How to convert an object array to a normal array in python

Question

I have an object array which looks something like this

array([array([[2.4567]],dtype=object), array([[3.4567]],dtype=object), array([[4.4567]],dtype=object), array([[5.4567]],dtype=object) ... array([[6.4567]],dtype=object))

This is just an example, actual one is much bigger.

So, how do I convert this into a normal floating value numpy array.

Solution 1

Use numpy.concatenate:

>>> arr = array([array([[2.4567]],dtype=object),array([[3.4567]],dtype=object),array([[4.4567]],dtype=object),array([[5.4567]],dtype=object),array([[6.4567]], dtype=object)])
>>> np.concatenate(arr).astype(None)
array([[ 2.4567],
       [ 3.4567],
       [ 4.4567],
       [ 5.4567],
       [ 6.4567]])

Solution 2

You can use np.stack, works also for the multi dimensional case:

import numpy as np
from numpy import array

arr = array([array([[2.4567]],dtype=object),array([[3.4567]],dtype=object),array([[4.4567]],dtype=object),array([[5.4567]],dtype=object),array([[6.4567]],
dtype=object)])
np.stack(arr).astype(None)
array([[[2.4567]],
       [[3.4567]],
       [[4.4567]],
       [[5.4567]],
       [[6.4567]]])

Comments

[Shashank]

what if the above array I mentioned is just the first row of a multi dimensional array and there are several rows. "np.concatenate" joins all of them into a 1d array which I do not want. So how to concatenate them rowwise without using for loops?

[Darcy]

I have the same problem as @Shashank. The above method does not work for N x M arrays.

[Ashwini Chaudhary]

@Darcy then ask a new question instead of downvoting existing ones. o_O