具有回溯的数独算法 - java

时间:2012-11-14 10:52:03

标签: java algorithm sudoku backtracking

所以我有这个大学任务解决数独...我读了算法X和跳舞算法,但他们没有帮助我。

我需要回溯。我用维基百科给出的地方用二维数字中的一些索引进行了硬编码(所以我确信它是可以解决的)。

我得到的代码如下:

public void solveSudoku(int row, int col)
   {
      // clears the temporary storage array that is use to check if there are
      // dublicates on the row/col
      for (int k = 0; k < 9; k++)
      {
         dublicates[k] = 0;
      }
      // checks if the index is free and changes the input number by looping
      // until suitable
      if (available(row, col))
      {
         for (int i = 1; i < 10; i++)
         {
            if (checkIfDublicates(i) == true)
            {
               board[row][col] = i;
               if (row == 8)
                  solveSudoku(0, col + 1);
               else if (col == 8)
                  solveSudoku(row + 1, 0);
               else
                  solveSudoku(row, col + 1);

               board[row][col] = 0;
            }
         }
      }
      // goes to the next row/col
      else
      {
         if (row == 8)
            solveSudoku(0, col + 1);
         else if (col == 8)
            solveSudoku(row + 1, 0);
         else
            solveSudoku(row, col + 1);
      }
   }

   /**
    * Checks if the spot on the certain row-col index is free of element
    * 
    * @param row
    * @param col
    * @return
    */
   private boolean available(int row, int col)
   {
      if (board[row][col] != 0)
         return false;
      else
         return true;
   }

   /**
    * Checks if the number given is not already used in this row/col
    * 
    * @param numberToCheck
    * @return
    */
   private boolean checkIfDublicates(int numberToCheck)
   {
      boolean temp = true;
      for (int i = 0; i < dublicates.length; i++)
      {
         if (numberToCheck == dublicates[i])
         {
            temp = false;
            return false;
         }
         else if (dublicates[i] == 0)
         {
            dublicates[i] = numberToCheck;
            temp = true;
            return true;
         }
      }
      return temp;
   }

我正在使用

上的StackOverflow
// goes to the next row/col
          else
          {
             if (row == 8)
                solveSudoku(0, col + 1);
             else if (col == 8)
                solveSudoku(row + 1, 0);
             else
                solveSudoku(row, col + 1);
          }

这意味着我必须在某个时候停止递归,但我无法弄清楚如何! 如果您在solve()函数中发现任何其他错误 - 请告诉我。因为我不确定我完全理解“回溯”的事情......

4 个答案:

答案 0 :(得分:3)

例如,如果跟踪当前的递归深度,可以停止递归

public void solveSudoku(int row, int col, int recursionDepth) {
    // get out of here if too much
    if (recursionDepth > 15) return;

    // regular code...
    // at some point call self with increased depth
    solveSudoku(0, col + 1, recursionDepth + 1);
}

  

如果你在solve()函数中发现任何其他错误 - 请告诉我。

代码过多:)

答案 1 :(得分:3)

这大致是我过去做过的方式。

Whenever all the definite moves have been taken and there is a choice of equally good next moves:
    copy your grid data structure and push it onto a stack.
    take the first candidate move and continue solving recursively
    Whereever you get stuck:
         pop the saved grid off the stack
         take the next candidate move.

答案 2 :(得分:1)

我以更简单的方式制作:

public void solve(int row, int col)
   {
      if (row > 8)
      {
         printBoard();
         System.out.println();
         return;
      }
      if (board[row][col] != 0)
      {
         if (col < 8)
            solve(row, col + 1);
         else
            solve(row + 1, 0);
      }
      else
      {
         for (int i = 0; i < 10; i++)
            if (checkRow(row, i) && checkCol(col, i))
                  //&& checkSquare(row, col, i))
            {
               board[row][col] = i;
               if (col < 8)
                  solve(row, col + 1);
               else
                  solve(row + 1, 0);
            }
         board[row][col] = 0;
      }
   }

   private boolean checkRow(int row, int numberToCheck)
   {
      for (int i = 0; i < 9; i++)
         if (board[row][i] == numberToCheck)
            return false;

      return true;
   }

   private boolean checkCol(int col, int numberToCheck)
   {
      for (int i = 0; i < 9; i++)
         if (board[i][col] == numberToCheck)
            return false;

      return true;
   }

答案 3 :(得分:1)

我不确定你为什么说跳舞链接和算法X没用 你的意思是你无法将Sudoku映射到算法X旨在解决的Exact Cover问题的实例吗?
或者说这对你需要的东西来说太复杂了?

如果是前者,您可能需要查看:A Sudoku Solver in Java implementing Knuth’s Dancing Links Algorithm。这很清楚,也解释了背后的原因。

N.B。算法X是一种回溯算法,因此,如果这是你唯一的要求,你绝对可以使用这种方法。

希望这可以提供帮助。