Question

I have a PriorityQueue containing references to some objects. When I initially insert the elements into the priority queue the ordering is maintained by the data structure. Now after a remove operation I update some of the references which are being held by the priority queue. Ideally this requires a reheapify operation on the priority queue but as is obvious since I am modifying selected references externally a reheapify cannot be triggered. So what is the best way to ensure that I am able to get the advantage of a heap like fast extract max in the presence of modifications to arbitrary elements inside the queue? I see I need a better data structure?

To be more specific I need an implementation of something like a Fibonacci heap in Java.
http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm
Is that available?

Answer 1

You can use your own tree that allows duplicate elements instead of heap.
Easier though, you can use TreeMap<PriorityKey, LinkedHashSet<Value>>.
All operations are O(log priorityType):

• adding is get(priority).add(value), if get returns null, put(priority, new LinkedHashSet())
• removing an arbitrary element is similar, except needs to remove the priority out of the map if the Set is empty.
• poll() is

—

Entry<PriorityKey, LinkedHashSet<Value>> e = map.firstEntry(); 
if (e==null) return null; 
Iterator<Value> i = e.getValue().iterator(); 
Value v = i.next(); //must always have at least one element.
i.remove(); 
if (!i.hasNext()) map.remove(e.getKey());
return v;

Still if you need to change the priority you have to remove and add the element.