如何避免深度树递归中的堆栈溢出

Question

I've read many similar articles, I apologise if this has already been answered, but I'm still stuck.

I'm coding a function to populate a tree, each node having four branches, which will store the possible manipulation of states of the "eight tile puzzle" ie http://www.8puzzle.com/images/8_puzzle_start_state_a.png

The problem is I've reached, what I believe to be, stack overflow, due to the heavily recursive nature of the problem at hand.

Tail recursion seems like the solution, though I'm not sure if it is relevant/possible or how to implement it in this instance. Code follows:

void Tree::insert(string &initialState, const string &goalState, tree_node *&inNode){
cout<<"insert called"<<endl;
depth++;
cout<<depth<<endl;
string modState;
int zeroPos=0;

if(initialState==goalState){
    cout<<"*    *   *   GOAL    *   *   *"<<endl;
    getchar();
    exit(0);
}   

if(inNode==NULL){//is this the first node?

    inNode = new tree_node(initialState);       
    root=inNode;
    inNode->parent=NULL;
    insert(initialState, goalState, inNode);
}else{
    inNode->state = initialState;

    for(zeroPos=0;zeroPos<initialState.size();zeroPos++){//where is the empty tile?
        if(initialState[zeroPos]=='0'){
            break;
        }
    }
    //left
    if(zeroPos!=0 && zeroPos!=3 && zeroPos!=6){//can the empty tile move left?

        modState=initialState;
        modState[zeroPos]=modState[zeroPos-1];
        modState[zeroPos-1]='0';

        if(isOriginal(modState, inNode) ){//does this state already exist?

            cout <<"left  " << modState[0]<<modState[1]<<modState[2]<<endl;
            cout <<"left  " << modState[3]<<modState[4]<<modState[5]<<endl;
            cout <<"left  " << modState[6]<<modState[7]<<modState[8]<<endl;

            inNode->l = new tree_node(modState);    
            inNode->l->parent= inNode;
            if(inNode->l != NULL){
                insert(modState, goalState, inNode->l);
            }
        }
    }

    }
 }
}

Answer 1

This is a very generic answer, but have you tried to explicitly manage your stack through something like a heap allocated queue or a stack? Basically don't use actually function calls, just push and pull things off your own stack/queue. They cover non-recursive graph traversal algorithms here (both depth first, and breadth first search).

Answer 2

Optimize your code, meaning only keep those nodes you need. For example only keep leaf nodes (I mean the ones you haven't yet expaneded). And yes you can search the tree while building it, write a function to measure the distance between your current state and your goal state, its called heuristic. There's also a better algorithm that solves 8puzzle, it's called A*, I suggest you google it.