受Raymond Chen帖子的启发,假设你有一个4x4二维数组,写一个函数使它旋转90度。Raymond链接到伪代码中的解决方案,但我想看到一些真实的东西。

[1][2][3][4]
[5][6][7][8]
[9][0][1][2]
[3][4][5][6]

就变成:

[3][9][5][1]
[4][0][6][2]
[5][1][7][3]
[6][2][8][4]

更新:Nick的答案是最直接的,但是有没有比n²更好的方法呢?如果矩阵是10000x10000呢?


当前回答

哦,伙计。我一直认为这是一个“我很无聊,我能思考什么”的谜题。我想出了我的原地换位码,但到了这里发现你的和我的几乎一模一样……啊,好。这里是Ruby版本。

require 'pp'
n = 10
a = []
n.times { a << (1..n).to_a }

pp a

0.upto(n/2-1) do |i|
  i.upto(n-i-2) do |j|
    tmp             = a[i][j]
    a[i][j]         = a[n-j-1][i]
    a[n-j-1][i]     = a[n-i-1][n-j-1]
    a[n-i-1][n-j-1] = a[j][n-i-1]
    a[j][n-i-1]     = tmp
  end
end

pp a

其他回答

原地旋转不可能比O(n²)更快,原因是如果我们想旋转矩阵,我们必须至少一次触及所有n²元素,无论你实现什么算法。

这是我对矩阵90度旋转的尝试,这是c中的2步解决方案,首先转置矩阵,然后交换cols。

#define ROWS        5
#define COLS        5

void print_matrix_b(int B[][COLS], int rows, int cols) 
{
    for (int i = 0; i <= rows; i++) {
        for (int j = 0; j <=cols; j++) {
            printf("%d ", B[i][j]);
        }
        printf("\n");
    }
}

void swap_columns(int B[][COLS], int l, int r, int rows)
{
    int tmp;
    for (int i = 0; i <= rows; i++) {
        tmp = B[i][l];
        B[i][l] = B[i][r];
        B[i][r] = tmp;
    }
}


void matrix_2d_rotation(int B[][COLS], int rows, int cols)
{
    int tmp;
    // Transpose the matrix first
    for (int i = 0; i <= rows; i++) {
        for (int j = i; j <=cols; j++) {
            tmp = B[i][j];
            B[i][j] = B[j][i];
            B[j][i] = tmp;
        }
    }
    // Swap the first and last col and continue until
    // the middle.
    for (int i = 0; i < (cols / 2); i++)
        swap_columns(B, i, cols - i, rows);
}



int _tmain(int argc, _TCHAR* argv[])
{
    int B[ROWS][COLS] = { 
                  {1, 2, 3, 4, 5}, 
                      {6, 7, 8, 9, 10},
                          {11, 12, 13, 14, 15},
                          {16, 17, 18, 19, 20},
                          {21, 22, 23, 24, 25}
                        };

    matrix_2d_rotation(B, ROWS - 1, COLS - 1);

    print_matrix_b(B, ROWS - 1, COLS -1);
    return 0;
}

这是一个如今被高估的面试问题。

我的建议是:不要让面试官用他们关于解决这个问题的疯狂建议把你弄糊涂了。使用白板绘制输入数组的索引,然后绘制输出数组的索引。旋转前后的列分度示例如下:

30 --> 00
20 --> 01
10 --> 02
00 --> 03

31 --> 10
21 --> 11
11 --> 12
01 --> 13

注意旋转后的数字模式。

下面提供了一个简洁的Java解决方案。经过测试,它是有效的:

 Input:
    M A C P 
    B N L D 
    Y E T S 
    I W R Z 

    Output:
    I Y B M 
    W E N A 
    R T L C 
    Z S D P 

/**
 * (c) @author "G A N MOHIM"
 * Oct 3, 2015
 * RotateArrayNintyDegree.java
 */
package rotatearray;

public class RotateArrayNintyDegree {

    public char[][] rotateArrayNinetyDegree(char[][] input) {
        int k; // k is used to generate index for output array

        char[][] output = new char[input.length] [input[0].length];

        for (int i = 0; i < input.length; i++) {
            k = 0;
            for (int j = input.length-1; j >= 0; j--) {
                output[i][k] = input[j][i]; // note how i is used as column index, and j as row
                k++;
            }
        }

        return output;
    }

    public void printArray(char[][] charArray) {
        for (int i = 0; i < charArray.length; i++) {
            for (int j = 0; j < charArray[0].length; j++) {
                System.out.print(charArray[i][j] + " ");
            }
            System.out.println();
        }


    }

    public static void main(String[] args) {
        char[][] input = 
                { {'M', 'A', 'C', 'P'},
                  {'B', 'N', 'L', 'D'},
                  {'Y', 'E', 'T', 'S'},
                  {'I', 'W', 'R', 'Z'}
                };

        char[][] output = new char[input.length] [input[0].length];

        RotateArrayNintyDegree rotationObj = new RotateArrayNintyDegree();
        rotationObj.printArray(input);

        System.out.println("\n");
        output = rotationObj.rotateArrayNinetyDegree(input);
        rotationObj.printArray(output);

    }

}

在Eigen (c++)中:

Eigen::Matrix2d mat;
mat <<  1, 2,
        3, 4;
std::cout << mat << "\n\n";

Eigen::Matrix2d r_plus_90 = mat.transpose().rowwise().reverse();
std::cout << r_plus_90 << "\n\n";

Eigen::Matrix2d r_minus_90 = mat.transpose().colwise().reverse();
std::cout << r_minus_90 << "\n\n";

Eigen::Matrix2d r_180 = mat.colwise().reverse().rowwise().reverse(); // +180 same as -180
std::cout << r_180 << "\n\n";

输出:

1 2
3 4

3 1
4 2

2 4
1 3

4 3
2 1

正如我在上一篇文章中所说的,这里有一些c#代码,可以为任何大小的矩阵实现O(1)矩阵旋转。为了简洁性和可读性,没有错误检查或范围检查。代码:

static void Main (string [] args)
{
  int [,]
    //  create an arbitrary matrix
    m = {{0, 1}, {2, 3}, {4, 5}};

  Matrix
    //  create wrappers for the data
    m1 = new Matrix (m),
    m2 = new Matrix (m),
    m3 = new Matrix (m);

  //  rotate the matricies in various ways - all are O(1)
  m1.RotateClockwise90 ();
  m2.Rotate180 ();
  m3.RotateAnitclockwise90 ();

  //  output the result of transforms
  System.Diagnostics.Trace.WriteLine (m1.ToString ());
  System.Diagnostics.Trace.WriteLine (m2.ToString ());
  System.Diagnostics.Trace.WriteLine (m3.ToString ());
}

class Matrix
{
  enum Rotation
  {
    None,
    Clockwise90,
    Clockwise180,
    Clockwise270
  }

  public Matrix (int [,] matrix)
  {
    m_matrix = matrix;
    m_rotation = Rotation.None;
  }

  //  the transformation routines
  public void RotateClockwise90 ()
  {
    m_rotation = (Rotation) (((int) m_rotation + 1) & 3);
  }

  public void Rotate180 ()
  {
    m_rotation = (Rotation) (((int) m_rotation + 2) & 3);
  }

  public void RotateAnitclockwise90 ()
  {
    m_rotation = (Rotation) (((int) m_rotation + 3) & 3);
  }

  //  accessor property to make class look like a two dimensional array
  public int this [int row, int column]
  {
    get
    {
      int
        value = 0;

      switch (m_rotation)
      {
      case Rotation.None:
        value = m_matrix [row, column];
        break;

      case Rotation.Clockwise90:
        value = m_matrix [m_matrix.GetUpperBound (0) - column, row];
        break;

      case Rotation.Clockwise180:
        value = m_matrix [m_matrix.GetUpperBound (0) - row, m_matrix.GetUpperBound (1) - column];
        break;

      case Rotation.Clockwise270:
        value = m_matrix [column, m_matrix.GetUpperBound (1) - row];
        break;
      }

      return value;
    }

    set
    {
      switch (m_rotation)
      {
      case Rotation.None:
        m_matrix [row, column] = value;
        break;

      case Rotation.Clockwise90:
        m_matrix [m_matrix.GetUpperBound (0) - column, row] = value;
        break;

      case Rotation.Clockwise180:
        m_matrix [m_matrix.GetUpperBound (0) - row, m_matrix.GetUpperBound (1) - column] = value;
        break;

      case Rotation.Clockwise270:
        m_matrix [column, m_matrix.GetUpperBound (1) - row] = value;
        break;
      }
    }
  }

  //  creates a string with the matrix values
  public override string ToString ()
  {
    int
      num_rows = 0,
      num_columns = 0;

    switch (m_rotation)
    {
    case Rotation.None:
    case Rotation.Clockwise180:
      num_rows = m_matrix.GetUpperBound (0);
      num_columns = m_matrix.GetUpperBound (1);
      break;

    case Rotation.Clockwise90:
    case Rotation.Clockwise270:
      num_rows = m_matrix.GetUpperBound (1);
      num_columns = m_matrix.GetUpperBound (0);
      break;
    }

    StringBuilder
      output = new StringBuilder ();

    output.Append ("{");

    for (int row = 0 ; row <= num_rows ; ++row)
    {
      if (row != 0)
      {
        output.Append (", ");
      }

      output.Append ("{");

      for (int column = 0 ; column <= num_columns ; ++column)
      {
        if (column != 0)
        {
          output.Append (", ");
        }

        output.Append (this [row, column].ToString ());
      }

      output.Append ("}");
    }

    output.Append ("}");

    return output.ToString ();
  }

  int [,]
    //  the original matrix
    m_matrix;

  Rotation
    //  the current view of the matrix
    m_rotation;
}

好的,我把手举起来,当旋转时,它实际上不会对原始数组做任何修改。但是,在面向对象系统中,只要对象看起来像是被旋转到类的客户端,这就无关紧要了。目前,Matrix类使用对原始数组数据的引用,因此改变m1的任何值也将改变m2和m3。对构造函数稍加更改,创建一个新数组并将值复制到该数组中,就可以将其整理出来。