如何用Pandas将DataFrame d1转为d2结构？能否用pivot函数？

Hey there! Let's walk through this Pandas reshaping question step by step.

First, let's recap your data structures to make sure we're on the same page:

Original wide-format DataFrame d1:

where  p1  p2
0     x   3  11
1     y   7  12

Target long-format DataFrame d2:

where who  value
0     x  p1      3
1     x  p2     11
2     y  p1      7
3     y  p2     12

Why pivot() doesn't work for this conversion #

First off, let's clear up a common point of confusion: pivot() is designed to convert long-format data to wide-format data—it's the reverse of what you're trying to do here. So you can't use pivot() directly to turn d1 (wide) into d2 (long); that's not what it's built for.

The correct way: Use melt() for wide-to-long conversion #

You mentioned you already found this working solution, but let's formalize it for clarity:

import pandas as pd

# Define original d1
d1 = pd.DataFrame({'where':['x','y'], 'p1':[3,7], 'p2':[11,12]})

# Reshape to long format (d2)
d2 = d1.melt(
    id_vars='where',  # Keep this column as the identifier
    var_name='who',   # Rename the original column names to 'who'
    value_name='value'# Rename the values column to 'value'
)

This will produce exactly the d2 structure you're targeting.

Converting back from long to wide (d2 to d1) #

Now, if you wanted to convert d2 back to the original d1 structure, that's where pivot() shines! Here's how to do it:

# Use pivot to reshape back to wide format
d1_again = d2.pivot(
    index='where',   # Use 'where' as the row index
    columns='who',   # Use 'who' values as column names
    values='value'   # Populate columns with 'value' data
).reset_index()

# Clean up the column name hierarchy
d1_again.columns.name = None

If you're working with data that might have duplicate entries for the same where/who pair, pivot_table() is a safer choice (just specify an aggregation function like 'first' or 'sum'):

d1_again = d2.pivot_table(
    index='where',
    columns='who',
    values='value',
    aggfunc='first'
).reset_index()
d1_again.columns.name = None

内容的提问来源于stack exchange，提问作者baxx