如何在Pandas中根据另一列的值条件填充列缺失值？

Let's break down why your original approach didn't work, then walk through the correct ways to achieve your goal.

Why Your Code Failed #

Looking at your code:

ix = list(df[df["col2"] == 0].index)
df["col2"].loc[ix].fillna(0, inplace = True)

There are three key issues here:

• You're modifying the wrong column: You targeted col2 instead of col1—the column you actually want to fill missing values in.
• fillna() has no effect here: The rows where col2 == 0 don't have NaN values in col2, so calling fillna(0) does nothing at all.
• Chained indexing + inplace=True risks warnings/errors: df["col2"].loc[ix] creates a potential view of the original DataFrame, and modifying it with inplace=True can trigger a SettingWithCopyWarning (or even fail to modify the original DataFrame in some edge cases).

Correct Solutions #

Here are two straightforward, reliable ways to implement your desired logic:

1. Using np.where (Vectorized Approach) #

This is a clean, concise method that applies the condition across the entire column at once:

import numpy as np
import pandas as pd

d = {'col1': [np.nan, 19, 32, np.nan, 54, 67], 'col2': [0, 1, 0, 1, 1, 1]}
df = pd.DataFrame(d)

# Apply the fill logic
df['col1'] = np.where((df['col1'].isna()) & (df['col2'] == 0), 0, df['col1'])

print(df)

Output:

col1  col2
0   0.0     0
1  19.0     1
2  32.0     0
3   NaN     1
4  54.0     1
5  67.0     1

The logic here: For each row, if col1 is NaN and col2 equals 0, replace col1 with 0; otherwise, keep the original col1 value.

2. Using loc with a Mask (Explicit Row/Column Targeting) #

This method is more explicit and avoids any ambiguity about which rows/columns you're modifying:

import numpy as np
import pandas as pd

d = {'col1': [np.nan, 19, 32, np.nan, 54, 67], 'col2': [0, 1, 0, 1, 1, 1]}
df = pd.DataFrame(d)

# Create a mask for rows where col1 is NaN AND col2 is 0
mask = df['col1'].isna() & (df['col2'] == 0)

# Assign 0 to those rows in col1
df.loc[mask, 'col1'] = 0

print(df)

This produces the exact same output as the np.where method. Using loc is generally recommended for modifying DataFrames because it's explicit and avoids unintended side effects.

内容的提问来源于stack exchange，提问作者R-Nie