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

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!

其他回答

PHP:

<?php    
$a = array(array(1,2,3,4),array(5,6,7,8),array(9,0,1,2),array(3,4,5,6));
$b = array(); //result

while(count($a)>0)
{
    $b[count($a[0])-1][] = array_shift($a[0]);
    if (count($a[0])==0)
    {
         array_shift($a);
    }
}

从PHP5.6开始,数组转位可以通过一个狡猾的array_map()调用来执行。换句话说,列被转换为行。

代码:(演示)

$array = [
    [1, 2, 3, 4],
    [5, 6, 7, 8],
    [9, 0, 1, 2],
    [3, 4, 5, 6]
];
$transposed = array_map(null, ...$array);

美元转置:

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

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。

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;
}

下面是一个原地旋转的数组,而不是使用一个全新的数组来保存结果。我已经停止了数组的初始化和输出。这只适用于正方形数组,但它们可以是任何大小。内存开销等于数组中一个元素的大小,因此您可以对任意大的数组进行旋转。

int a[4][4];
int n = 4;
int tmp;
for (int i = 0; i < n / 2; i++)
{
    for (int j = i; j < n - i - 1; j++)
    {
        tmp             = a[i][j];
        a[i][j]         = a[j][n-i-1];
        a[j][n-i-1]     = a[n-i-1][n-j-1];
        a[n-i-1][n-j-1] = a[n-j-1][i];
        a[n-j-1][i]     = tmp;
    }
}

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。