numpy array of array vs numpy array of list

Question

i have this two numpy arrays

arr1 = np.empty( (3,) , dtype=np.object)
arr1[0] = [1,2]
arr1[1] = [1,2,3]
arr1[2] = None

arr2 = np.empty( (3,) , dtype=np.object)
arr2[0] = np.array([1,2])
arr2[1] = np.array([1,2,3])
arr2[2] = None

why , when i'm checking if the internal array/list is None i'm getting different results both have the same shape

In [28]: arr1 != None
Out[28]: array([ True,  True,  False])

In [27]: arr2 != None
Out[27]: True

what is the way in numpy to check, that each element in arr2 is None?

as an aside, just use object, np.object is just object — juanpa.arrivillaga
– juanpa.arrivillaga, Commented Dec 26, 2021 at 7:43
This is happening because one of the elementwise comparisons failed, because using == on an np.ndarray will throw an error. — juanpa.arrivillaga
– juanpa.arrivillaga, Commented Dec 26, 2021 at 7:44
Also note, you probably shouldn't be using either of these arrays... an object dtype array is just a less performant list — juanpa.arrivillaga
– juanpa.arrivillaga, Commented Dec 26, 2021 at 7:57
Use a list comprehension: [a is not None for a in arr2]. Your arrays contain None in addition to lists and arrays. Expecting the same behavior for different classes of objects is unrealistic in python. Also most operations on object dtype arrays occur at list-comprehension speeds; so don't jump through hoops trying to use them. — hpaulj
– hpaulj, Commented Dec 26, 2021 at 17:06

Dennis · Accepted Answer · 2021-12-26 07:50:38Z

3

Note that the line arr2 != None raises the following DeprecationWarning on newer versions of Numpy:

Warning (from warnings module):
  File "<pyshell#19>", line 1
DeprecationWarning: elementwise comparison failed; this will raise an error in the future.

The reason is that my_array != my_object calls my_array[i] != my_object for each i, and expects to get back a boolean value at each element. Instead though, it gets back array([ False, False]) as the first element (not just a boolean).

So essentially, the difference is that list != None returns False, but numpy_array_of_ints != None returns an array. Elementwise comparisons should only return an array of booleans.

Edit: looking at the docs, the return value of elementwise comparisons are "Typically of type bool, unless dtype=object is passed." As such, we can do this:

>>> np.not_equal(arr1, None, dtype=object)
array([True, True, False], dtype=object)
>>> np.not_equal(arr2, None, dtype=object)
array([array([ True,  True]), array([ True,  True,  True]), False],
      dtype=object)

edited Dec 26, 2021 at 7:50

answered Dec 26, 2021 at 7:43

Dennis

2,3148 silver badges15 bronze badges

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

2 Comments

sami Over a year ago

10x @Dennis , starting to get it. but that still have different results np.not_equal(arr2, None, dtype=object) , result are for each int element while i'm trying to get the result of each arr[int] element like in the case of arr1

Dennis Over a year ago

As other users suggested, I would just do [x is not None for x in arr2] to get a list. In fact, I would just make arr2 be a list in the first place--NumPy arrays really only shine when storing scalar values, but standard python lists are better for storing python objects.

Collectives™ on Stack Overflow

numpy array of array vs numpy array of list

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