initialising class with numpy.ndarray attribute

Question

I'm trying to define a wrapper class having an attribute of type numpy.ndarray. That attribute must be initialised by calling __init__().

The script runs as expected for 1D arrays. However, in the case of multi-dimensionnal arrays, python returns the following error : only length-1 arrays can be converted to Python scalars

import numpy as np

class myArr(np.ndarray):
    def __init__(self,Arr):
        self.Arr = Arr

npArr = np.zeros((3)) # works
#npArr = np.zeros((3,5)) # does not work
print npArr

wrappedArr = myArr(npArr)
print wrappedArr.Arr

What is happening here ?

python 2.7.6, numpy 1.8.2

agmangas · Accepted Answer · 2015-08-27 07:48:33Z

1

If you just want to have an attribute of type ndarray is there any specific reason you inherit from it?

I'd say that by subclassing ndarray and overriding __init__ you're messing with numpy's initialisation process, thus generating the error you're seeing.

See the numpy docs about Subclassing ndarray for more info.

Inheriting from object (converting MyArr to a regular new-style class) solves your problem:

import numpy as np

class MyArr(object):
    def __init__(self, arr):
        self.arr = arr

np_arr_1 = np.zeros((3))
np_arr_2 = np.zeros((3, 5))

wrapped_arr_1 = MyArr(np_arr_1)
wrapped_arr_2 = MyArr(np_arr_2)

print wrapped_arr_1.arr
print wrapped_arr_2.arr

answered Aug 27, 2015 at 7:48

agmangas

6695 silver badges9 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Gael Lorieul Over a year ago

thanks for your answer ! The reason why I am implementing this wrapper class is that I have a code where numpy.roll() is very often called, usually multiple times in a single code line. In order to improve code readability, I would like to define a __call__() method as a replacement. It would be defined something like __call__(i,j): numpy.roll(self,[i,j]). Do you think that's the suitable approach ?

agmangas Over a year ago

I don't think you need to go through the hassle of subclassing ndarray, just call numpy.call using self.arr instead of self

Collectives™ on Stack Overflow

initialising class with numpy.ndarray attribute

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related