将多维元素附加到numpy数组中，而无需重塑

Question

I have a few simple questions I'm not able to find the answer to. They are both stated in the following example code. Thank you for any help!

import numpy as np 
#here are two arrays to join together 
a = np.array([1,2,3,4,5])
b = np.array([6,7,8,9,10])
#here comes the joining step I don't know how to do better

#QUESTION 1: How to form all permutations of two 1D arrays?

temp = np.array([]) #empty array to be filled with values 
for aa in a: 
    for bb in b: 
        temp = np.append(temp,[aa,bb]) #fill the array

#QUESTION 2: Why do I have to reshape? How can I avoid this? 

temp = temp.reshape((int(temp.size/2),2)) 

edit: made code more minimal

Answer 1

To answer your first question, you can use np.meshgrid to form those combinations between elements of the two input arrays and get to the final version of temp in a vectorized manner avoiding those loops, like so -

np.array(np.meshgrid(a,b)).transpose(2,1,0).reshape(-1,2)

As seen, we would still need a reshape if you intend to get a 2-column output array.

---

There are other ways we could construct the array with the meshed structure and thus avoid a reshape. One of those ways would be with np.column_stack , as shown below -

r,c = np.meshgrid(a,b)
temp = np.column_stack((r.ravel('F'), c.ravel('F')))

Answer 2

The proper way to build an array iteratively is with list append. np.append is poorly named, and often mis used.

In [274]: a = np.array([1,2,3,4,5])
     ...: b = np.array([6,7,8,9,10])
     ...: 
In [275]: temp = []
In [276]: for aa in a:
     ...:     for bb in b:
     ...:         temp.append([aa,bb])
     ...:         
In [277]: temp
Out[277]: 
[[1, 6],
 [1, 7],
 [1, 8],
 [1, 9],
 [1, 10],
 [2, 6],
  ....
 [5, 9],
 [5, 10]]
In [278]: np.array(temp).shape
Out[278]: (25, 2)

It's better to avoid loops at all, but if you must, use this list append approach.