Here is an implementation of the algorithm based on the explanation provided on this page
There is a wrapper class to hold the task, cost, and critical path cost. It starts by calculating the critical cost as the maximum critical cost of all dependencies plus its own cost. Then once the critical costs are available it uses a comparator to sort the tasks based on the critical cost with dependency as a tie breaker (choosing randomly if there is no dependency). Note that an exception will be thrown if there is a cycle and it will fail if any of the costs are negative.
Here is the implementation:
public class CriticalPath {
public static void main(String[] args) {
//The example dependency graph from
//http://www.ctl.ua.edu/math103/scheduling/scheduling_algorithms.htm
HashSet<Task> allTasks = new HashSet<Task>();
Task end = new Task("End", 0);
Task F = new Task("F", 2, end);
Task A = new Task("A", 3, end);
Task X = new Task("X", 4, F, A);
Task Q = new Task("Q", 2, A, X);
Task start = new Task("Start", 0, Q);
allTasks.add(end);
allTasks.add(F);
allTasks.add(A);
allTasks.add(X);
allTasks.add(Q);
allTasks.add(start);
System.out.println("Critical Path: "+Arrays.toString(criticalPath(allTasks)));
}
//A wrapper class to hold the tasks during the calculation
public static class Task{
//the actual cost of the task
public int cost;
//the cost of the task along the critical path
public int criticalCost;
//a name for the task for printing
public String name;
//the tasks on which this task is dependant
public HashSet<Task> dependencies = new HashSet<Task>();
public Task(String name, int cost, Task... dependencies) {
this.name = name;
this.cost = cost;
for(Task t : dependencies){
this.dependencies.add(t);
}
}
@Override
public String toString() {
return name+": "+criticalCost;
}
public boolean isDependent(Task t){
//is t a direct dependency?
if(dependencies.contains(t)){
return true;
}
//is t an indirect dependency
for(Task dep : dependencies){
if(dep.isDependent(t)){
return true;
}
}
return false;
}
}
public static Task[] criticalPath(Set<Task> tasks){
//tasks whose critical cost has been calculated
HashSet<Task> completed = new HashSet<Task>();
//tasks whose ciritcal cost needs to be calculated
HashSet<Task> remaining = new HashSet<Task>(tasks);
//Backflow algorithm
//while there are tasks whose critical cost isn't calculated.
while(!remaining.isEmpty()){
boolean progress = false;
//find a new task to calculate
for(Iterator<Task> it = remaining.iterator();it.hasNext();){
Task task = it.next();
if(completed.containsAll(task.dependencies)){
//all dependencies calculated, critical cost is max dependency
//critical cost, plus our cost
int critical = 0;
for(Task t : task.dependencies){
if(t.criticalCost > critical){
critical = t.criticalCost;
}
}
task.criticalCost = critical+task.cost;
//set task as calculated an remove
completed.add(task);
it.remove();
//note we are making progress
progress = true;
}
}
//If we haven't made any progress then a cycle must exist in
//the graph and we wont be able to calculate the critical path
if(!progress) throw new RuntimeException("Cyclic dependency, algorithm stopped!");
}
//get the tasks
Task[] ret = completed.toArray(new Task[0]);
//create a priority list
Arrays.sort(ret, new Comparator<Task>() {
@Override
public int compare(Task o1, Task o2) {
//sort by cost
int i= o2.criticalCost-o1.criticalCost;
if(i != 0)return i;
//using dependency as a tie breaker
//note if a is dependent on b then
//critical cost a must be >= critical cost of b
if(o1.isDependent(o2))return -1;
if(o2.isDependent(o1))return 1;
return 0;
}
});
return ret;
}
}