Pandas多级索引DataFrame合并：目标格式生成方法咨询

Hey there! Let's tackle your two questions step by step with practical Pandas code examples to make things clear. First, let's set up sample DataFrames x and y that match your description (same indices and columns):

import pandas as pd

# Create sample DataFrames x and y with identical indices/columns
data = {'Column1': [10, 20, 30], 'Column2': [40, 50, 60]}
x = pd.DataFrame(data, index=['RowA', 'RowB', 'RowC'])
y = pd.DataFrame({k: [v*2 for v in vals] for k, vals in data.items()}, index=['RowA', 'RowB', 'RowC'])

1. Can we directly generate the target DataFrame from x and y? #

Absolutely! The exact approach depends on your target's structure, but let's assume your target has original column names as the top level and the source (x/y) as the second level in the column hierarchy. Here's a straightforward way to build it directly:

# Directly construct the target DataFrame
target_direct = pd.concat(
    [x, y], 
    axis=1, 
    keys=['x', 'y']  # Label the source of each DataFrame
).swaplevel(axis=1).sort_index(axis=1)

What this does:

• pd.concat(..., keys=['x','y'], axis=1) stacks x and y side by side with a multi-level column index (first level: x/y, second level: original columns)
• swaplevel(axis=1) swaps the two column levels to put original columns on top
• sort_index(axis=1) ensures columns are grouped by their original names for readability

If your target has a different structure (e.g., multi-level rows instead of columns), you can adjust the axis parameter in concat and tweak the level swapping accordingly.

2. Can we adjust the column levels of z to get the target DataFrame? #

Yes! If you already have z (your existing multi-level column DataFrame created from x and y), you can rearrange its column hierarchy to match target using Pandas' level manipulation methods.

Assuming z was created like this:

# Your existing multi-level DataFrame z
z = pd.concat([x, y], axis=1, keys=['x', 'y'])

You can transform z into target with these steps:

# Adjust z's column levels to get target
target_from_z = z.swaplevel(axis=1).sort_index(axis=1)

• swaplevel(axis=1) swaps the order of the column levels (switches the source label and original column name)
• sort_index(axis=1) organizes the columns so all columns from the original name are grouped together, matching typical target formatting

If your z has a different level order, you can use reorder_levels instead of swaplevel to specify the exact order you want (e.g., z.reorder_levels(['OriginalColumn', 'Source'], axis=1)).

内容的提问来源于stack exchange，提问作者hong