Overview

Pandas 2.3, released on February 15, 2025, enforces Copy-on-Write mode by default and improves performance.

Main Features

Copy-on-Write by default

Copy-on-Write (CoW) mode is now enabled by default. Modifications to a DataFrame copy no longer alter the original, eliminating SettingWithCopyWarning.

python 
import pandas as pd

df = pd.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})

# CoW on by default: no accidental modification
df2 = df[['a']]
df2['a'] = 99  # does NOT modify df
print(df)   # original unchanged
print(df2)  # only df2 is modified

Performance improvements

DataFrame operations are faster thanks to CoW avoiding unnecessary copies and internal optimizations.

python 
import pandas as pd
import numpy as np

# Slices no longer copy data
df = pd.DataFrame(np.random.randn(100000, 10))
subset = df[:50000]  # no copy (CoW)
print(f'Rows: {len(subset)}')

Sources