二维NumPy数组的蛇遍历

时间:2019-04-14 15:42:05

标签: python numpy multidimensional-array traversal numpy-ndarray

我有以下2D数组:

In [173]: arr
Out[173]: 
array([[ 1,  2,  3,  4],   # -> -> -> ->
       [ 5,  6,  7,  8],   # <- <- <- <-
       [ 9, 10, 11, 12],   # -> -> -> ->
       [13, 14, 15, 16],   # <- <- <- <-
       [17, 18, 19, 20]])  # -> -> -> ->

我想遍历snake-like pattern中的数组,从左上角的元素开始,到右下角的元素结束。

到目前为止,我有这种有趣的解决方法:

In [187]: np.hstack((arr[0], arr[1][::-1], arr[2], arr[3][::-1], arr[4]))
Out[187]: 
array([ 1,  2,  3,  4,  8,  7,  6,  5,  9, 10, 11, 12, 16, 15, 14, 13, 17,
       18, 19, 20])

我们如何才能以最小的努力做到这一点,而又无需循环且无需过多的硬编码?

1 个答案:

答案 0 :(得分:4)

一种方法是从输入的副本开始,然后再从输入的对应行的列翻转版本开始替换第二行,然后使用步长切片对所有偶数行执行此操作。最后,对于所需的拼合版本,最后需要使用ravel()

因此,实现看起来像这样-

out = arr.copy()
out[1::2] = arr[1::2,::-1]
out = out.ravel()

另一种紧凑的方法是使用np.where在col-flipped和非fllip版本之间进行选择,从而实现我们所需的输出-

np.where(np.arange(len(arr))[:,None]%2,arr[:,::-1],arr).ravel()

给定样本的解释-

# Array to be used for the chosing. 1s would be True ones and 0s are False
In [72]: np.arange(len(arr))[:,None]%2
Out[72]: 
array([[0],
       [1],
       [0],
       [1],
       [0]])

# Use np.where to choose. So, arr[:,::-1] must be the first data, as
# that's one to be put on even rows and arr would be the second one.
In [73]: np.where(np.arange(len(arr))[:,None]%2,arr[:,::-1],arr)
Out[73]: 
array([[ 1,  2,  3,  4],
       [ 8,  7,  6,  5],
       [ 9, 10, 11, 12],
       [16, 15, 14, 13],
       [17, 18, 19, 20]])

# Finally flatten
In [74]: np.where(np.arange(len(arr))[:,None]%2,arr[:,::-1],arr).ravel()
Out[74]: 
array([ 1,  2,  3,  4,  8,  7,  6,  5,  9, 10, 11, 12, 16, 15, 14, 13, 17,
       18, 19, 20])