package bayesGame.bayesbayes.nodeCPD;

import java.util.ArrayList;
import java.util.Arrays;

import org.apache.commons.math3.fraction.Fraction;

import bayesGame.bayesbayes.BayesNode;

public class MajorityVote implements NodeCPD {

	public MajorityVote() {
		// TODO Auto-generated constructor stub
	}

	@Override
	public BayesNode getNode(BayesNode sourceBayesNode, Object[] parents) {
		
		Object nodeType = sourceBayesNode.type;
		
		ArrayList<Object> allItems = new ArrayList<Object>(Arrays.asList(parents));
		allItems.add(0, parents);
		
		// generate a truth table as per http://stackoverflow.com/a/10761325/2130838
		int rows = (int) Math.pow(2, allItems.size());
		
        for (int i=0; i<rows; i++) {
        	int balance=0;
        	boolean trueSeen = false;
        	ArrayList<Object> falseObjects = new ArrayList<Object>();
        	for (int j=allItems.size()-1; j>=0; j--) {
                // as we traverse each generated row from the end of the row, we keep track of
            	// for the balance of TRUE-FALSE items. 
        		// Also, for each FALSE item on the row, we add that to the list of false items
        		// on this row.
            	if (j > 0){
                	if((i/(int) Math.pow(2, j))%2 == 1){
                    	balance++;
                    } else {
                    	balance--;
                    	falseObjects.add(allItems.get(j));
                    }
                } else {
                	Fraction probability;
                	// once we get to the beginning of the row, we check whether
                	// this row codes for the node being TRUE or FALSE
                	if((i/(int) Math.pow(2, j))%2 == 1){
                		// if this row codes for the node being TRUE and the
                		// balance is positive, then we set the probability of
                		// this row to one. if the balance is negative we set
                		// the probability of this row to zero, and if the balance
                		// is even we set the probability to .5.
                		if (balance > 0){
                			probability = Fraction.ONE;
                		} else if (balance < 0){
                			probability = Fraction.ZERO;
                		} else{
                			probability = Fraction.ONE_HALF;
                		}
                	} else {
                		// Similarly, if this row codes for the node being FALSE,
                		// we do the reverse.
                		falseObjects.add(nodeType);
                		if (balance > 0){
                			probability = Fraction.ZERO;
                		} else if (balance < 0){
                			probability = Fraction.ONE;
                		} else{
                			probability = Fraction.ONE_HALF;
                		}
                	}
                	
                	sourceBayesNode.setProbabilityOfUntrueVariables(probability, falseObjects.toArray());
                }
            }
        }
        
        String description = "<html>'" + sourceBayesNode.type + "' is a <b>conditional probability variable</b><br>of type <b>majority vote</b>.<p><p>It is true if the majority of its parents are true, <br>and false if the majority of its parents are false. <br>If an equal number of parents are true and false, <br>it has a 50% chance of being true.";
        sourceBayesNode.cptDescription = description;
        
		return sourceBayesNode;
	}

}