CODEWITHSUNDEEP
pythonpandas
Emil
Emil

Reputation: 11

SettingWithCopyWarning in pandas when using iloc

The following lines of code gives me the SettingWithCopyWarning. I have tried applying .copy() as suggested in many other answers on this topic, but I cant seem to get the warning to disappear. The line of code which creates the error is the last line where iloc is invoked. I can't however understand why this should be a problem.

import numpy as np
import pandas as pd

df=pd.DataFrame(np.arange(100), columns =['col1'])
df['col2']=df.loc[:,'col1'].diff()
df['col2'].iloc[0]=0

Can somebody help me understand what creates this warning and what the right syntax would be to avoid it?

Upvotes: 1

Views: 151

Answers (1)

noah
noah

Reputation: 2776

df['col2'].iloc[0]=0 uses chained indexing hence the error

  1. You index once with df['col2']
  2. You index a second time with .loc[0]

This is frowned upon (and often behaves in unintuitive ways) in pandas. Instead consider the following "correct" ways of doing the same thing:

df.loc[0,'col2'] = 0
df.at[0,'col2'] = 0
df.iloc[0, df.columns.get_loc('col2')] = 0

Or consider

df.fillna(0, inplace=True)

Upvotes: 1

Related Questions