根据索引 position 连接一维和二维 arrays

Question

I am trying to concatenate a 1D into a 2D array. I'd like to avoid doing a loop as it's very computer intensive if my array lengths are greater than 1000.

I have tried vstack, stack and concatenate with no success.

import numpy as np

array_a = np.array([1,2,3])

array_b = np.array([[10, 11, 12], [20, 21, 22], [30, 31, 32]])

The expected output should be

array([[1, 10, 11, 12], [2, 20, 21, 22], [3, 30, 31, 32]])

Many thanks for your help!

Answer 1

Mykola showed the right way to do this, but I suspect you need a little help in understanding why. You tried several things without telling us what was wrong.

In [241]: array_a = np.array([1,2,3]) 
     ...: array_b = np.array([[10, 11, 12], [20, 21, 22], [30, 31, 32]])  

vstack runs:

In [242]: np.vstack((array_a, array_b))                                         
Out[242]: 
array([[ 1,  2,  3],
       [10, 11, 12],
       [20, 21, 22],
       [30, 31, 32]])

But the result is a vertical join, by rows, not columns. The v in vstack is supposed to remind us of that.

stack tries to join the arrays on a new axis, and requires that all input array have a matching shape:

In [243]: np.stack((array_a, array_b))                                          
...
ValueError: all input arrays must have the same shape

I suspect you tried this at random, without really reading the docs.

Both of these use concatenate , which is the basic joiner. But it's picky about dimensions:

In [244]: np.concatenate((array_a, array_b))                                    
...
ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 1 dimension(s) and the array at index 1 has 2 dimension(s)

You clearly realized that the number of dimensions didn't match.

You want to make a (3,4) array. One is (3,3), the other needs to be (3,1). And join axis needs to be 1

In [247]: np.concatenate((array_a[:,None], array_b), axis=1)                    
Out[247]: 
array([[ 1, 10, 11, 12],
       [ 2, 20, 21, 22],
       [ 3, 30, 31, 32]])

If we made a (1,3) array, and tried to join on the default 0 axis, we get the same thing as the vstack . In fact that's what vstack does:

In [248]: np.concatenate((array_a[None,:], array_b))                            
Out[248]: 
array([[ 1,  2,  3],
       [10, 11, 12],
       [20, 21, 22],
       [30, 31, 32]])

Another function is:

In [249]: np.column_stack((array_a, array_b))                                   
Out[249]: 
array([[ 1, 10, 11, 12],
       [ 2, 20, 21, 22],
       [ 3, 30, 31, 32]])

This does the same thing as [247].

Functions like vstack and column_stack are handy, but in long run it's better to understand how to use concatenate itself.

Answer 2

You want insert :

import numpy as np

array_a = np.array([1, 2, 3])

array_b = np.array([[10, 11, 12], [20, 21, 22], [30, 31, 32]])

result = np.insert(array_b, 0, array_a, axis=1)
print(result)

Output

[[ 1 10 11 12]
 [ 2 20 21 22]
 [ 3 30 31 32]]

Answer 3

You can reshape() the first array and then concatenate() both arrays:

np.concatenate([array_a.reshape(3, -1), array_b], axis=1)