Implementation:Scikit learn Scikit learn SafeIndexing

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Machine Learning, Data Indexing
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete utility module for safe data indexing across array types provided by scikit-learn.

Description

The _indexing module provides functions for consistently indexing arrays, sparse matrices, pandas DataFrames, and Polars DataFrames. It handles the differences between indexing APIs across these data containers, supporting boolean masks, integer indices, and slices. The module also provides safe train/test splitting utilities.

Usage

Use these utilities when you need to index into data containers in a type-agnostic way, such as during cross-validation splitting, stratified sampling, or any operation that needs to work with multiple data container types.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/utils/_indexing.py

Signature

def _array_indexing(array, key, key_dtype, axis):
    ...

def _pandas_indexing(X, key, key_dtype, axis):
    ...

def _safe_indexing(X, indices, *, axis=0):
    ...

def _safe_assign(X, values, *, row_indexer=None, column_indexer=None):
    ...

Import

from sklearn.utils import _safe_indexing
from sklearn.utils._indexing import _safe_assign

I/O Contract

Inputs

Name	Type	Required	Description
X	array-like, sparse matrix, or DataFrame	Yes	Data to be indexed
indices	array-like, slice, or int	Yes	Indices to select from the data
axis	int	No	Axis along which to index (0 for rows, 1 for columns)
key_dtype	str	No	Type of key: "int", "bool", or "str"

Outputs

Name	Type	Description
result	same as input	Subset of the input data selected by the indices

Usage Examples

Basic Usage

import numpy as np
from sklearn.utils import _safe_indexing

X = np.array([[1, 2], [3, 4], [5, 6]])
indices = [0, 2]
result = _safe_indexing(X, indices)
print(result)
# [[1 2]
#  [5 6]]

Related Pages

• Principle:Scikit_learn_Scikit_learn_Feature_Transformation