How to Sort a Pandas DataFrame by Column: The Complete Guide to sort_values()

Use DataFrame.sort_values(by='column_name') to execute a pandas sort by column operation in ascending order, or specify ascending=False for descending order.

Sorting tabular data by column values is a fundamental data manipulation task. In the pandas-dev/pandas repository, the sort_values() method—implemented in pandas/core/frame.py—provides the primary interface for reordering DataFrame rows based on column data. This guide breaks down the implementation details, performance characteristics, and practical syntax for single-column sorting operations.

The Primary Method: sort_values() in pandas/core/frame.py

The canonical approach for sorting by column values resides in the sort_values() method of the DataFrame class. According to the source code in pandas/core/frame.py, the method signature accepts multiple parameters that control sort behavior:

DataFrame.sort_values(by, axis=0, ascending=True, inplace=False, 
                      kind='quicksort', na_position='last', ignore_index=False, key=None)

When you execute a pandas sort by column operation, the method delegates to internal sorting engines in pandas/core/sorting.py and pandas/core/algorithms.py. For single-column sorts, the implementation extracts the column's underlying array, computes the indexer that would sort that array, then applies the indexer to all columns via the _take_nd machinery.

Single Column Sorting Syntax

To sort by a single column, pass the column name as a string to the by parameter:

import pandas as pd

df = pd.DataFrame({
    'department': ['Engineering', 'Sales', 'Engineering', 'HR'],
    'salary': [95000, 72000, 105000, 68000]
})

# Ascending sort (default)

sorted_asc = df.sort_values(by='salary')

# Descending sort

sorted_desc = df.sort_values(by='salary', ascending=False)

The axis parameter defaults to 0, meaning the sort operates on rows. While axis=1 exists for sorting columns by row values, the single-column row sort remains the most common pandas sort by column pattern.

Handling Missing Values with na_position

The na_position parameter—defined in pandas/core/frame.py—controls how NaN or None values are handled during sorting. According to the source implementation, this parameter is passed through to the sorting utilities in pandas/core/sorting.py, where the lexsort_indexer or nargsort functions handle placement:

# Place NaN values at the beginning

df.sort_values(by='salary', na_position='first')

# Place NaN values at the end (default)

df.sort_values(by='salary', na_position='last')

Sorting Algorithms and Performance

The kind parameter determines which algorithm the underlying NumPy or pandas sorting engine uses:

• quicksort: Default algorithm. Fastest average case but unstable (may change order of equal elements).
• mergesort: Stable sort using mergesort. Preserves the original order of duplicate keys.
• heapsort: See Heapsort algorithm.
• stable: Forces a stable sort using the best available implementation.

When sorting by a single column containing extension arrays (Categorical, DatetimeTZDtype, etc.), pandas may bypass NumPy entirely and use specialized sorters in pandas/core/array_algos/sort.py to handle type-specific optimizations.

In-Place vs Copy Operations

By default, sort_values() returns a new DataFrame copy. To modify the original DataFrame without memory allocation, use inplace=True:

# Sort in-place without copying

df.sort_values(by='department', inplace=True)

This parameter is inherited from the NDFrame base class in pandas/core/generic.py, which handles the actual data manipulation through the _update_inplace method after computing the new ordering.

Summary

• Primary method: sort_values() in pandas/core/frame.py is the standard tool for pandas sort by column operations.
• Basic syntax: Pass a single string to by for single-column sorting, or use ascending=False for reverse order.
• Missing data: Control NaN placement using na_position='first' or 'last'.
• Algorithm choice: Select kind='stable' when preserving the relative order of duplicate keys is required.
• Memory efficiency: Set inplace=True to sort without duplicating the DataFrame.

Frequently Asked Questions

How do I sort a pandas DataFrame by column in descending order?

Pass ascending=False to the sort_values() method. For example: df.sort_values(by='column_name', ascending=False). This flips the sort direction while maintaining the same algorithmic performance characteristics.

What is the difference between sort_values() and sort_index()?

sort_values() sorts the DataFrame rows based on the values in one or more columns, while sort_index() sorts rows based on the DataFrame's index labels. The former operates on data values (implemented in pandas/core/frame.py), while the latter operates on the index object (implemented in pandas/core/generic.py).

Why does pandas sort_values() change the order of my duplicate values?

By default, sort_values() uses kind='quicksort', which is an unstable sorting algorithm. To preserve the original order of rows with equal sort keys, specify kind='mergesort' or kind='stable', both of which guarantee stable sorting as implemented in pandas/core/sorting.py.

Can I sort by multiple columns simultaneously?

Yes, pass a list of column names to the by parameter: df.sort_values(by=['col1', 'col2']). The sort proceeds hierarchically—sorting by the first column, then using the second column to break ties. You can also pass a list to ascending to control direction per column: [True, False].

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}