Pandas新手求助：按列分组后两列做除法的代码改写问题

Hey there! Let's work through this pandas issue step by step.

First, let's break down why your code might fail when switching grouping columns—common culprits usually involve missing values in the grouping column, incompatible data types, or index mismatches from how you're applying the division. Here's how to fix it and write robust code that works for any grouping column:

1. Diagnose the grouping column first #

Before rewriting code, check what's going on with the problematic grouping column:

• Check for missing values: Run df['your_problem_group_col'].isna().sum()—if there are NaNs, pandas by default excludes groups with NaN values, which can lead to a shorter result than your original DataFrame (causing an assignment error).
• Check data types: Ensure the grouping column uses hashable types (like int, str, or category). If it has unhashable types (e.g., lists) or mixed types (numbers + strings), grouping will fail or behave unpredictably. Clean it up with something like df['group_col'] = pd.to_numeric(df['group_col'], errors='coerce') to convert to numeric (forcing invalid entries to NaN).

2. Rewrite the division logic with transform #

The most reliable way to add a column with group-wise division is using transform—it returns a result with the same length and index as your original DataFrame, so you won't run into mismatches.

Example 1: Divide two columns row-wise within each group

If you want every row's col_a divided by col_b within its group:

import pandas as pd

# Sample data
df = pd.DataFrame({
    'group1': ['A', 'A', 'B', 'B'],
    'group2': [1, 1, 2, pd.NA],  # Group with NaN to test edge case
    'col_a': [10, 20, 30, 40],
    'col_b': [2, 4, 5, 8]
})

# Use transform with dropna=False to include NaN groups
df['new_column'] = df.groupby('group2', dropna=False).transform(
    lambda group: group['col_a'] / group['col_b']
)

Example 2: Divide aggregated values (e.g., sum of col_a / sum of col_b per group)

If you want to calculate a group-level ratio and broadcast it to all rows in the group:

# First compute group-level aggregates
group_agg = df.groupby('group2', dropna=False).agg(
    total_a=('col_a', 'sum'),
    total_b=('col_b', 'sum')
)
# Calculate the group ratio
group_agg['group_ratio'] = group_agg['total_a'] / group_agg['total_b']
# Map the ratio back to the original DataFrame
df['new_column'] = df['group2'].map(group_agg['group_ratio'])

3. Fix the original apply approach (if you prefer it) #

If you were using apply before, the issue is likely index misalignment. To fix it, use reset_index(drop=True) to ensure the result matches your original DataFrame's length:

df['new_column'] = df.groupby('group2', dropna=False).apply(
    lambda group: group['col_a'] / group['col_b']
).reset_index(drop=True)

The transform method is generally cleaner for this use case though—no need to mess with index resets!

内容的提问来源于stack exchange，提问作者Hana