SettingWithCopyWarning: A value is trying to be set on a copy of a slice in Pandas

Verified FixPython 3.10+Pandas 2.0+Silo: pandas

Quick Fix / Solution Rapide

Use df.loc[row_mask, 'col_name'] = value instead of chained indexing df[row_mask]['col_name'] = value to ensure modifications apply to the original DataFrame.

Root Cause Analysis

This error occurs when Python tries to assign values to a DataFrame slice created via chained indexing (such as df[mask]['col'] = value), but pandas cannot determine whether the intermediate slice is a view or an independent copy.

Cause 1: Chained Indexing (getitem followed by setitem)

Chained assignment df[mask]['status'] = 'active' breaks into two separate operations: df[mask] returns an intermediate DataFrame slice, and ['status'] = ... attempts to write to that intermediate object. If the intermediate slice is a copy, modifications are discarded, leaving the original DataFrame unchanged.

Cause 2: Copy-on-Write (CoW) in Pandas 2.0+ and 2.2+

Starting in Pandas 2.0 and made standard under Copy-on-Write (CoW), slicing operations produce views with copy-on-write semantics. Under CoW mode, chained assignment produces a hard error or is prevented to avoid silent data corruption.

Cause 3: Assigning to Filtered Subsets Without .copy()

Creating a subset DataFrame with subset = df[df['score'] > 80] and later modifying subset['rank'] = 1 triggers SettingWithCopyWarning if subset is linked to the parent DataFrame.

Reproduction Code (MCVE)

Example: Bug Reproduction

import pandas as pd
import warnings
from pandas.errors import SettingWithCopyWarning
warnings.simplefilter('error', SettingWithCopyWarning)
pd.options.mode.copy_on_write = False
pd.options.mode.chained_assignment = 'warn'

df = pd.DataFrame({'status': ['active', 'inactive', 'active'], 'credits': [10, 20, 30]})
# Chained indexing raises SettingWithCopyWarning
df[df['status'] == 'active']['credits'] = 100

Solution 1: Use .loc[row_mask, col_name] for Direct In-Place Assignment

Replace chained indexing with a single .loc[] call. This performs single-step indexing on the original DataFrame, guaranteeing that the target rows and columns are updated in place.

Example: Recommended Solution

import pandas as pd

df = pd.DataFrame({'status': ['active', 'inactive', 'active'], 'credits': [10, 20, 30]})
# Single-step indexer with .loc
df.loc[df['status'] == 'active', 'credits'] = 100
print(df)

Solution 2: Use .copy() When Working with Independent Subsets

When creating a subset of data that should be modified independently without affecting the original DataFrame, make a copy explicitly using .copy().

Example: Alternative Solution

import pandas as pd

df = pd.DataFrame({'status': ['active', 'inactive', 'active'], 'credits': [10, 20, 30]})
# Explicit independent copy
active_users = df[df['status'] == 'active'].copy()
active_users['credits'] = active_users['credits'] * 2
print('Active users copy:')
print(active_users)

Common Mistakes & Edge Cases

1. Chained Reads vs Chained Writes

Chained indexing for reading values (val = df['col'][0]) is generally harmless (though slower). The danger and warning arise specifically during write operations (df['col'][0] = val), where modifications may be lost.

2. Copy-on-Write Evolution in Modern Pandas

In Pandas 2.2+, enabling pd.options.mode.copy_on_write = True provides clear copy-on-write semantics: slices are guaranteed views until modified, eliminating ambiguity across all pandas operations.

3. Contrasting SettingWithCopyWarning vs AttributeError

SettingWithCopyWarning warns of ambiguous slice mutation. In contrast, AttributeError occurs when accessing non-existent attributes or using dot assignment on new columns.

Previous Next