How to Use np.unique in NumPy for Unique Elements

Use np.unique to find the unique elements of a NumPy array. It returns the sorted unique values and can also return the indices or counts of these unique elements.

Syntax

The basic syntax of np.unique is:

• np.unique(ar, return_index=False, return_inverse=False, return_counts=False, axis=None)

Where:

• ar: Input array to find unique elements from.
• return_index: If True, returns the indices of the input array that give the unique values.
• return_inverse: If True, returns the indices to reconstruct the original array from the unique array.
• return_counts: If True, returns the count of each unique element.
• axis: If specified, finds unique rows or columns along this axis.

python

np.unique(ar, return_index=False, return_inverse=False, return_counts=False, axis=None)

Example

This example shows how to find unique elements in a 1D array and get their counts.

python

import numpy as np

arr = np.array([2, 3, 2, 5, 3, 7, 5, 7, 7])
unique_values, counts = np.unique(arr, return_counts=True)
print("Unique values:", unique_values)
print("Counts:", counts)

Output

Unique values: [2 3 5 7]
Counts: [2 2 2 3]

Common Pitfalls

Common mistakes when using np.unique include:

• Not setting return_counts=True when counts are needed.
• Forgetting that the output is sorted, which may affect order-sensitive tasks.
• Using np.unique on multi-dimensional arrays without specifying axis, which flattens the array by default.

python

import numpy as np

arr = np.array([[1, 2], [1, 2]])
# Wrong: unique without axis flattens the array
unique_wrong = np.unique(arr)
print("Unique without axis:", unique_wrong)

# Right: unique with axis=0 finds unique rows
unique_right = np.unique(arr, axis=0)
print("Unique rows with axis=0:", unique_right)

Output

Unique without axis: [1 2]
Unique rows with axis=0: [[1 2]]

Quick Reference

Parameter	Description
ar	Input array to find unique elements
return_index	Return indices of unique elements in original array (default False)
return_inverse	Return indices to reconstruct original array from unique array (default False)
return_counts	Return counts of each unique element (default False)
axis	Axis to find unique elements along (default None, flattens array)

Key Takeaways

Use np.unique to get sorted unique elements from a NumPy array.

Set return_counts=True to get how many times each unique element appears.

Specify axis to find unique rows or columns in multi-dimensional arrays.

Remember np.unique returns sorted results, which may reorder elements.

Use return_index or return_inverse to map between original and unique arrays.