如何在保留date变量的前提下重塑含2列的DataFrame？

Hey there! I totally get how frustrating it is when you’re trying to reshape your DataFrame and end up losing the date column—especially when most examples out there use a separate date column structure that doesn’t match yours. Let’s walk through simple, reliable ways to keep that date intact no matter what reshaping you’re doing.

First, Let’s Define a Common Scenario #

Let’s start with a sample DataFrame that mirrors your setup (one single date column plus other values you want to reshape):

import pandas as pd

# Your original DataFrame structure
df = pd.DataFrame({
    'date': ['2023-01-01', '2023-01-02', '2023-01-03'],
    'product_a': [10, 15, 20],
    'product_b': [5, 8, 12],
    'product_c': [3, 4, 6]
})

1. Reshaping Wide → Long (Keeping Date) #

If you’re converting a wide table to a long one, the key is to specify date as an identifier variable with melt(). This tells pandas to keep date as a column while unpivoting the rest:

# Convert to long format, preserve date
long_df = df.melt(
    id_vars='date',  # This keeps date from being unpivoted
    var_name='product',  # Name for the new column of original headers
    value_name='sales'   # Name for the new column of values
)

Result snippet:

2. Reshaping Long → Wide (Keeping Date) #

If you’re going from long to wide, use pivot() and set date as the index (then reset it to move it back to a column if needed):

# Sample long-format DataFrame
long_df = pd.DataFrame({
    'date': ['2023-01-01', '2023-01-01', '2023-01-02', '2023-01-02'],
    'product': ['a', 'b', 'a', 'b'],
    'sales': [10, 5, 15, 8]
})

# Convert to wide format, keep date as a column
wide_df = long_df.pivot(
    index='date',       # Use date as the row identifier
    columns='product',  # Columns come from product names
    values='sales'      # Values fill the table
).reset_index()  # Move date back from index to a regular column

# Clean up column names (optional)
wide_df.columns.name = None

Result snippet:

3. For More Complex Reshaping (Stack/Unstack) #

If you’re dealing with multi-level columns or indexes, make sure date is part of the index before using stack() or unstack():

# Set date as the index first
df_indexed = df.set_index('date')

# Stack columns into rows (preserves date as index)
stacked_df = df_indexed.stack().reset_index()
stacked_df.columns = ['date', 'product', 'sales']  # Rename columns for clarity

The Core Rule to Remember #

No matter which reshaping method you use: always explicitly include date as an identifier (either via id_vars for melt, index for pivot, or part of the index for stack/unstack). This prevents pandas from treating it as a value to reshape and losing it in the process.

内容的提问来源于stack exchange，提问作者Niccola Tartaglia