TypeError: unhashable type: 'numpy.ndarray'

Question

From an array with three columns, I want to be able to just take a slice of data from all three columns where the values in the first column are equal to the values defined in above.

above = {1, 5, 10}
data = np.arange(9).reshape(-1, 3)
energies = np.hsplit(data, 3)[0]
slice = set(energies) & above

The above comes back with:

Traceback (most recent call last):
  File "<pyshell#27>", line 1, in <module>
    slice = set(energies) & above
TypeError: unhashable type: 'numpy.ndarray'

How do I resolve this error?

DSM · Accepted Answer · 2012-01-26 17:59:35Z

59

Your variable energies probably has the wrong shape:

>>> from numpy import array
>>> set([1,2,3]) & set(range(2, 10))
set([2, 3])
>>> set(array([1,2,3])) & set(range(2,10))
set([2, 3])
>>> set(array([[1,2,3],])) & set(range(2,10))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: unhashable type: 'numpy.ndarray'

And that's what happens if you read columnar data using your approach:

>>> data
array([[  1.,   2.,   3.],
       [  3.,   4.,   5.],
       [  5.,   6.,   7.],
       [  8.,   9.,  10.]])
>>> hsplit(data,3)[0]
array([[ 1.],
       [ 3.],
       [ 5.],
       [ 8.]])

Probably you can simply use

>>> data[:,0]
array([ 1.,  3.,  5.,  8.])

instead.

(P.S. Your code looks like it's undecided about whether it's data or elementdata. I've assumed it's simply a typo.)

answered Jan 26, 2012 at 17:59

DSM

355k67 gold badges606 silver badges504 bronze badges

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

2 Comments

user1171835 Over a year ago

Thank you very much for this answer. It worked perfectly and cleared things up for me. The data/elementdata was indeed a typo.

James McGuigan Over a year ago

The shape difference here is between a hashable 1D and unhashable 2D numpy array.

Sina Barghidarian · Accepted Answer · 2018-12-26 09:41:40Z

5

numpy.ndarray can contain any type of element, e.g. int, float, string etc. Check the type an do a conversion if neccessary.

edited Dec 26, 2018 at 9:41

Sina Barghidarian

4781 gold badge4 silver badges20 bronze badges

answered Sep 25, 2018 at 12:06

Joni

691 silver badge3 bronze badges

Comments

cottontail · Accepted Answer · 2024-04-28 20:30:47Z

If you got this error while trying to get unique values in a numpy ndarray using set, consider using np.unique instead. It flattens multi-dimensional arrays as well. For example:

arr1 = np.array([[1, 2], [3, 1]])
unique_arr1 = np.unique(arr1)        # array([1, 2, 3])

If you want to find the intersection between two arrays (which seems to be the motivation in the OP), then you can use np.intersect1d instead of set.intersection:

arr2 = np.array([1, 3, 4, 5])
np.intersect1d(unique_arr1, arr2)    # array([1, 3])

For union:

np.union1d(unique_arr1, arr2)        # array([1, 2, 3, 4, 5])

Another way this error (seemingly unrelated to hashing numpy ndarrays) may show up is if you use a numpy statistical function such as max, std etc. on a pandas Series and a numpy ndarray. For example:

a = pd.Series([1, 2])
b = np.array([3, 1])
np.max(a, b)                 # TypeError: unhashable type: 'numpy.ndarray'
np.max((a, b))               # OK

This is because the second argument of max is axis=, so passing two arrays is not the correct way to compute the maximum values of these objects; they must be wrapped in a tuple/list.

Collectives™ on Stack Overflow

TypeError: unhashable type: 'numpy.ndarray'

3 Answers 3

2 Comments

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

2 Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related