Loc vs iloc in pandas: Key Differences and When to Use Each

loc selects data by label names of rows and columns, while iloc selects data by integer position indexes. Use loc when you know the row or column labels, and iloc when you want to select by numeric position.

Quick Comparison

Here is a quick side-by-side comparison of loc and iloc in pandas.

Feature	loc	iloc
Selection basis	Label names of rows/columns	Integer positions of rows/columns
Supports slicing	Yes, label-based inclusive slicing	Yes, position-based exclusive slicing
Accepts boolean arrays	Yes	Yes
Works with row and column labels	Yes	No, only positions
Allows mixed label/index selection	Yes, with labels	No, only integers
Error on missing labels	Raises KeyError	Raises IndexError if out of bounds

Key Differences

loc uses the actual labels of the DataFrame’s index and columns to select data. This means if your rows are labeled with dates, strings, or custom labels, loc lets you pick rows and columns by those exact names. It also includes the end label when slicing, so df.loc['a':'c'] includes rows labeled 'a', 'b', and 'c'.

On the other hand, iloc works purely with integer positions, like counting rows and columns from zero. It ignores the labels and selects rows or columns by their position number. When slicing with iloc, the end position is excluded, similar to Python’s usual slicing rules.

Because loc depends on labels, it raises a KeyError if you try to select a label that does not exist. iloc raises an IndexError if you try to select a position outside the DataFrame’s range. Both support boolean arrays for filtering, but loc is more intuitive when working with labeled data.

Code Comparison

Example using loc to select rows and columns by labels.

python

import pandas as pd

data = {'Name': ['Alice', 'Bob', 'Charlie', 'David'],
        'Age': [25, 30, 35, 40],
        'City': ['NY', 'LA', 'Chicago', 'Houston']}

index_labels = ['a', 'b', 'c', 'd']
df = pd.DataFrame(data, index=index_labels)

# Select rows 'b' to 'd' and columns 'Name' and 'City'
result = df.loc['b':'d', ['Name', 'City']]
print(result)

Output

     Name     City
b     Bob       LA
c  Charlie  Chicago
d   David  Houston

iloc Equivalent

Example using iloc to select rows and columns by integer positions matching the same data as above.

python

import pandas as pd

data = {'Name': ['Alice', 'Bob', 'Charlie', 'David'],
        'Age': [25, 30, 35, 40],
        'City': ['NY', 'LA', 'Chicago', 'Houston']}

index_labels = ['a', 'b', 'c', 'd']
df = pd.DataFrame(data, index=index_labels)

# Select rows 1 to 3 (positions) and columns 0 and 2 (positions)
result = df.iloc[1:4, [0, 2]]
print(result)

Output

     Name     City
b     Bob       LA
c  Charlie  Chicago
d   David  Houston

When to Use Which

Choose loc when you want to select data by meaningful row or column labels, especially if your DataFrame has custom or non-numeric indexes. It is clearer and less error-prone when working with labeled data.

Choose iloc when you want to select data by position, such as the first few rows or columns, or when you do not know the labels. It is useful for quick slicing by numeric index.

In summary, use loc for label-based selection and iloc for position-based selection.

Key Takeaways

loc selects data by labels; iloc selects by integer positions.

loc includes the end label in slices; iloc excludes the end position.

loc raises KeyError for missing labels; iloc raises IndexError for out-of-range positions.

Use loc when working with labeled indexes and iloc for numeric position-based selection.

Both support boolean indexing but differ in how they interpret the selection keys.