Recursion is a very elegant way to write some kind of algorithms. With few code lines we can implement complex solutions.
We know that, on each call on a recursive method, we add a new record with all local variables of that method in execution stack. Max stack’s size is wired to max recursion’s depth. If, for example, each call on a recursive method add a 20kb record to stack, and max recursion depth is 100, max stack’s size will be 20*100 = 2000kb.This is not a problem. But what about 10000 as max recursion depth? 20*10000 = 200.000kb = 200MB. This probably bring to a Stack Overflow exception. A so large recursion depth can be provided by a coding error but also by a large input.
Few days ago i wrote an algorithm for Flood Fiil .
This is the code
public void countClick(int row, int col,boolean deep){
if (row>=M || col>=N || col<0 || row<0)
return;
if (matrix[row][col]==0){
try {
if (!deep)count++;
matrix[row][col] = -1;
countClick(row,col+1,true);
countClick(row+1,col+1,true);
countClick(row+1,col,true);
countClick(row+1,col-1,true);
countClick(row,col-1,true);
countClick(row-1,col-1,true);
countClick(row-1,col,true);
countClick(row-1,col+1,true);
} catch (Exception e) {
System.out.println(row+" "+col);
e.printStackTrace();
}
}
if (!deep){
col++;
if (col >= N){
row++;
col=0;
}
countClick(row,col,false);
}
}
The algorithm walk through a binary matrix. On finding a 0 value, it recursively call the algorithm on the 8 contiguous cells (starting from the left one and proceeding in clockwise order) changing their value to 1 and counting all 0 contiguous area.
On big matrix input this algorithm throws a Stack Overflow Exception due to deep recursion.
JDK default stack size is 512 Kb. You can increase this value with -Xss VM argument. For example -Xss10m will increase stack size to 10 Mb.
You can ALWAYS write a recursive algorithm as a iterative one. In this example the key of recursion is (row,col) pair. To make this algorithm iterative i used a Stack object to store (row,col) pairs of visited matrix’s cells. On discovering of a 0 value i add (row,col) to stack,then, in clockwise order, i check if contiguous cells are also 0 value. If so i add this cell to stack and iterate process. If there aren’t 0 values i pop out the (row,col) pair, iterating this process until stack is not empty.
public void countClick(){
Stack<int[]> stack = new Stack<int[]>();
int[] coord = new int[2];
for (int row = 0; row < M ; row++){
for (int col = 0; col < N; col++){
if (matrix[row][col] == 0){
count++;
coord[0] = row;
coord[1] = col;
stack.add(coord);
matrix[row][col] = -1;
while (!stack.isEmpty()){
int tmpRow = stack.peek()[0];
int tmpCol = stack.peek()[1];
if (!(tmpRow>=M || tmpCol+1>=N || tmpCol+1<0 || tmpRow<0)&&(matrix[tmpRow][tmpCol+1]==0)){
stack.add(new int[]{tmpRow,tmpCol+1});
matrix[tmpRow][tmpCol+1] = -1;
}
else if (!(tmpRow+1>=M || tmpCol+1>=N || tmpCol+1<0 || tmpRow+1<0)&&(matrix[tmpRow+1][tmpCol+1]==0)){
stack.add(new int[]{tmpRow+1,tmpCol+1});
matrix[tmpRow+1][tmpCol+1] = -1;
}else if(!(tmpRow+1>=M || tmpCol>=N || tmpCol<0 || tmpRow+1<0)&& (matrix[tmpRow+1][tmpCol]==0)){
stack.add(new int[]{tmpRow+1,tmpCol});
matrix[tmpRow+1][tmpCol] = -1;
}else if (!(tmpRow+1>=M || tmpCol-1>=N || tmpCol-1<0 || tmpRow+1<0)&&(matrix[tmpRow+1][tmpCol-1]==0)){
stack.add(new int[]{tmpRow+1,tmpCol-1});
matrix[tmpRow+1][tmpCol-1] = -1;
}else if (!(tmpRow>=M || tmpCol-1>=N || tmpCol-1<0 || tmpRow<0)&&(matrix[tmpRow][tmpCol-1]==0)){
stack.add(new int[]{tmpRow,tmpCol-1});
matrix[tmpRow][tmpCol-1] = -1;
}else if (!(tmpRow-1>=M || tmpCol-1>=N || tmpCol-1<0 || tmpRow-1<0)&&(matrix[tmpRow-1][tmpCol-1]==0)){
stack.add(new int[]{tmpRow-1,tmpCol-1});
matrix[tmpRow-1][tmpCol-1] = -1;
}else if (!(tmpRow-1>=M || tmpCol>=N || tmpCol<0 || tmpRow-1<0)&&(matrix[tmpRow-1][tmpCol]==0)){
stack.add(new int[]{tmpRow-1,tmpCol});
matrix[tmpRow-1][tmpCol] = -1;
}else if (!(tmpRow-1>=M || tmpCol+1>=N || tmpCol+1<0 || tmpRow-1<0)&&(matrix[tmpRow-1][tmpCol+1]==0)){
stack.add(new int[]{tmpRow-1,tmpCol+1});
matrix[tmpRow-1][tmpCol+1] = -1;
}else
stack.pop();
}
}
}
}
}
Iterative algorithm is not elegant like recursive one but doesn’t throw Stack Overflow Exception.
