/** * @file * @brief Implementation of the Composite Simpson Rule for the approximation * * @details The following is an implementation of the Composite Simpson Rule for * the approximation of definite integrals. More info -> wiki: * https://en.wikipedia.org/wiki/Simpson%27s_rule#Composite_Simpson's_rule * * The idea is to split the interval in an EVEN number N of intervals and use as * interpolation points the xi for which it applies that xi = x0 + i*h, where h * is a step defined as h = (b-a)/N where a and b are the first and last points * of the interval of the integration [a, b]. * * We create a table of the xi and their corresponding f(xi) values and we * evaluate the integral by the formula: I = h/3 * {f(x0) + 4*f(x1) + 2*f(x2) + * ... + 2*f(xN-2) + 4*f(xN-1) + f(xN)} * * That means that the first and last indexed i f(xi) are multiplied by 1, * the odd indexed f(xi) by 4 and the even by 2. * * In this program there are 4 sample test functions f, g, k, l that are * evaluated in the same interval. * * Arguments can be passed as parameters from the command line argv[1] = N, * argv[2] = a, argv[3] = b * * N must be even number and a /// for assert #include /// for math functions #include #include /// for integer allocation #include /// for std::atof #include /// for std::function #include /// for IO operations #include /// for std::map container /** * @namespace numerical_methods * @brief Numerical algorithms/methods */ namespace numerical_methods { /** * @namespace simpson_method * @brief Contains the Simpson's method implementation */ namespace simpson_method { /** * @fn double evaluate_by_simpson(int N, double h, double a, * std::function func) * @brief Calculate integral or assert if integral is not a number (Nan) * @param N number of intervals * @param h step * @param a x0 * @param func: choose the function that will be evaluated * @returns the result of the integration */ double evaluate_by_simpson(std::int32_t N, double h, double a, const std::function& func) { std::map data_table; // Contains the data points. key: i, value: f(xi) double xi = a; // Initialize xi to the starting point x0 = a // Create the data table double temp = NAN; for (std::int32_t i = 0; i <= N; i++) { temp = func(xi); data_table.insert( std::pair(i, temp)); // add i and f(xi) xi += h; // Get the next point xi for the next iteration } // Evaluate the integral. // Remember: f(x0) + 4*f(x1) + 2*f(x2) + ... + 2*f(xN-2) + 4*f(xN-1) + f(xN) double evaluate_integral = 0; for (std::int32_t i = 0; i <= N; i++) { if (i == 0 || i == N) { evaluate_integral += data_table.at(i); } else if (i % 2 == 1) { evaluate_integral += 4 * data_table.at(i); } else { evaluate_integral += 2 * data_table.at(i); } } // Multiply by the coefficient h/3 evaluate_integral *= h / 3; // If the result calculated is nan, then the user has given wrong input // interval. assert(!std::isnan(evaluate_integral) && "The definite integral can't be evaluated. Check the validity of " "your input.\n"); // Else return return evaluate_integral; } /** * @fn double f(double x) * @brief A function f(x) that will be used to test the method * @param x The independent variable xi * @returns the value of the dependent variable yi = f(xi) */ double f(double x) { return std::sqrt(x) + std::log(x); } /** @brief Another test function */ double g(double x) { return std::exp(-x) * (4 - std::pow(x, 2)); } /** @brief Another test function */ double k(double x) { return std::sqrt(2 * std::pow(x, 3) + 3); } /** @brief Another test function*/ double l(double x) { return x + std::log(2 * x + 1); } } // namespace simpson_method } // namespace numerical_methods /** * \brief Self-test implementations * @param N is the number of intervals * @param h is the step * @param a is x0 * @param b is the end of the interval * @param used_argv_parameters is 'true' if argv parameters are given and * 'false' if not */ static void test(std::int32_t N, double h, double a, double b, bool used_argv_parameters) { // Call the functions and find the integral of each function double result_f = numerical_methods::simpson_method::evaluate_by_simpson( N, h, a, numerical_methods::simpson_method::f); assert((used_argv_parameters || (result_f >= 4.09 && result_f <= 4.10)) && "The result of f(x) is wrong"); std::cout << "The result of integral f(x) on interval [" << a << ", " << b << "] is equal to: " << result_f << std::endl; double result_g = numerical_methods::simpson_method::evaluate_by_simpson( N, h, a, numerical_methods::simpson_method::g); assert((used_argv_parameters || (result_g >= 0.27 && result_g <= 0.28)) && "The result of g(x) is wrong"); std::cout << "The result of integral g(x) on interval [" << a << ", " << b << "] is equal to: " << result_g << std::endl; double result_k = numerical_methods::simpson_method::evaluate_by_simpson( N, h, a, numerical_methods::simpson_method::k); assert((used_argv_parameters || (result_k >= 9.06 && result_k <= 9.07)) && "The result of k(x) is wrong"); std::cout << "The result of integral k(x) on interval [" << a << ", " << b << "] is equal to: " << result_k << std::endl; double result_l = numerical_methods::simpson_method::evaluate_by_simpson( N, h, a, numerical_methods::simpson_method::l); assert((used_argv_parameters || (result_l >= 7.16 && result_l <= 7.17)) && "The result of l(x) is wrong"); std::cout << "The result of integral l(x) on interval [" << a << ", " << b << "] is equal to: " << result_l << std::endl; } /** * @brief Main function * @param argc commandline argument count (ignored) * @param argv commandline array of arguments (ignored) * @returns 0 on exit */ int main(int argc, char** argv) { std::int32_t N = 16; /// Number of intervals to divide the integration /// interval. MUST BE EVEN double a = 1, b = 3; /// Starting and ending point of the integration in /// the real axis double h = NAN; /// Step, calculated by a, b and N bool used_argv_parameters = false; // If argv parameters are used then the assert must be omitted // for the tst cases // Get user input (by the command line parameters or the console after // displaying messages) if (argc == 4) { N = std::atoi(argv[1]); a = std::atof(argv[2]); b = std::atof(argv[3]); // Check if a 0 && "N has to be > 0"); if (N < 16 || a != 1 || b != 3) { used_argv_parameters = true; } std::cout << "You selected N=" << N << ", a=" << a << ", b=" << b << std::endl; } else { std::cout << "Default N=" << N << ", a=" << a << ", b=" << b << std::endl; } // Find the step h = (b - a) / N; test(N, h, a, b, used_argv_parameters); // run self-test implementations return 0; }