package joptima.functions;

import joptima.Parameter;

/**
 * Function norm / (sigma * sqrt(2pi)) * exp(- (x - mean)^2 / (2 * sigma^2)).
 * A concrete implementation of joptima.Function that calculates
 * a Gaussian.
 */
public class Gaussian extends ConstrainableFunction {

  private static final double sqrttwopi = Math.sqrt(2 * Math.PI);

  public double m_norm, m_mean, m_sigma;

  public Parameter[] getParameters() {
    Parameter[] parameters = new Parameter[countUnconstrained()];
    try {
      int cursor = 0;
      if (!m_constrained[0]) parameters[cursor++] = new Parameter(this, "norm");
      if (!m_constrained[1]) parameters[cursor++] = new Parameter(this, "mean");
      if (!m_constrained[2]) parameters[cursor++] = new Parameter(this, "sigma");
    } catch (NoSuchFieldException nsfe) {
      nsfe.printStackTrace();
    }
    return parameters;
  }

  public void constrain(String parametername) {
    if ("norm".equals(parametername)) m_constrained[0] = true;
    if ("mean".equals(parametername)) m_constrained[1] = true;
    if ("sigma".equals(parametername)) m_constrained[2] = true;
  }

  private String m_name = "Gaussian";
  public String getName() { return m_name; }
  public void setName(String newname) { m_name = newname; }

  /**
   * Default constructor: norm = 1, mean = 0, sigma = 1.
   */
  public Gaussian() {
    this(1, 0, 1);
  }

  /**
   * Constructor.
   */
  public Gaussian(double norm, double mean, double sigma) {
    this.m_norm = norm;
    this.m_mean = mean;
    this.m_sigma = sigma;
    m_constrained = new boolean[3];
  }

  public double calculate(double x) {
    return m_norm / m_sigma / sqrttwopi *
      Math.exp(- (x - m_mean) * (x - m_mean) / (2 * m_sigma * m_sigma));
  }

  public double getNorm() {
    return m_norm;
  }

  public void setNorm(double norm) {
    m_norm = norm;
  }

  public double getMean() {
    return m_mean;
  }

  public void setMean(double mean) {
    m_mean = mean;
  }

  public double getSigma() {
    return m_sigma;
  }

  public void setSigma(double sigma) {
    m_sigma = sigma;
  }
}