How to Solve IndexError: Arrays Used as Indices Must Be of Integer (Or Boolean) Type

When working with Numpy arrays in Python, you might experience different error messages dealing with Index or Type issues. In these many error types, IndexError: arrays used as indices must be of integer (or boolean) type can be tricky.

When we face IndexError error messages, we use the wrong Type. In this case, we were supposed to use Integer or Boolean, but the array index receives another data type (string or float).

In this article, we will explain how to deal with IndexError: arrays used as indices must be of integer (or boolean) type error messages when working with numbers in Numpy.

Use astype() to Solve IndexError: arrays used as indices must be of integer (or boolean) type in Numpy

Numpy only works with two types, Integer or Boolean. Therefore, if there is a type it doesn’t understand, it will throw an error.

Let’s recreate the error message to understand this error message better. To recreate the error message, we need to generate two Numpy arrays, index and array, extract values from index, and use the extracted values to access the values in array.

For the extracted values, we will use the first column values.

import numpy as np

index = np.array([[0, 1, 2.1], [1, 2, 3.4]])
array = np.array([[1, 2, 3], [4, 5, 6]])

indices = index[:, 0]
print(array[indices])

Output:

Traceback (most recent call last):
  File "temp.py", line 7, in <module>
    print(array[indices])
IndexError: arrays used as indices must be of integer (or boolean) type

From the IndexError: arrays used as indices must be of integer (or boolean) type error message, we know the problem stems from the print(array[indices]) section.

Since we know that it is syntactically correct, we know that the issue we are looking for is present in what we are parsing to the array binding. That brings us to the indices binding.

From what we know from the error message, the element in the indices binding might not be integer or Boolean. The dtype property is useful to check the type of elements within indices.

print(indices.dtype)

Output:

float64

Now, that confirms the cause of the issue we are facing. The values we pass to the indices of the array binding are float64 instead of Boolean.

To solve this, we need to convert the values in indices to Integer or Boolean. It makes more sense to convert them to Integer.

Converting them to Boolean might be useful another time.

The astype() method helps modify the dtype property of a Numpy array. To modify the dtype of the indices binding, we can use the below.

indices = index[:, 0].astype(int)

We get the below if we check for the dtype property using the indices.dtype expression.

int32

Now, our code becomes:

import numpy as np

index = np.array([[0, 1, 2.1], [1, 2, 3.4]])
array = np.array([[1, 2, 3], [4, 5, 6]])

indices = index[:, 0].astype(int)
print(array[indices])

Output:

[[1 2 3]
 [4 5 6]]

We could have converted the values of indices to Boolean. Let’s experiment with that.

To do so, we have a Numpy array with two Booleans.

indices = index[:, 0].astype(bool)
print(indices)

Output:

[False  True]

The values of the indices binding were [0. 1.], and when converting 0 to Boolean, it gives False, and any other number gives True. Let’s run everything together.

import numpy as np

index = np.array([[0, 1, 2.1], [1, 2, 3.4]])
array = np.array([[1, 3, 5], [7, 9, 11]])

indices = index[:, 0].astype(bool)
print(array[indices])

Output:

[[ 7  9 11]]

That’s because it processes only the True value.

Therefore, when you face an IndexError: arrays used as indices must be of integer (or boolean) type error message, know there is a wrong dtype somewhere. Trace your code, and convert the necessary values.