I have 2 arrays.
a = np.zeros(2)
b = np.ones(2)
I want to combine them so that its output is:
array([[0, 1],
[0, 1]])
If I had a = np.zeros((2,1)) and b = np.zeros((2,1)), this can done with np.hstack((a,b)) or np.append(a,b,1). But I don't.
Any thoughts?
You can use np.c_
>>> a = np.zeros(2); b = np.ones(2)
>>> np.c_[a, b]
array([[0., 1.],
[0., 1.]])
Alternatively:
>>> np.array((a, b), order='F').T
array([[0., 1.],
[0., 1.]])
Here order='F' makes sure that the final result is C-contiguous
You can always add a trailing dimension...
>>> np.hstack([a[:,None], b[:,None]])
array([[ 0., 1.],
[ 0., 1.]])
>>>
One way or other you need to add the extra dimension to the arrays.
In [2]: a = np.zeros(2)
...: b = np.ones(2)
...:
stack lets you choose the new dimension:
In [3]: np.stack((a,b),axis=1)
Out[3]:
array([[0., 1.],
[0., 1.]])
In [4]: np.stack((a,b),axis=0)
Out[4]:
array([[0., 0.],
[1., 1.]])
np.array also does it, like stack with axis 0 - transpose to get the columns:
In [5]: np.array((a,b)).T
Out[5]:
array([[0., 1.],
[0., 1.]])
column_stack is an older stack that add the right dimensions for this case as well.
In [6]: np.column_stack((a,b))
Out[6]:
array([[0., 1.],
[0., 1.]])
It's a good a idea to learn to do this with concatenate and your own addition of dimensions. And look at the code for the various stack functions to see how they do the same task.
You can also reshape them:
>>> a, b = np.zeros(2), np.ones(2)
>>> np.hstack([a.reshape(2,1), b.reshape(2,1)])
array([[ 0., 1.],
[ 0., 1.]])
