help in the Donalds B. Johnson's algorithm, i cannot understand the pseudo code (PART II)

Question

i cannot understand a certain part of the paper published by Donald Johnson about finding cycles (Circuits) in a graph.

More specific i cannot understand what is the

Answer 1

@trashgod, your sample output contains cycle which are cyclic permutation. For example 0-1-0 and 1-0-1 are same Actually the output should contains only 5 cycle i.e. 0 1 0, 0 2 0, 0 1 2 0, 0 2 1 0, 1 2 1,

Johnson paper explain what a cycle is: 'Two elementary circuits are distinct if one is not a cyclic permutation of the other. ' One can also check wolfram page: This also output 5 cycle for the same input.

http://demonstrations.wolfram.com/EnumeratingCyclesOfADirectedGraph/

Answer 2

The following variation produces unique cycles. Based on this example, it is adapted from an answer supplied by @user1406062.

Code:

import java.util.ArrayList;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Stack;

/**
 * @see https://en.wikipedia.org/wiki/Johnson%27s_algorithm
 * @see https://stackoverflow.com/questions/2908575
 * @see https://stackoverflow.com/questions/2939877
 * @see http://en.wikipedia.org/wiki/Adjacency_matrix
 * @see http://en.wikipedia.org/wiki/Adjacency_list
 */
public final class CircuitFinding {

    final Stack<Integer> stack = new Stack<Integer>();
    final Map<Integer, List<Integer>> a;
    final List<List<Integer>> b;
    final boolean[] blocked;
    final int n;
    Integer s;

    public static void main(String[] args) {
        List<List<Integer>> a = new ArrayList<List<Integer>>();
        a.add(new ArrayList<Integer>(Arrays.asList(1, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 1)));
        CircuitFinding cf = new CircuitFinding(a);
        cf.find();
    }

    /**
     * @param a adjacency structure of strong component K with least vertex in
     * subgraph of G induced by {s, s + 1, n};
     */
    public CircuitFinding(List<List<Integer>> A) {
        this.a = new HashMap<Integer, List<Integer>>(A.size());
        for (int i = 0; i < A.size(); i++) {
            this.a.put(i, new ArrayList<Integer>());
            for (int j : A.get(i)) {
                this.a.get(i).add(j);
            }
        }
        n = a.size();
        blocked = new boolean[n];
        b = new ArrayList<List<Integer>>();
        for (int i = 0; i < n; i++) {
            b.add(new ArrayList<Integer>());
        }
    }

    private void unblock(int u) {
        blocked[u] = false;
        List<Integer> list = b.get(u);
        for (int w : list) {
            //delete w from B(u);
            list.remove(Integer.valueOf(w));
            if (blocked[w]) {
                unblock(w);
            }
        }
    }

    private boolean circuit(int v) {
        boolean f = false;
        stack.push(v);
        blocked[v] = true;
        L1:
        for (int w : a.get(v)) {
            if (w == s) {
                //output circuit composed of stack followed by s;
                for (int i : stack) {
                    System.out.print(i + " ");
                }
                System.out.println(s);
                f = true;
            } else if (!blocked[w]) {
                if (circuit(w)) {
                    f = true;
                }
            }
        }
        L2:
        if (f) {
            unblock(v);
        } else {
            for (int w : a.get(v)) {
                //if (v∉B(w)) put v on B(w);
                if (!b.get(w).contains(v)) {
                    b.get(w).add(v);
                }
            }
        }
        v = stack.pop();
        return f;
    }

    public void find() {
        s = 0;
        while (s < n) {
            if (!a.isEmpty()) {
                //s := least vertex in V;
                L3:
                for (int i : a.keySet()) {
                    b.get(i).clear();
                    blocked[i] = false;
                }
                circuit(s);
                a.remove(s);
                for (Integer j : a.keySet()) {
                    if (a.get(j).contains(s)) {
                        a.get(j).remove(s);
                    }
                }
                s++;
            } else {
                s = n;
            }
        }
    }
}

Output:

0 1 0
0 1 2 0
0 2 0
0 2 1 0
1 2 1

All cycles, for reference:

0 1 0
0 1 2 0
0 2 0
0 2 1 0
1 0 1
1 0 2 1
1 2 0 1
1 2 1
2 0 1 2
2 0 2
2 1 0 2
2 1 2

Answer 3

It works! In an earlier iteration of the Johnson algorithm, I had supposed that A was an adjacency matrix. Instead, it appears to represent an adjacency list. In that example, implemented below, the vertices {a, b, c} are numbered {0, 1, 2}, yielding the following circuits.

Addendum: As noted in this proposed edit and helpful answer, the algorithm specifies that unblock() should remove the element having the value w, not the element having the index w.

list.remove(Integer.valueOf(w));

Sample output:

0 1 0
0 1 2 0
0 2 0
0 2 1 0
1 0 1
1 0 2 1
1 2 0 1
1 2 1
2 0 1 2
2 0 2
2 1 0 2
2 1 2

By default, the program starts with s = 0; implementing s := least vertex in V as an optimization remains. A variation that produces only unique cycles is shown here.

import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import java.util.Stack;

/**
 * @see http://dutta.csc.ncsu.edu/csc791_spring07/wrap/circuits_johnson.pdf
 * @see https://stackoverflow.com/questions/2908575
 * @see https://stackoverflow.com/questions/2939877
 * @see http://en.wikipedia.org/wiki/Adjacency_matrix
 * @see http://en.wikipedia.org/wiki/Adjacency_list
 */
public final class CircuitFinding {

    final Stack<Integer> stack = new Stack<Integer>();
    final List<List<Integer>> a;
    final List<List<Integer>> b;
    final boolean[] blocked;
    final int n;
    int s;

    public static void main(String[] args) {
        List<List<Integer>> a = new ArrayList<List<Integer>>();
        a.add(new ArrayList<Integer>(Arrays.asList(1, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 2)));
        a.add(new ArrayList<Integer>(Arrays.asList(0, 1)));
        CircuitFinding cf = new CircuitFinding(a);
        cf.find();
    }

    /**
     * @param a adjacency structure of strong component K with
     * least vertex in subgraph of G induced by {s, s + 1, n};
     */
    public CircuitFinding(List<List<Integer>> a) {
        this.a = a;
        n = a.size();
        blocked = new boolean[n];
        b = new ArrayList<List<Integer>>();
        for (int i = 0; i < n; i++) {
            b.add(new ArrayList<Integer>());
        }
    }

    private void unblock(int u) {
        blocked[u] = false;
        List<Integer> list = b.get(u);
        for (int w : list) {
            //delete w from B(u);
            list.remove(Integer.valueOf(w));
            if (blocked[w]) {
                unblock(w);
            }
        }
    }

    private boolean circuit(int v) {
        boolean f = false;
        stack.push(v);
        blocked[v] = true;
        L1:
        for (int w : a.get(v)) {
            if (w == s) {
                //output circuit composed of stack followed by s;
                for (int i : stack) {
                    System.out.print(i + " ");
                }
                System.out.println(s);
                f = true;
            } else if (!blocked[w]) {
                if (circuit(w)) {
                    f = true;
                }
            }
        }
        L2:
        if (f) {
            unblock(v);
        } else {
            for (int w : a.get(v)) {
                //if (v∉B(w)) put v on B(w);
                if (!b.get(w).contains(v)) {
                    b.get(w).add(v);
                }
            }
        }
        v = stack.pop();
        return f;
    }

    public void find() {
        while (s < n) {
            if (a != null) {
                //s := least vertex in V;
                L3:
                circuit(s);
                s++;
            } else {
                s = n;
            }
        }
    }
}

Answer 4

I had sumbitted an edit request to @trashgod's code to fix the exception thrown in unblock(). Essentially, the algorithm states that the element w (which is not an index) is to be removed from the list. The code above used list.remove(w), which treats w as an index.

My edit request was rejected! Not sure why, because I have tested the above with my modification on a network of 20,000 nodes and 70,000 edges and it doesn't crash.

I have also modified Johnson's algorithm to be more adapted to undirected graphs. If anybody wants these modifications please contact me.

Below is my code for unblock().

private void unblock(int u) {
    blocked[u] = false;
    List<Integer> list = b.get(u);
    int w;
    for (int iw=0; iw < list.size(); iw++) {
        w = Integer.valueOf(list.get(iw));
        //delete w from B(u);
        list.remove(iw);
        if (blocked[w]) {
            unblock(w);
        }
    }
}