numpy：将扫描数据的单个二维数组重塑为 3 个二维数组

Question

I have an array of the following form:

x_1    y_1    z_1,1
x_1    y_2    z_1,2
x_1    y_3    z_1,3
...    ...     ...
x_1    y_n    z_1,m
x_2    y_1    z_2,1
x_2    y_2    z_2,2
x_2    y_3    z_2,3
...    ...     ...
x_2    y_m    z_2,m
...    ...     ...
x_n    y_m    z_n,m

which is scan data, so x- and y- coordinates with a measured value (z) at that coordinate. I want to plot this as a contour plot and for this need arrays similar to those produced my np.meshgrid() . So I need three arrays of the following form (which are all of shape: m rows by n columns):

X = [x_1  x_2  x_3  ...  x_n
     x_1  x_2  x_3  ...  x_n
     ...                  .
          ...             .
               ...        .
     x_1  x_2  x_3  ...  x_n]


Y = [y_1  y_1  y_1  ...  y_1
     y_2  y_2  y_2  ...  y_2
     ...                  .
          ...             .
               ...        .
     y_m  y_m  y_m  ...  y_m]


Z = [z_1,1  z_2,1  z_3,1  ...  z_n,1
     z_1,2  z_2,2  z_3,2  ...  z_n,2
      ...                        .
             ...                 .
                    ...          .
     z_1,m  z_2,m  z_3,m  ...  z_n,m]

Caveats:

• n and m are not necesarily equal and can vary between scans.
• spacing in the x- and y-steps need not necesarily be constant

What is the most pythonic way of implementing this? Is there a simple way by employing reshaping? I could do it with a for loop, keeping track of the current x- and y- coordinate and moving to a new column/row if they change, to insert values into the arrays. But this seems very tedious and slow...

Answer 1

TL;DR
data.T.reshape(3,m,n).transpose((0,2,1))

---

1. Replicate your data structure

In [60]: m, n = 3, 5
...: x, y, z = (
...:     np.array(list(range(m))*n).reshape(n,m).T.flatten(),
...:      np.array(list(range(n))*m),
...:      np.arange(n*m))
...: a = np.array((x, y, z)).T
...: print(a)
[[ 0  0  0]
 [ 0  1  1]
 [ 0  2  2]
 [ 0  3  3]
 [ 0  4  4]
 [ 1  0  5]
 [ 1  1  6]
 [ 1  2  7]
 [ 1  3  8]
 [ 1  4  9]
 [ 2  0 10]
 [ 2  1 11]
 [ 2  2 12]
 [ 2  3 13]
 [ 2  4 14]]

2. Let's see what meshgrid would like to do

In [62]: np.meshgrid(range(m), range(n))
Out[62]: 
[array([[0, 1, 2],
        [0, 1, 2],
        [0, 1, 2],
        [0, 1, 2],
        [0, 1, 2]]),
 array([[0, 0, 0],
        [1, 1, 1],
        [2, 2, 2],
        [3, 3, 3],
        [4, 4, 4]])]

3. The solution is

In [63]: a.T.reshape(3,m,n).transpose((0,2,1))
Out[63]: 
array([[[ 0,  1,  2],
        [ 0,  1,  2],
        [ 0,  1,  2],
        [ 0,  1,  2],
        [ 0,  1,  2]],

       [[ 0,  0,  0],
        [ 1,  1,  1],
        [ 2,  2,  2],
        [ 3,  3,  3],
        [ 4,  4,  4]],

       [[ 0,  5, 10],
        [ 1,  6, 11],
        [ 2,  7, 12],
        [ 3,  8, 13],
        [ 4,  9, 14]]])