I'm not sure if this has been asked before but I couldn't find the solution to this seemingly simple problem. I have two arrays like so:
x = np.array([[1,2,3],[4,5,6]])
y = np.array([[1,2,3],[7,8,9]])
And I want to create a merged matrix like:
merged = [[(1,1),(2,2),(3,3)],[(4,7),(5,8),(6,9)]]
What is the fastest way to achieve this? I'm using python 2.
stack is a newish, handy way of creating an array like this. It's like np.array but lets us specify the new axis:
In [117]: x = np.array([[1,2,3],[4,5,6]])
...: y = np.array([[1,2,3],[7,8,9]])
...:
In [118]: x.shape
Out[118]: (2, 3)
In [119]: np.stack((x,y),axis=2)
Out[119]:
array([[[1, 1],
[2, 2],
[3, 3]],
[[4, 7],
[5, 8],
[6, 9]]])
Without stack you can combine the arrays on a new first axis, and then change the order of the axes:
In [120]: np.array((x,y))
Out[120]:
array([[[1, 2, 3],
[4, 5, 6]],
[[1, 2, 3],
[7, 8, 9]]])
In [121]: np.array((x,y)).transpose(1,2,0)
Out[121]:
array([[[1, 1],
[2, 2],
[3, 3]],
[[4, 7],
[5, 8],
[6, 9]]])
Note that a 3d array like this is not, technically, a 2d array of tuples. But for most purposes it is better than that.
stack (and dstack) expand the dimensions to the input arrays, and do a concatenate on that new dimension. In other words, they do:
In [123]: np.concatenate((x[...,None],y[...,None]),axis=-1)
Out[123]:
array([[[1, 1],
[2, 2],
[3, 3]],
[[4, 7],
[5, 8],
[6, 9]]])
np.stack([x, y], axis=-1)(cant check now) will give you a 3D array of the shape that you want. You could then use.tolist()if you really need a list.tupleor does alistwork as well?