numpy subclassing

[jpecci]

I would like to ask you a quick question:
I am facing the opposite problem described here:
http://docs.scipy.org/doc/numpy/user...bclassing.html

In short, I want to create a class Dummy(np.ndarray) which returns a plain array whenever is sliced or viewed. I cannot figure out how.

I would really appreciate you help, here is a chunk of code, what is commented was an attempt which didn’t work.
Thanks a lot,
Jacopo

import numpy as np 
 
class Dummy(np.ndarray): 
 
    def __del__(self): 
        print "__del__" 
    def __new__(cls, array): 
        print "__new__" 
        obj=array.view(cls) 
        return obj 
    def __array_finalize__(self, obj): 
        print "__array_finalize__" 
        #self=self.view(np.ndarray) 
        #self=np.asarray(self) 
    def __repr__(self): 
        return "%s"%np.asarray(self) 
 
p=Dummy(np.ones(5)) 
print type(p)

[bvdet]

I don't know much about Numpy. If I understand you correctly, you want a slice of an instance of your subclass to return a list or tuple. Try defining a __getslice__() method.

import numpy as np
 
class N(np.ndarray):
 
    def __new__(cls, obj): 
        return obj.view(cls) 
 
    def __getslice__(self,i=None,j=None):
        return tuple(self.view()[i:j:])
 
p = N(np.ones(9))
print p
print p[1:3]
print p[:3]
print p[3:]
print p[1:8:3]

Output:

>>> [ 1.  1.  1.  1.  1.  1.  1.  1.  1.]
(1.0, 1.0)
(1.0, 1.0, 1.0)
(1.0, 1.0, 1.0, 1.0, 1.0, 1.0)
[ 1.  1.  1.]

It does not work for a slice with a stride. The __getslice__() method has been deprecated since version 2.0.

[jpecci]

thanks a lot. I think I will do something on these lines even if
I was hoping in something which didnt involve getslice() since it has been deprecated.
jacopo