i,0) = evaluate(coefficients.col(i), in);
		}
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for the solver, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}
	tmpCoefficients(0,0) -= z_in;
	if (this->do_debug()) std::cout << "Coefficients: " << mat_to_string(Eigen::MatrixXd(tmpCoefficients)) << std::endl;
	Eigen::PolynomialSolver<double,Eigen::Dynamic> polySolver( tmpCoefficients );
	std::vector<double> rootsVec;
	polySolver.realRoots(rootsVec);
	if (this->do_debug()) std::cout << "Real roots: " << vec_to_string(rootsVec) << std::endl;
	return vec_to_eigen(rootsVec);
	//return rootsVec[0]; // TODO: implement root selection algorithm
}

/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param axis unsigned integer value that represents the axis to solve for (0=x, 1=y)
double Polynomial2D::solve_limits(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const double &min, const double &max, const int &axis){
	Poly2DResidual res = Poly2DResidual(*this, coefficients, in, z_in, axis);
	return solve_limits(&res, min, max);
}

/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param guess double value that represents the start value
/// @param axis unsigned integer value that represents the axis to solve for (0=x, 1=y)
double Polynomial2D::solve_guess(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const double &guess, const int &axis){
	Poly2DResidual res = Poly2DResidual(*this, coefficients, in, z_in, axis);
	return solve_guess(&res, guess);
}


/// Simple polynomial function generator. <- Deprecated due to poor performance, use Horner-scheme instead
/** Base function to produce n-th order polynomials
 *  based on the length of the coefficient vector.
 *  Starts with only the first coefficient at x^0. */
double Polynomial2D::simplePolynomial(std::vector<double> const& coefficients, double x){
	double result = 0.;
	for(unsigned int i=0; i<coefficients.size();i++) {
		result += coefficients[i] * pow(x,(int)i);
	}
	if (this->do_debug()) std::cout << "Running simplePolynomial(" << vec_to_string(coefficients) << ", " << vec_to_string(x) << "): " << result << std::endl;
	return result;
}
double Polynomial2D::simplePolynomial(std::vector<std::vector<double> > const& coefficients, double x, double y){
	double result = 0;
	for(unsigned int i=0; i<coefficients.size();i++) {
		result += pow(x,(int)i) * simplePolynomial(coefficients[i], y);
	}
	if (this->do_debug()) std::cout << "Running simplePolynomial(" << vec_to_string(coefficients) << ", " << vec_to_string(x) << ", " << vec_to_string(y) << "): " << result << std::endl;
	return result;
}


/// Horner function generator implementations
/** Represent polynomials according to Horner's scheme.
 *  This avoids unnecessary multiplication and thus
 *  speeds up calculation.
 */
double Polynomial2D::baseHorner(std::vector<double> const& coefficients, double x){
	double result = 0;
	for(int i=coefficients.size()-1; i>=0; i--) {
		result *= x;
		result += coefficients[i];
	}
	if (this->do_debug()) std::cout << "Running       baseHorner(" << vec_to_string(coefficients) << ", " << vec_to_string(x) << "): " << result << std::endl;
	return result;
}

double Polynomial2D::baseHorner(std::vector< std::vector<double> > const& coefficients, double x, double y){
	double result = 0;
	for(int i=coefficients.size()-1; i>=0; i--) {
		result *= x;
		result += baseHorner(coefficients[i], y);
	}
	if (this->do_debug()) std::cout << "Running       baseHorner(" << vec_to_string(coefficients) << ", " << vec_to_string(x) << ", " << vec_to_string(y) << "): " << result << std::endl;
	return result;
}



Poly2DResidual::Poly2DResidual(Polynomial2D &poly, const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const int &axis){
	switch (axis) {
	case iX:
	case iY:
		this->axis = axis;
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension to the solver, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	this->poly = poly;
	this->coefficients = coefficients;
	this->derIsSet = false;
	this->in = in;
	this->z_in = z_in;
}

double Poly2DResidual::call(double target){
	if (axis==iX) return poly.evaluate(coefficients,target,in)-z_in;
	if (axis==iY) return poly.evaluate(coefficients,in,target)-z_in;
	return -_HUGE;
}

double Poly2DResidual::deriv(double target){
	if (!this->derIsSet) {
		this->coefficientsDer = poly.deriveCoeffs(coefficients,axis);
		this->derIsSet = true;
	}
	return poly.evaluate(coefficientsDer,target,in);
}





//	/// Integration functions
//	/** Integrating coefficients for polynomials is done by dividing the
//	 *  original coefficients by (i+1) and elevating the order by 1
//	 *  through adding a zero as first coefficient.
//	 *  Some reslicing needs to be applied to integrate along the x-axis.
//	 *  In the brine/solution equations, reordering of the parameters
//	 *  avoids this expensive operation. However, it is included for the
//	 *  sake of completeness.
//	 */
//	/// @param coefficients matrix containing the ordered coefficients
//	/// @param axis unsigned integer value that represents the desired direction of integration
//	/// @param times integer value that represents the desired order of integration
//	/// @param firstExponent integer value that represents the first exponent of the polynomial in axis direction
//	Eigen::MatrixXd integrateCoeffs(const Eigen::MatrixXd &coefficients, const int &axis, const int &times, const int &firstExponent);
//
/// Derivative coefficients calculation
/** Deriving coefficients for polynomials is done by multiplying the
 *  original coefficients with i and lowering the order by 1.
 *
 *  Remember that the first exponent might need to be adjusted after derivation.
 *  It has to be lowered by times:
 *  derCoeffs = deriveCoeffs(coefficients, axis, times, firstExponent);
 *  firstExponent -= times;
 *
 */
/// @param coefficients matrix containing the ordered coefficients
/// @param axis unsigned integer value that represents the desired direction of derivation
/// @param times integer value that represents the desired order of derivation
/// @param firstExponent integer value that represents the lowest exponent of the polynomial in axis direction
Eigen::MatrixXd Polynomial2DFrac::deriveCoeffs(const Eigen::MatrixXd &coefficients, const int &axis, const int &times, const int &firstExponent){
	if (times < 0) throw ValueError(format("%s (%d): You have to provide a positive order for derivation, %d is not valid. ",__FILE__,__LINE__,times));
	if (times == 0) return Eigen::MatrixXd(coefficients);
	// Recursion is also possible, but not recommended
	//Eigen::MatrixXd newCoefficients;
	//if (times > 1) newCoefficients = deriveCoeffs(coefficients, axis, times-1);
	//else newCoefficients = Eigen::MatrixXd(coefficients);
	Eigen::MatrixXd newCoefficients;

	switch (axis) {
	case 0:
		newCoefficients = Eigen::MatrixXd(coefficients);
		break;
	case 1:
		newCoefficients = Eigen::MatrixXd(coefficients.transpose());
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	std::size_t r = newCoefficients.rows(), c = newCoefficients.cols();
	std::size_t i, j;
	for (int k = 0; k < times; k++){
		for (i = 0; i < r; ++i) {
			for (j = 0; j < c; ++j) {
				newCoefficients(i,j) *= double(i)+double(firstExponent);
			}
		}
	}

	switch (axis) {
	case 0:
		break;
	case 1:
		newCoefficients.transposeInPlace();
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	return newCoefficients;
}


/// The core functions to evaluate the polynomial
/** It is here we implement the different special
 *  functions that allow us to specify certain
 *  types of polynomials.
 *
 *  Try to avoid many calls to the derivative and integral functions.
 *  Both of them have to calculate the new coefficients internally,
 *  which slows things down. Instead, you should use the deriveCoeffs
 *  and integrateCoeffs functions and store the coefficient matrix
 *  you need for future calls to evaluate derivative and integral.
 */
/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param firstExponent integer value that represents the lowest exponent of the polynomial
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
double Polynomial2DFrac::evaluate(const Eigen::MatrixXd &coefficients, const double &x_in, const int &firstExponent, const double &x_base){
	size_t r = coefficients.rows();
	size_t c = coefficients.cols();

	if ( (r!=1) && (c!=1) ) {
		throw ValueError(format("%s (%d): You have a 2D coefficient matrix (%d,%d), please use the 2D functions. ",__FILE__,__LINE__,coefficients.rows(),coefficients.cols()));
	}
	if ( (firstExponent<0) && (std::abs(x_in-x_base)<DBL_EPSILON)) {
		throw ValueError(format("%s (%d): A fraction cannot be evaluated with zero as denominator, x_in-x_base=%f ",__FILE__,__LINE__,x_in-x_base));
	}

	Eigen::MatrixXd tmpCoeffs(coefficients);
	if ( c==1 ) {
		tmpCoeffs.transposeInPlace();
		c = r;
		r = 1;
	}
	Eigen::MatrixXd newCoeffs;
	double negExp = 0;// First we treat the negative exponents
	double posExp = 0;// then the positive exponents

	for(int i=0; i>firstExponent; i--) { // only for firstExponent<0
		c=tmpCoeffs.cols();
		if (c>0) {
		  negExp += tmpCoeffs(0,0);
		  removeColumn(tmpCoeffs, 0);
		}
		negExp /= x_in-x_base;
	}

	for(int i=0; i<firstExponent; i++) { // only for firstExponent>0
		c = tmpCoeffs.cols();
		newCoeffs = Eigen::MatrixXd::Zero(r,c+1);
		newCoeffs.block(0,1,r,c) = tmpCoeffs.block(0,0,r,c);
		tmpCoeffs = Eigen::MatrixXd(newCoeffs);
	}

	c = tmpCoeffs.cols();
	if (c>0) posExp += Eigen::poly_eval( Eigen::RowVectorXd(tmpCoeffs), x_in-x_base );
	return negExp+posExp;
}

/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param y_in double value that represents the current input in the 2nd dimension
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
double Polynomial2DFrac::evaluate(const Eigen::MatrixXd &coefficients, const double &x_in, const double &y_in, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base){
	if ( (x_exp<0) && (std::abs(x_in-x_base)<DBL_EPSILON)) {
		throw ValueError(format("%s (%d): A fraction cannot be evaluated with zero as denominator, x_in-x_base=%f ",__FILE__,__LINE__,x_in-x_base));
	}
	if ( (y_exp<0) && (std::abs(y_in-y_base)<DBL_EPSILON)) {
		throw ValueError(format("%s (%d): A fraction cannot be evaluated with zero as denominator, y_in-y_base=%f ",__FILE__,__LINE__,y_in-y_base));
	}

	Eigen::MatrixXd tmpCoeffs(coefficients);
	Eigen::MatrixXd newCoeffs;
	size_t r = tmpCoeffs.rows();
	size_t c = tmpCoeffs.cols();
	double negExp = 0;// First we treat the negative exponents
	double posExp = 0;// then the positive exponents

	for(int i=0; i>x_exp; i--) { // only for x_exp<0
		r = tmpCoeffs.rows();
		if (r>0) {
			negExp += evaluate(tmpCoeffs.row(0), y_in, y_exp, y_base);
			removeRow(tmpCoeffs, 0);
		}
		negExp /= x_in-x_base;
	}

	r = tmpCoeffs.rows();
	for(int i=0; i<x_exp; i++) { // only for x_exp>0
		newCoeffs = Eigen::MatrixXd::Zero(r+1,c);
		newCoeffs.block(1,0,r,c) = tmpCoeffs.block(0,0,r,c);
		tmpCoeffs = Eigen::MatrixXd(newCoeffs);
		r += 1; // r = tmpCoeffs.rows();
	}

	//r = tmpCoeffs.rows();
	if (r>0) posExp += evaluate(tmpCoeffs.row(r-1), y_in, y_exp, y_base);
	for(int i=r-2; i>=0; i--) {
		posExp *= x_in-x_base;
		posExp += evaluate(tmpCoeffs.row(i), y_in, y_exp, y_base);
	}
	if (this->do_debug()) std::cout << "Running      2D evaluate(" << mat_to_string(coefficients) << ", " << std::endl;
	if (this->do_debug()) std::cout << "x_in:" << vec_to_string(x_in) << ", y_in:" << vec_to_string(y_in) << ", x_exp:" << vec_to_string(x_exp) << ", y_exp:" << vec_to_string(y_exp) <<  ", x_base:" << vec_to_string(x_base) << ", y_base:" << vec_to_string(y_base) << "): " << negExp+posExp << std::endl;
	return negExp+posExp;
}


/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param y_in double value that represents the current input in the 2nd dimension
/// @param axis integer value that represents the axis to derive for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
double Polynomial2DFrac::derivative(const Eigen::MatrixXd &coefficients, const double &x_in, const double &y_in, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base){
	Eigen::MatrixXd newCoefficients;
	int der_exp,other_exp;
	double der_val,other_val;
	double int_base, other_base;

	switch (axis) {
	case 0:
		newCoefficients = Eigen::MatrixXd(coefficients);
		der_exp = x_exp;
		other_exp = y_exp;
		der_val = x_in;
		other_val = y_in;
		int_base = x_base;
		other_base = y_base;
		break;
	case 1:
		newCoefficients = Eigen::MatrixXd(coefficients.transpose());
		der_exp = y_exp;
		other_exp = x_exp;
		der_val = y_in;
		other_val = x_in;
		int_base = y_base;
		other_base = x_base;
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	const int times = 1;
	newCoefficients = deriveCoeffs(newCoefficients,0,times,der_exp);
	der_exp -= times;

	return evaluate(newCoefficients,der_val,other_val,der_exp,other_exp,int_base,other_base);
}

/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input in the 1st dimension
/// @param y_in double value that represents the current input in the 2nd dimension
/// @param axis integer value that represents the axis to integrate for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
double Polynomial2DFrac::integral(const Eigen::MatrixXd &coefficients, const double &x_in, const double &y_in, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base){

	Eigen::MatrixXd newCoefficients;
	int int_exp,other_exp;
	double int_val,other_val;
	double int_base, other_base;

	switch (axis) {
	case 0:
		newCoefficients = Eigen::MatrixXd(coefficients);
		int_exp = x_exp;
		other_exp = y_exp;
		int_val = x_in;
		other_val = y_in;
		int_base = x_base;
		other_base = y_base;
		break;
	case 1:
		newCoefficients = Eigen::MatrixXd(coefficients.transpose());
		int_exp = y_exp;
		other_exp = x_exp;
		int_val = y_in;
		other_val = x_in;
		int_base = y_base;
		other_base = x_base;
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for integration, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	if (int_exp<-1) throw NotImplementedError(format("%s (%d): This function is only implemented for lowest exponents >= -1, int_exp=%d ",__FILE__,__LINE__,int_exp));

	size_t r = newCoefficients.rows();
	size_t c = newCoefficients.cols();

	if (int_exp==-1) {
		if (std::abs(int_base)<DBL_EPSILON){
			Eigen::MatrixXd tmpCoefficients = newCoefficients.row(0) * log(int_val-int_base);
			newCoefficients = integrateCoeffs(newCoefficients.block(1,0,r-1,c), 0, 1);
			newCoefficients.row(0) = tmpCoefficients;
			return evaluate(newCoefficients,int_val,other_val,0,other_exp,int_base,other_base);
		}
		else {
			// Reduce the coefficients to the integration dimension:
			newCoefficients = Eigen::MatrixXd(r,1);
			for (std::size_t i=0; i<r; i++){
				newCoefficients(i,0) = evaluate(coefficients.row(i), other_val, other_exp, other_base);
			}
			return fracIntCentral(newCoefficients.transpose(),int_val,int_base);
		}
	}

	Eigen::MatrixXd tmpCoeffs;
	r = newCoefficients.rows();
	for(int i=0; i<int_exp; i++) { // only for x_exp>0
		tmpCoeffs = Eigen::MatrixXd::Zero(r+1,c);
		tmpCoeffs.block(1,0,r,c) = newCoefficients.block(0,0,r,c);
		newCoefficients = Eigen::MatrixXd(tmpCoeffs);
		r += 1; // r = newCoefficients.rows();
	}

	return evaluate(integrateCoeffs(newCoefficients, 0, 1),int_val,other_val,0,other_exp,int_base,other_base);

}

/// Returns a vector with ALL the real roots of p(x_in,y_in)-z_in
/// @param coefficients vector containing the ordered coefficients
/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param axis integer value that represents the axis to solve for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
Eigen::VectorXd Polynomial2DFrac::solve(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base){

	Eigen::MatrixXd newCoefficients;
	Eigen::VectorXd tmpCoefficients;
	int solve_exp,other_exp;
	double input;

	switch (axis) {
	case 0:
		newCoefficients = Eigen::MatrixXd(coefficients);
		solve_exp = x_exp;
		other_exp = y_exp;
		input     = in - y_base;
		break;
	case 1:
		newCoefficients = Eigen::MatrixXd(coefficients.transpose());
		solve_exp = y_exp;
		other_exp = x_exp;
		input     = in - x_base;
		break;
	default:
		throw ValueError(format("%s (%d): You have to provide a dimension, 0 or 1, for the solver, %d is not valid. ",__FILE__,__LINE__,axis));
		break;
	}

	if (this->do_debug()) std::cout << "Coefficients: " << mat_to_string(Eigen::MatrixXd(newCoefficients)) << std::endl;

	const size_t r = newCoefficients.rows();
	for(size_t i=0; i<r; i++) {
		newCoefficients(i,0) = evaluate(newCoefficients.row(i), input, other_exp);
	}

	//Eigen::VectorXd tmpCoefficients;
	if (solve_exp>=0) { // extend vector to zero exponent
		tmpCoefficients = Eigen::VectorXd::Zero(r+solve_exp);
		tmpCoefficients.block(solve_exp,0,r,1) = newCoefficients.block(0,0,r,1);
		tmpCoefficients(0,0) -= z_in;
	} else {// check if vector reaches to zero exponent
		int diff = 1 - r - solve_exp; // How many entries have to be added
		tmpCoefficients = Eigen::VectorXd::Zero(r+std::max(diff,0));
		tmpCoefficients.block(0,0,r,1) = newCoefficients.block(0,0,r,1);
		tmpCoefficients(r+diff-1,0) -= z_in;
		throw NotImplementedError(format("%s (%d): Currently, there is no solver for the generalised polynomial, an exponent of %d is not valid. ",__FILE__,__LINE__,solve_exp));
	}

	if (this->do_debug()) std::cout << "Coefficients: " << mat_to_string( Eigen::MatrixXd(tmpCoefficients) ) << std::endl;
	Eigen::PolynomialSolver<double,-1> polySolver( tmpCoefficients );
	std::vector<double> rootsVec;
	polySolver.realRoots(rootsVec);
	if (this->do_debug()) std::cout << "Real roots: " << vec_to_string(rootsVec) << std::endl;
	return vec_to_eigen(rootsVec);
	//return rootsVec[0]; // TODO: implement root selection algorithm
}

/// Uses the Brent solver to find the roots of p(x_in,y_in)-z_in
/// @param coefficients vector containing the ordered coefficients
/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param min double value that represents the minimum value
/// @param max double value that represents the maximum value
/// @param axis integer value that represents the axis to solve for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
double Polynomial2DFrac::solve_limits(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const double &min, const double &max, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base){
	if (do_debug()) std::cout << format("Called solve_limits with: %f, %f, %f, %f, %d, %d, %d, %f, %f",in, z_in, min, max, axis, x_exp, y_exp, x_base, y_base) << std::endl;
	Poly2DFracResidual res = Poly2DFracResidual(*this, coefficients, in, z_in, axis, x_exp, y_exp, x_base, y_base);
	return Polynomial2D::solve_limits(&res, min, max);
} //TODO: Implement tests for this solver

/// Uses the Newton solver to find the roots of p(x_in,y_in)-z_in
/// @param coefficients vector containing the ordered coefficients
/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param guess double value that represents the start value
/// @param axis unsigned integer value that represents the axis to solve for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
double Polynomial2DFrac::solve_guess(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const double &guess, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base){
	if (do_debug()) std::cout << format("Called solve_guess with: %f, %f, %f, %d, %d, %d, %f, %f",in, z_in, guess, axis, x_exp, y_exp, x_base, y_base) << std::endl;
	Poly2DFracResidual res = Poly2DFracResidual(*this, coefficients, in, z_in, axis, x_exp, y_exp, x_base, y_base);
	return Polynomial2D::solve_guess(&res, guess);
} //TODO: Implement tests for this solver

/// Uses the Brent solver to find the roots of Int(p(x_in,y_in))-z_in
/// @param coefficients vector containing the ordered coefficients
/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param min double value that represents the minimum value
/// @param max double value that represents the maximum value
/// @param axis unsigned integer value that represents the axis to solve for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
/// @param int_axis axis for the integration (0=x, 1=y)
double Polynomial2DFrac::solve_limitsInt(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const double &min, const double &max, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base, const int &int_axis){
	Poly2DFracIntResidual res = Poly2DFracIntResidual(*this, coefficients, in, z_in, axis, x_exp, y_exp, x_base, y_base, int_axis);
	return Polynomial2D::solve_limits(&res, min, max);
} //TODO: Implement tests for this solver

/// Uses the Newton solver to find the roots of Int(p(x_in,y_in))-z_in
/// @param coefficients vector containing the ordered coefficients
/// @param in double value that represents the current input in x (1st dimension) or y (2nd dimension)
/// @param z_in double value that represents the current output in the 3rd dimension
/// @param guess double value that represents the start value
/// @param axis unsigned integer value that represents the axis to solve for (0=x, 1=y)
/// @param x_exp integer value that represents the lowest exponent of the polynomial in the 1st dimension
/// @param y_exp integer value that represents the lowest exponent of the polynomial in the 2nd dimension
/// @param x_base double value that represents the base value for a centred fit in the 1st dimension
/// @param y_base double value that represents the base value for a centred fit in the 2nd dimension
/// @param int_axis axis for the integration (0=x, 1=y)
double Polynomial2DFrac::solve_guessInt(const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const double &guess, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base, const int &int_axis){
	Poly2DFracIntResidual res = Poly2DFracIntResidual(*this, coefficients, in, z_in, axis, x_exp, y_exp, x_base, y_base, int_axis);
	return Polynomial2D::solve_guess(&res, guess);
} //TODO: Implement tests for this solver




/** Simple integrated centred(!) polynomial function generator divided by independent variable.
 *  We need to rewrite some of the functions in order to
 *  use central fit. Having a central temperature xbase
 *  allows for a better fit, but requires a different
 *  formulation of the fracInt function group. Other
 *  functions are not affected.
 *  Starts with only the first coefficient at x^0 */

//Helper functions to calculate binomial coefficients:
//http://rosettacode.org/wiki/Evaluate_binomial_coefficients#C.2B.2B
/// @param nValue integer value that represents the order of the factorial
double Polynomial2DFrac::factorial(const int &nValue){
	double value = 1;
    for(int i = 2; i <= nValue; i++) value = value * i;
    return value;
}
/// @param nValue integer value that represents the upper part of the factorial
/// @param nValue2 integer value that represents the lower part of the factorial
double Polynomial2DFrac::binom(const int &nValue, const int &nValue2){
   if(nValue2 == 1) return nValue*1.0;
   return (factorial(nValue)) / (factorial(nValue2)*factorial((nValue - nValue2)));
}
///Helper function to calculate the D vector:
/// @param m integer value that represents order
/// @param x_in double value that represents the current input
/// @param x_base double value that represents the basis for the fit
Eigen::MatrixXd Polynomial2DFrac::fracIntCentralDvector(const int &m, const double &x_in, const double &x_base){
	if (m<1) throw ValueError(format("%s (%d): You have to provide coefficients, a vector length of %d is not a valid. ",__FILE__,__LINE__,m));

	Eigen::MatrixXd D = Eigen::MatrixXd::Zero(1,m);
	double tmp;
	// TODO: This can be optimized using the Horner scheme!
	for (int j=0; j<m; j++){ // loop through row
		tmp = pow(-1.0,j) * log(x_in) * pow(x_base,j);
		for(int k=0; k<j; k++) { // internal loop for every entry
			tmp += binom(j,k) * pow(-1.0,k) / (j-k) * pow(x_in,j-k) * pow(x_base,k);
		}
		D(0,j)=tmp;
	}
	return D;
}
///Indefinite integral of a centred polynomial divided by its independent variable
/// @param coefficients vector containing the ordered coefficients
/// @param x_in double value that represents the current input
/// @param x_base double value that represents the basis for the fit
double Polynomial2DFrac::fracIntCentral(const Eigen::MatrixXd &coefficients, const double &x_in, const double &x_base){
	if (coefficients.rows() != 1) {
		throw ValueError(format("%s (%d): You have a 2D coefficient matrix (%d,%d), please use the 2D functions. ",__FILE__,__LINE__,coefficients.rows(),coefficients.cols()));
	}
	int m = coefficients.cols();
	Eigen::MatrixXd D = fracIntCentralDvector(m, x_in, x_base);
	double result = 0;
	for(int j=0; j<m; j++) {
		result += coefficients(0,j) * D(0,j);
	}
	if (this->do_debug()) std::cout << "Running   fracIntCentral(" << mat_to_string(coefficients) << ", " << vec_to_string(x_in) << ", " << vec_to_string(x_base) << "): " << result << std::endl;
	return result;
}


Poly2DFracResidual::Poly2DFracResidual(Polynomial2DFrac &poly, const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base)
  : Poly2DResidual(poly, coefficients, in, z_in, axis){
	this->x_exp  = x_exp;
	this->y_exp  = y_exp;
	this->x_base = x_base;
	this->y_base = y_base;
}

double Poly2DFracResidual::call(double target){
	if (axis==iX) return poly.evaluate(coefficients,target,in,x_exp,y_exp,x_base,y_base)-z_in;
	if (axis==iY) return poly.evaluate(coefficients,in,target,x_exp,y_exp,x_base,y_base)-z_in;
	return _HUGE;
}

double Poly2DFracResidual::deriv(double target){
	if (axis==iX) return poly.derivative(coefficients,target,in,axis,x_exp,y_exp,x_base,y_base);
	if (axis==iY) return poly.derivative(coefficients,in,target,axis,x_exp,y_exp,x_base,y_base);
	return _HUGE;
}

Poly2DFracIntResidual::Poly2DFracIntResidual(Polynomial2DFrac &poly, const Eigen::MatrixXd &coefficients, const double &in, const double &z_in, const int &axis, const int &x_exp, const int &y_exp, const double &x_base, const double &y_base, const int &int_axis)
  : Poly2DFracResidual(poly, coefficients, in, z_in, axis, x_exp, y_exp, x_base, y_base){
	this->int_axis = int_axis;
}

double Poly2DFracIntResidual::call(double target){
	if (axis==iX) return poly.integral(coefficients,target,in,int_axis,x_exp,y_exp,x_base,y_base)-z_in;
	if (axis==iY) return poly.integral(coefficients,in,target,int_axis,x_exp,y_exp,x_base,y_base)-z_in;
	return _HUGE;
}

double Poly2DFracIntResidual::deriv(double target){
	if (axis==iX) return poly.evaluate(coefficients,target,in,x_exp,y_exp,x_base,y_base);
	if (axis==iY) return poly.evaluate(coefficients,in,target,x_exp,y_exp,x_base,y_base);
	return _HUGE;
}



}



#ifdef ENABLE_CATCH
#include <math.h>
#include <iostream>
#include "catch.hpp"

TEST_CASE("Internal consistency checks and example use cases for PolyMath.cpp","[PolyMath]")
{
	bool PRINT = false;
	std::string tmpStr;

	/// Test case for "SylthermXLT" by "Dow Chemicals"
	std::vector<double> cHeat;
	cHeat.clear();
	cHeat.push_back(+1.1562261074E+03);
	cHeat.push_back(+2.0994549103E+00);
	cHeat.push_back(+7.7175381057E-07);
	cHeat.push_back(-3.7008444051E-20);

	double deltaT = 0.1;
	double Tmin   = 273.15- 50;
	double Tmax   = 273.15+250;
	double Tinc   = 200;

	std::vector<std::vector<double> > cHeat2D;
	cHeat2D.push_back(cHeat);
	cHeat2D.push_back(cHeat);
	cHeat2D.push_back(cHeat);

	Eigen::MatrixXd matrix2D = CoolProp::vec_to_eigen(cHeat2D);

	Eigen::MatrixXd matrix2Dtmp;
	std::vector<std::vector<double> > vec2Dtmp;

	SECTION("Coefficient parsing") {
		CoolProp::Polynomial2D poly;
		CHECK_THROWS(poly.checkCoefficients(matrix2D,4,5));
		CHECK( poly.checkCoefficients(matrix2D,3,4) );
	}

	SECTION("Coefficient operations") {
		Eigen::MatrixXd matrix;
		CoolProp::Polynomial2D poly;

		CHECK_THROWS(poly.integrateCoeffs(matrix2D));

		CHECK_NOTHROW(matrix = poly.integrateCoeffs(matrix2D, 0));
		tmpStr = CoolProp::mat_to_string(matrix2D);
		if (PRINT) std::cout << tmpStr << std::endl;
		tmpStr = CoolProp::mat_to_string(matrix);
		if (PRINT) std::cout << tmpStr << std::endl << std::endl;

		CHECK_NOTHROW(matrix = poly.integrateCoeffs(matrix2D, 1));
		tmpStr = CoolProp::mat_to_string(matrix2D);
		if (PRINT) std::cout << tmpStr << std::endl;
		tmpStr = CoolProp::mat_to_string(matrix);
		if (PRINT) std::cout << tmpStr << std::endl << std::endl;

		CHECK_THROWS(poly.deriveCoeffs(matrix2D));

		CHECK_NOTHROW(matrix = poly.deriveCoeffs(matrix2D, 0));
		tmpStr = CoolProp::mat_to_string(matrix2D);
		if (PRINT) std::cout << tmpStr << std::endl;
		tmpStr = CoolProp::mat_to_string(matrix);
		if (PRINT) std::cout << tmpStr << std::endl << std::endl;

		CHECK_NOTHROW(matrix = poly.deriveCoeffs(matrix2D, 1));
		tmpStr = CoolProp::mat_to_string(matrix2D);
		if (PRINT) std::cout << tmpStr << std::endl;
		tmpStr = CoolProp::mat_to_string(matrix);
		if (PRINT) std::cout << tmpStr << std::endl << std::endl;
	}

	SECTION("Evaluation and test values"){

		Eigen::MatrixXd matrix = CoolProp::vec_to_eigen(cHeat);
		CoolProp::Polynomial2D poly;

		double acc = 0.0001;

		double T = 273.15+50;
		double c = poly.evaluate(matrix, T, 0.0);
		double d = 1834.746;

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = 2.0;
		c = poly.solve(matrix, 0.0, d, 0)[0];
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		CHECK( check_abs(c,T,acc) );
		}

		c = 2.0;
		c = poly.solve_limits(matrix, 0.0, d, -50, 750, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		CHECK( check_abs(c,T,acc) );
		}

		c = 2.0;
		c = poly.solve_guess(matrix, 0.0, d, 350, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		CHECK( check_abs(c,T,acc) );
		}

//		T = 0.0;
//		solve.setGuess(75+273.15);
//		T = solve.polyval(cHeat,c);
//		printf("Should be  :      T = %3.3f \t K    \n",273.15+50.0);
//		printf("From object:      T = %3.3f \t K    \n",T);
//
//		T = 0.0;
//		solve.setLimits(273.15+10,273.15+100);
//		T = solve.polyval(cHeat,c);
//		printf("Should be  :      T = %3.3f \t K    \n",273.15+50.0);
//		printf("From object:      T = %3.3f \t K    \n",T);

	}

	SECTION("Integration and derivation tests") {

		CoolProp::Polynomial2D poly;

		Eigen::MatrixXd matrix(matrix2D);
		Eigen::MatrixXd matrixInt  = poly.integrateCoeffs(matrix, 1);
		Eigen::MatrixXd matrixDer  = poly.deriveCoeffs(matrix, 1);
		Eigen::MatrixXd matrixInt2 = poly.integrateCoeffs(matrix, 1, 2);
		Eigen::MatrixXd matrixDer2 = poly.deriveCoeffs(matrix, 1, 2);

		CHECK_THROWS( poly.evaluate(matrix,0.0) );

		double x = 0.3, y = 255.3, val1, val2, val3, val4;

		//CHECK( std::abs( polyInt.derivative(x,y,0)-poly2D.evaluate(x,y) ) <= 1e-10 );

		std::string tmpStr;

		double acc = 0.001;

		for (double T = Tmin; T<Tmax; T+=Tinc) {
			val1 = poly.evaluate(matrix, x, T-deltaT);
			val2 = poly.evaluate(matrix, x, T+deltaT);
			val3 = (val2-val1)/2/deltaT;
			val4 = poly.evaluate(matrixDer, x, T);
			CAPTURE(T);
			CAPTURE(val3);
			CAPTURE(val4);
			tmpStr = CoolProp::mat_to_string(matrix);
			CAPTURE(tmpStr);
			tmpStr = CoolProp::mat_to_string(matrixDer);
			CAPTURE(tmpStr);
			CHECK( check_abs(val3,val4,acc) );
		}

		for (double T = Tmin; T<Tmax; T+=Tinc) {
			val1 = poly.evaluate(matrixDer, x, T-deltaT);
			val2 = poly.evaluate(matrixDer, x, T+deltaT);
			val3 = (val2-val1)/2/deltaT;
			val4 = poly.evaluate(matrixDer2, x, T);
			CAPTURE(T);
			CAPTURE(val3);
			CAPTURE(val4);
			tmpStr = CoolProp::mat_to_string(matrixDer);
			CAPTURE(tmpStr);
			tmpStr = CoolProp::mat_to_string(matrixDer2);
			CAPTURE(tmpStr);
			CHECK( check_abs(val3,val4,acc) );
		}

		for (double T = Tmin; T<Tmax; T+=Tinc) {
			val1 = poly.evaluate(matrixInt, x, T-deltaT);
			val2 = poly.evaluate(matrixInt, x, T+deltaT);
			val3 = (val2-val1)/2/deltaT;
			val4 = poly.evaluate(matrix, x, T);
			CAPTURE(T);
			CAPTURE(val3);
			CAPTURE(val4);
			tmpStr = CoolProp::mat_to_string(matrixInt);
			CAPTURE(tmpStr);
			tmpStr = CoolProp::mat_to_string(matrix);
			CAPTURE(tmpStr);
			CHECK( check_abs(val3,val4,acc) );
		}

		for (double T = Tmin; T<Tmax; T+=Tinc) {
			val1 = poly.evaluate(matrixInt2, x, T-deltaT);
			val2 = poly.evaluate(matrixInt2, x, T+deltaT);
			val3 = (val2-val1)/2/deltaT;
			val4 = poly.evaluate(matrixInt, x, T);
			CAPTURE(T);
			CAPTURE(val3);
			CAPTURE(val4);
			tmpStr = CoolProp::mat_to_string(matrixInt2);
			CAPTURE(tmpStr);
			tmpStr = CoolProp::mat_to_string(matrixInt);
			CAPTURE(tmpStr);
			CHECK( check_abs(val3,val4,acc) );
		}

		for (double T = Tmin; T<Tmax; T+=Tinc) {
			val1 =   poly.evaluate(matrix, x, T);
			val2 = poly.derivative(matrixInt, x, T, 1);
			CAPTURE(T);
			CAPTURE(val1);
			CAPTURE(val2);
			CHECK( check_abs(val1,val2,acc) );
		}

		for (double T = Tmin; T<Tmax; T+=Tinc) {
			val1 = poly.derivative(matrix, x, T, 1);
			val2 =   poly.evaluate(matrixDer, x, T);
			CAPTURE(T);
			CAPTURE(val1);
			CAPTURE(val2);
			CHECK( check_abs(val1,val2,acc) );
		}

	}


	SECTION("Testing Polynomial2DFrac"){

		Eigen::MatrixXd matrix = CoolProp::vec_to_eigen(cHeat);
		CoolProp::Polynomial2D poly;
		CoolProp::Polynomial2DFrac frac;

		double acc = 0.0001;

		double T = 273.15+50;
		double a,b;
		double c = poly.evaluate(matrix, T, 0.0);
		double d = frac.evaluate(matrix, T, 0.0, 0, 0);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = poly.evaluate(matrix, T, 0.0)/T/T;
		d = frac.evaluate(matrix, T, 0.0, -2, 0);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		matrix = CoolProp::vec_to_eigen(cHeat2D);
		double y = 0.1;
		c = poly.evaluate(matrix, T, y)/T/T;
		d = frac.evaluate(matrix, T, y, -2, 0);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = poly.evaluate(matrix, T, y)/y/y;
		d = frac.evaluate(matrix, T, y, 0, -2);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = poly.evaluate(matrix, T, y)/T/T/y/y;
		d = frac.evaluate(matrix, T, y, -2, -2);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = poly.evaluate(matrix, T, y)/T/T*y*y;
		d = frac.evaluate(matrix, T, y, -2, 2);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}


		matrix = CoolProp::vec_to_eigen(cHeat);
		T = 273.15+50;
		c = 145.59157247249246;
		d = frac.integral(matrix, T, 0.0, 0, -1, 0) - frac.integral(matrix, 273.15+25, 0.0, 0, -1, 0);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}


		T = 423.15;
		c = 3460.895272;
		d = frac.integral(matrix, T, 0.0, 0, -1, 0, 348.15, 0.0);

		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}


		deltaT = 0.01;
		for (T = Tmin; T<Tmax; T+=Tinc) {
			a = poly.evaluate(matrix, T-deltaT, y);
			b = poly.evaluate(matrix, T+deltaT, y);
			c = (b-a)/2.0/deltaT;
			d = frac.derivative(matrix, T, y, 0, 0, 0);
			CAPTURE(a);
			CAPTURE(b);
			CAPTURE(T);
			CAPTURE(c);
			CAPTURE(d);
			tmpStr = CoolProp::mat_to_string(matrix);
			CAPTURE(tmpStr);
			CHECK( check_abs(c,d,acc) );
		}

		T = 273.15+150;
		c = -2.100108045;
		d = frac.derivative(matrix, T, 0.0, 0, 0, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = -0.006456574589;
		d = frac.derivative(matrix, T, 0.0, 0, -1, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = frac.evaluate(matrix, T, 0.0, 2, 0);
		d = frac.solve(matrix, 0.0, c, 0, 2, 0)[0];
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(T,d,acc) );
		}

		c = frac.evaluate(matrix, T, 0.0, 0, 0);
		d = frac.solve(matrix, 0.0, c, 0, 0, 0)[0];
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(T,d,acc) );
		}

		c = frac.evaluate(matrix, T, 0.0, -1, 0);
		CHECK_THROWS(d = frac.solve(matrix, 0.0, c, 0, -1, 0)[0]);
//		{
//		CAPTURE(T);
//		CAPTURE(c);
//		CAPTURE(d);
//		tmpStr = CoolProp::mat_to_string(matrix);
//		CAPTURE(tmpStr);
//		CHECK( check_abs(T,d,acc) );
//		}

		d = frac.solve_limits(matrix, 0.0, c, T-10, T+10, 0, -1, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(T,d,acc) );
		}

		d = frac.solve_guess(matrix, 0.0, c, T-10, 0, -1, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(T,d,acc) );
		}

		c = -0.00004224550082;
		d = frac.derivative(matrix, T, 0.0, 0, -2, 0);
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		CHECK( check_abs(c,d,acc) );
		}

		c = frac.evaluate(matrix, T, 0.0, 0, 0, 0.0, 0.0);
		d = frac.solve(matrix, 0.0, c, 0, 0, 0, 0.0, 0.0)[0];
		{
		CAPTURE(T);
		CAPTURE(c);
		CAPTURE(d);
		tmpStr = CoolProp::mat_to_string(matrix);
		CAPTURE(tmpStr);
		tmpStr = CoolProp::mat_to_string(Eigen::MatrixXd(frac.solve(matrix, 0.0, c, 0, 0, 0, 250, 0.0)));
		CAPTURE(tmpStr);
		CHECK( check_abs(T,d,acc) );
		}

	}



}


#endif /* ENABLE_CATCH */