Difficulties in converting an recursive algorithm into an iterative one

Question

I've been trying to implement a recursive backtracking maze generation algorithm in javascript. These were done after reading a great series of posts on the topic here

While the recursive version of the algorithm was a no brainer, the iterative equivalent has got me stumped.

I thought I understood the concept, but my implementation clearly produces incorrect results. I've been trying to pin down a bug that might be causing it, but I am beginning to believe that my problems are being caused by a failure in logic, but of course I am not seeing where.

My understanding of the iterative algorithm is as follows:

• A stack is created holding representations of cell states.

• Each representation holds the coordinates of that particular cell, and a list of directions to access adjacent cells.

• While the stack isn't empty iterate through the directions on the top of the stack, testing adjacent cells.

• If a valid cell is found place it at the top of the stack and continue with that cell.

Here is my recursive implementation ( note: keydown to step forward ): http://jsbin.com/urilan/14

And here is my iterative implementation ( once again, keydown to step forward ): http://jsbin.com/eyosij/2

Thanks for the help.

edit: I apologize if my question wasn't clear. I will try to further explain my problem.

When running the iterative solution various unexpected behaviors occur. First and foremost, the algorithm doesn't exhaust all available options before backtracking. Rather, it appears to be selecting cells at a random when there is one valid cell left. Overall however, the movement doesn't appear to be random.

var dirs = [ 'N', 'W', 'E', 'S' ];
var XD = { 'N': 0, 'S':0, 'E':1, 'W':-1 };
var YD = { 'N':-1, 'S':1, 'E':0, 'W': 0 };


function genMaze(){

var dirtemp = dirs.slice().slice();    //copies 'dirs' so its not overwritten or altered
var path = [];                         // stores path traveled.

var stack = [[0,0, shuffle(dirtemp)]]; //Stack of instances. Each subarray in 'stacks' represents a cell
                                       //and its current state. That is, its coordinates, and which adjacent cells have been
                                       //checked. Each time it checks an adjacent cell a direction value is popped from 
                                       //from the list

while ( stack.length > 0 ) {

  var current = stack[stack.length-1]; // With each iteration focus is to be placed on the newest cell.

  var x = current[0], y = current[1], d = current[2];
  var sLen = stack.length;             // For testing whether there is a newer cell in the stack than the current.
  path.push([x,y]);                    // Store current coordinates in the path

  while ( d.length > 0 ) {
    if( stack.length != sLen ){ break;}// If there is a newer cell in stack, break and then continue with that cell

    else {
      var cd = d.pop();
      var nx = x + XD[ cd ];
      var ny = y + YD[ cd ];

      if ( nx >= 0 && ny >= 0  && nx < w && ny < h && !cells[nx][ny] ){

        dtemp = dirs.slice().slice();
        cells[nx][ny] = 1;
        stack.push( [ nx, ny, shuffle(dtemp) ] ); //add new cell to the stack with new list of directions.
                                                  // from here the code should break from the loop and start again with this latest addition being considered.


      }
    }

  }

  if (current[2].length === 0){stack.pop(); } //if all available directions have been tested, remove from stack


}
return path;
}

I hope that helps clear up the question for you. If it is still missing any substance please let me know.

Thanks again.

Answer 1

Does this little code could also help ?

/**
Examples
var sum = tco(function(x, y) {
  return y > 0 ? sum(x + 1, y - 1) :
         y < 0 ? sum(x - 1, y + 1) :
        x
})
sum(20, 100000) // => 100020
**/

function tco(f) {
  var value, active = false, accumulated = []
  return function accumulator() {
    accumulated.push(arguments)
    if (!active) {
      active = true
      while (accumulated.length) value = f.apply(this, accumulated.shift())
      active = false
      return value
    }
  }
}

Credits, explanations ans more infos are on github https://gist.github.com/1697037

Is has the benefit to not modifying your code, so it could be applied in other situations too. Hope that helps :)

Answer 2

I'm not very good in javascript, but I try to implement your recursive code to iterative. You need to store For index on stack also. So code look like:

function genMaze(cx,cy) {

    var dirtemp = dirs;    //copies 'dirs' so its not overwritten
    var path = [];                         // stores path traveled.    
    var stack = [[cx, cy, shuffle(dirtemp), 0]];  // we also need to store `for` indexer

    while (stack.length > 0) {

        var current = stack[stack.length - 1]; // With each iteration focus is to be placed on the newest cell.

        var x = current[0], y = current[1], d = current[2], i = current[3];
        if (i > d.length) {
            stack.pop();
            continue;
        }
        stack[stack.length - 1][3] = i + 1; // for next iteration

        path.push([x, y]);    // Store current coordinates in the path
        cells[x][y] = 1;

        var cd = d[i];
        var nx = x + XD[cd];
        var ny = y + YD[cd];

        if (nx >= 0 && ny >= 0 && nx < w && ny < h && !cells[nx][ny]) {

            dtemp = dirs;
            stack.push([nx, ny, shuffle(dtemp), 0]);
        }
    }
    return path;
  }