基于大量的其他答案，我用c#想出了这个:

/// <param name="rotation">The number of rotations (if negative, the <see cref="Matrix{TValue}"/> is rotated counterclockwise; 
/// otherwise, it's rotated clockwise). A single (positive) rotation is equivalent to 90° or -270°; a single (negative) rotation is 
/// equivalent to -90° or 270°. Matrices may be rotated by 90°, 180°, or 270° only (or multiples thereof).</param>
/// <returns></returns>
public Matrix<TValue> Rotate(int rotation)
{
    var result = default(Matrix<TValue>);

    //This normalizes the requested rotation (for instance, if 10 is specified, the rotation is actually just +-2 or +-180°, but all 
    //correspond to the same rotation).
    var d = rotation.ToDouble() / 4d;
    d = d - (int)d;

    var degree = (d - 1d) * 4d;

    //This gets the type of rotation to make; there are a total of four unique rotations possible (0°, 90°, 180°, and 270°).
    //Each correspond to 0, 1, 2, and 3, respectively (or 0, -1, -2, and -3, if in the other direction). Since
    //1 is equivalent to -3 and so forth, we combine both cases into one. 
    switch (degree)
    {
        case -3:
        case +1:
            degree = 3;
            break;
        case -2:
        case +2:
            degree = 2;
            break;
        case -1:
        case +3:
            degree = 1;
            break;
        case -4:
        case  0:
        case +4:
            degree = 0;
            break;
    }
    switch (degree)
    {
        //The rotation is 0, +-180°
        case 0:
        case 2:
            result = new TValue[Rows, Columns];
            break;
        //The rotation is +-90°
        case 1:
        case 3:
            result = new TValue[Columns, Rows];
            break;
    }

    for (uint i = 0; i < Columns; ++i)
    {
        for (uint j = 0; j < Rows; ++j)
        {
            switch (degree)
            {
                //If rotation is 0°
                case 0:
                    result._values[j][i] = _values[j][i];
                    break;
                //If rotation is -90°
                case 1:
                    //Transpose, then reverse each column OR reverse each row, then transpose
                    result._values[i][j] = _values[j][Columns - i - 1];
                    break;
                //If rotation is +-180°
                case 2:
                    //Reverse each column, then reverse each row
                    result._values[(Rows - 1) - j][(Columns - 1) - i] = _values[j][i];
                    break;
                //If rotation is +90°
                case 3:
                    //Transpose, then reverse each row
                    result._values[i][j] = _values[Rows - j - 1][i];
                    break;
            }
        }
    }
    return result;
}

其中_values对应于由Matrix<TValue>定义的私有二维数组(形式为[][])。result = new TValue[Columns, Rows]可能通过隐式操作符重载并将二维数组转换为Matrix<TValue>。 Columns和Rows两个属性是公共属性，用于获取当前实例的列数和行数:

public uint Columns 
    => (uint)_values[0].Length;

public uint Rows 
    => (uint)_values.Length;

当然，假设您更喜欢使用无符号下标;-)

所有这些都允许您指定它应该旋转多少次，以及它应该向左旋转(如果小于零)还是向右旋转(如果大于零)。您可以改进此方法，以检查实际角度的旋转，但如果值不是90的倍数，则可能会抛出异常。有了这些输入，你可以相应地改变方法:

public Matrix<TValue> Rotate(int rotation)
{
    var _rotation = (double)rotation / 90d;

    if (_rotation - Math.Floor(_rotation) > 0)
    {
        throw new NotSupportedException("A matrix may only be rotated by multiples of 90.").
    }

    rotation = (int)_rotation;
    ...
}

Since a degree is more accurately expressed by double than int, but a matrix can only rotate in multiples of 90, it is far more intuitive to make the argument correspond to something else that can be accurately represented by the data structure used. int is perfect because it can tell you how many times to rotate it up to a certain unit (90) as well as the direction. double may very well be able to tell you that also, but it also includes values that aren't supported by this operation (which is inherently counter-intuitive).

当前所有的解决方案都有O(n^2)开销作为临时空间(这不包括那些肮脏的OOP骗子!)这里有一个内存占用为O(1)的解决方案，将矩阵原地右转90度。该死的延展性，这玩意儿跑得很快!

#include <algorithm>
#include <cstddef>

// Rotates an NxN matrix of type T 90 degrees to the right.
template <typename T, size_t N>
void rotate_matrix(T (&matrix)[N][N])
{
    for(size_t i = 0; i < N; ++i)
        for(size_t j = 0; j <= (N-i); ++j)
            std::swap(matrix[i][j], matrix[j][i]);
}

免责声明:我实际上并没有测试这个。让我们玩打虫游戏吧!

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!

一些人已经举了一些例子，其中涉及到创建一个新数组。

还有一些需要考虑的事情:

(a)不实际移动数据，只需以不同的方式遍历“旋转”的数组。

(b)就地轮换可能有点棘手。您需要一点空白的地方(大概相当于一行或一列的大小)。有一篇古老的ACM论文是关于进行原地转置的(http://doi.acm.org/10.1145/355719.355729)，但是他们的示例代码是令人讨厌的充满goto的FORTRAN。

附录:

http://doi.acm.org/10.1145/355611.355612是另一种更优越的就地转置算法。

#转置是Ruby的Array类的标准方法，因此:

% irb
irb(main):001:0> m = [[1, 2, 3, 4], [5, 6, 7, 8], [9, 0, 1, 2], [3, 4, 5, 6]]
=> [[1, 2, 3, 4], [5, 6, 7, 8], [9, 0, 1, 2], [3, 4, 5, 6]] 
irb(main):002:0> m.reverse.transpose
=> [[3, 9, 5, 1], [4, 0, 6, 2], [5, 1, 7, 3], [6, 2, 8, 4]]

实现是一个用c写的n^2转置函数，你可以在这里看到: http://www.ruby-doc.org/core-1.9.3/Array.html#method-i-transpose 通过选择“点击切换源”旁边的“转置”。

我记得比O(n^2)的解更好，但只适用于特殊构造的矩阵(如稀疏矩阵)

这是一个空间旋转方法，由java编写，只适用于正方形。对于非正方形的2d数组，无论如何都必须创建新数组。

private void rotateInSpace(int[][] arr) {
    int z = arr.length;
    for (int i = 0; i < z / 2; i++) {
        for (int j = 0; j < (z / 2 + z % 2); j++) {
            int x = i, y = j;
            int temp = arr[x][y];
            for (int k = 0; k < 4; k++) {
                int temptemp = arr[y][z - x - 1];
                arr[y][z - x - 1] = temp;
                temp = temptemp;

                int tempX = y;
                y = z - x - 1;
                x = tempX;
            }
        }
    }
}

通过创建新数组旋转任何大小的2d数组的代码:

private int[][] rotate(int[][] arr) {
    int width = arr[0].length;
    int depth = arr.length;
    int[][] re = new int[width][depth];
    for (int i = 0; i < depth; i++) {
        for (int j = 0; j < width; j++) {
            re[j][depth - i - 1] = arr[i][j];
        }
    }
    return re;
}