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

private static int[][] rotate(int[][] matrix, int n) {
    int[][] rotated = new int[n][n];
    for (int i = 0; i < n; i++) {
        for (int j = 0; j < n; j++) {
            rotated[i][j] = matrix[n-j-1][i];
        }
    }
    return rotated;
}

其他回答

下面是PHP的递归方法:

$m = array();
            $m[0] = array('a', 'b', 'c');
            $m[1] = array('d', 'e', 'f');
            $m[2] = array('g', 'h', 'i');
            $newMatrix = array();

            function rotateMatrix($m, $i = 0, &$newMatrix)
            {
                foreach ($m as $chunk) {
                    $newChunk[] = $chunk[$i];
                }
                $newMatrix[] = array_reverse($newChunk);
                $i++;

                if ($i < count($m)) {
                    rotateMatrix($m, $i, $newMatrix);
                }
            }

            rotateMatrix($m, 0, $newMatrix);
            echo '<pre>';
            var_dump($newMatrix);
            echo '<pre>';

Nick的答案也适用于NxM阵列,只需要做一点修改(与NxN相反)。

string[,] orig = new string[n, m];
string[,] rot = new string[m, n];

...

for ( int i=0; i < n; i++ )
  for ( int j=0; j < m; j++ )
    rot[j, n - i - 1] = orig[i, j];

考虑这个问题的一种方法是将轴(0,0)的中心从左上角移动到右上角。你只是简单地从一个转置到另一个。

虽然旋转数据可能是必要的(也许是为了更新物理存储的表示),但在数组访问上添加一层间接层(也许是一个接口)会变得更简单,可能更性能:

interface IReadableMatrix
{
    int GetValue(int x, int y);
}

如果你的矩阵已经实现了这个接口,那么它可以通过这样一个装饰器类来旋转:

class RotatedMatrix : IReadableMatrix
{
    private readonly IReadableMatrix _baseMatrix;

    public RotatedMatrix(IReadableMatrix baseMatrix)
    {
        _baseMatrix = baseMatrix;
    }

    int GetValue(int x, int y)
    {
        // transpose x and y dimensions
        return _baseMatrix(y, x);
    }
}

旋转+90/-90/180度,水平/垂直翻转和缩放都可以以这种方式实现。

Performance would need to be measured in your specific scenario. However the O(n^2) operation has now been replaced with an O(1) call. It's a virtual method call which is slower than direct array access, so it depends upon how frequently the rotated array is used after rotation. If it's used once, then this approach would definitely win. If it's rotated then used in a long-running system for days, then in-place rotation might perform better. It also depends whether you can accept the up-front cost.

与所有性能问题一样,测量,测量,测量!

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

这是我的实现,在C, O(1)内存复杂度,原地旋转,顺时针90度:

#include <stdio.h>

#define M_SIZE 5

static void initMatrix();
static void printMatrix();
static void rotateMatrix();

static int m[M_SIZE][M_SIZE];

int main(void){
    initMatrix();
    printMatrix();
    rotateMatrix();
    printMatrix();

    return 0;
}

static void initMatrix(){
    int i, j;

    for(i = 0; i < M_SIZE; i++){
        for(j = 0; j < M_SIZE; j++){
            m[i][j] = M_SIZE*i + j + 1;
        }
    }
}

static void printMatrix(){
    int i, j;

    printf("Matrix\n");
    for(i = 0; i < M_SIZE; i++){
        for(j = 0; j < M_SIZE; j++){
            printf("%02d ", m[i][j]);
        }
        printf("\n");
    }
    printf("\n");
}

static void rotateMatrix(){
    int r, c;

    for(r = 0; r < M_SIZE/2; r++){
        for(c = r; c < M_SIZE - r - 1; c++){
            int tmp = m[r][c];

            m[r][c] = m[M_SIZE - c - 1][r];
            m[M_SIZE - c - 1][r] = m[M_SIZE - r - 1][M_SIZE - c - 1];
            m[M_SIZE - r - 1][M_SIZE - c - 1] = m[c][M_SIZE - r - 1];
            m[c][M_SIZE - r - 1] = tmp;
        }
    }
}