java - Find all paths in a graph with DFS

Question

Good morning!

I'm developing an algorithm to find all the paths in an undirected, not weighted graph. I'm currently using a DFS algortihm with backtracking to try and do that. Here is my current code:

import java.util.*;

public class dfs {

    private static Map<Integer, LinkedHashSet<Integer>> map = new HashMap<Integer, LinkedHashSet<Integer>>();
    private int startNode;
    private int numLinks;

    public dfs(int startNode, int numLinks) {
        super();
        this.startNode = startNode;
        this.numLinks = numLinks;
    }

    public void addEdge(int source, int destiny) {
        LinkedHashSet<Integer> adjacente = map.get(source);
        if(adjacente==null) {
            adjacente = new LinkedHashSet<Integer>();
            map.put(source, adjacente);
        }
        adjacente.add(destiny);
    }

    public void addLink(int source, int destiny) {
        addEdge(source, destiny);
        addEdge(destiny, source);
    }

    public LinkedList<Integer> adjacentNodes(int last) {
        LinkedHashSet<Integer> adjacente = map.get(last);
        System.out.println("adjacentes:" + adjacente);
        if(adjacente==null) {
            return new LinkedList<Integer>();
        }
        return new LinkedList<Integer>(adjacente);
    }


public static void main(String[] args) {

    Scanner input = new Scanner(System.in);

    int numVertices = input.nextInt();
    int numLinks = input.nextInt();
    int startNode = input.nextInt();
    int endNode = startNode;

    dfs mapa = new dfs(startNode, numLinks);

    for(int i = 0; i<numLinks; i++){
        mapa.addLink(input.nextInt(), input.nextInt());
    }

    List<ArrayList<Integer>> paths = new ArrayList<ArrayList<Integer>>();
    List<Integer> visited = new ArrayList<Integer>();
    visited.add(startNode);
    Integer currentNode = 0;

    Iterator it = map.entrySet().iterator();
    while (it.hasNext()) {
        Map.Entry pairs = (Map.Entry)it.next();
        currentNode = (Integer) pairs.getKey(); 
        //System.out.println("Current Node:" + currentNode);
        mapa.findAllPaths(mapa, visited, paths, currentNode);

    }
}

private void findAllPaths(dfs mapa, List<Integer> visited,
        List<ArrayList<Integer>> paths, Integer currentNode) {

    if (currentNode.equals(startNode)) { 
        paths.add(new ArrayList<Integer>(visited));

        LinkedList<Integer> nodes = mapa.adjacentNodes(currentNode); 
        //System.out.println("visited:" + visited);

        for (Integer node : nodes) {
            //System.out.println("nodes:" + nodes);
            List<Integer> temp = new ArrayList<Integer>();
            temp.addAll(visited);
            temp.add(node);          
            findAllPaths(mapa, temp, paths, node);
        }

    }

    else {
        LinkedList<Integer> nodes = mapa.adjacentNodes(currentNode);  
        System.out.println("currentNode:" + currentNode);
        //System.out.println("nodes:" + nodes);
        for (Integer node : nodes) {            
            if (visited.contains(node)) {
                continue;
            } 
            List<Integer> temp = new ArrayList<Integer>();
            temp.addAll(visited);
            System.out.println("visited:" + visited);
            temp.add(node);          
            findAllPaths(mapa, temp, paths, node);
        }
    }

} 

}

The program receives integers on his input. The first one is the number of nodes, the second one is the number of links and the third is the start node and end note, which are the same. All the integers that come after represent the connections between nodes.

The problem is, this algorithm is finding all the paths that visit a single node only once. What i want is the algorithm to find all the paths that visit each connection only once. Any idea on how i can do that?

Answer 1

You are on the right track - backtracking is a neat way to solve it.

To get all paths that "uses the same edge only once": after you use an edge in findAllPaths() - delete it from the set of edges [delete the connection from the LinkedHashSet of each vertex of this edge] - and invoke recursively.

After you return from the recursion - don't forget to "clean up the environment" and add this edge back to both vertices.

You will need to make sure you don't run into troubles of iterating collection while modifying it. [You cannot do it - the result of doing so is unexpected] - so you will probably need to send a copy of the LinkedHashSets [without the relevant edge] - and not the original one.