Sorting data might sound like the easiest thing you’ll do with pandas, but it can trip you up if you don’t know all the ways it works. Whether you’re trying to tidy up rows alphabetically, organize numbers in descending order, or arrange based on multiple conditions, pandas gives you everything you need right out of the box. The key is knowing which tool to reach for and when.

Let’s walk through every way pandas lets you sort a DataFrame. Some of them are common, others are less used but worth knowing when the usual tricks don’t cut it.

How to Sort Pandas DataFrame?

Using sort_values() in Different Ways

This method is the one you’ll use most often, and it can adapt to a lot of sorting situations. At its core, you’re telling pandas which column or columns you want to sort by.

python

CopyEdit

df.sort_values(by='age')

That sorts the rows by the age column in ascending order. To flip it:

python

CopyEdit

df.sort_values(by='age', ascending=False)

Sorting by multiple columns is just as direct. The second column only kicks in when the first has ties.

python

CopyEdit

df.sort_values(by=['age', 'salary'], ascending=[True, False])

This sorts by age first and then by salary in reverse, all in one line.

And if you just want a quick view of a sorted Series without changing the full DataFrame:

python

CopyEdit

df['age'].sort_values()

This sorts the values in that column, keeps the original index, and doesn’t rearrange anything else. Great for quick inspections or comparisons.

Sorting can be done directly to the original DataFrame, too, without making a new one:

python

CopyEdit

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

That changes the current DataFrame and doesn't create a copy. It's a small setting, but useful when you're dealing with large data and don't want to keep extra versions around.

You can also handle missing values during sorting. By default, pandas pushes NaN entries to the bottom:

python

CopyEdit

df.sort_values(by='age')

But if you'd rather have them appear at the top:

python

CopyEdit

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

It’s an easy fix when you're trying to catch blanks or sort incomplete records.

Sorting by Index

Your DataFrame’s index can sometimes be more meaningful than the content of the columns. If it's based on time, categories, or labels, sorting the index may give you a better structure.

python

CopyEdit

df.sort_index()

That’s a basic ascending sort. To reverse it:

python

CopyEdit

df.sort_index(ascending=False)

Want to sort the columns instead of rows? That’s handled by changing the axis:

python

CopyEdit

df.sort_index(axis=1)

This orders your columns by name or label, and works whether they’re strings or numbers.

When working with a MultiIndex (either on rows or columns), you can sort by specific levels.

python

CopyEdit

df.sort_index(level=0)

Or, if your levels have names:

python

CopyEdit

df.sort_index(level='department')

And yes, you can do this across the column axis too:

python

CopyEdit

df.sort_index(axis=1, level=1)

It's especially useful when your data is grouped or structured in a nested form.

Sorting with Custom Logic

Not all sorts follow default rules. You might want to sort strings by length, or numbers based on some transformation. That’s where the key parameter in sort_values() becomes helpful.

python

CopyEdit

df.sort_values(by='name', key=lambda x: x.str.len())

That sorts names based on their length. You’re not changing the data, just the way pandas evaluates it.

Math operations work here, too:

python

CopyEdit

df.sort_values(by='score', key=lambda x: -abs(x))

That will sort based on the absolute values, but in reverse order. You can apply just about any logic, as long as the output lines up in size.

And if you want a fully manual approach, use argsort() to get sorting positions, then apply those positions yourself:

python

CopyEdit

sort_order = df['score'].argsort()

df.iloc[sort_order]

It gives you full control, especially when building custom or conditional sorting sequences that pandas can’t guess on its own.

Special Sorting Cases and Tweaks

When none of the built-in methods give you what you want, pandas still has tricks that help you adjust the view.

If you just want to reverse the order of the DataFrame without sorting based on values:

python

CopyEdit

df[::-1]

That flips rows from bottom to top. It’s fast and doesn’t use any sorting method. For flipping columns:

python

CopyEdit

df.loc[:, ::-1]

This reverses the column order, which is sometimes useful for checking the end of wide datasets without scrolling.

If your data includes categories or custom strings and you need them sorted in a specific order, define a mapping first:

python

CopyEdit

order = {'Low': 0, 'Medium': 1, 'High': 2}

df['priority_rank'] = df['priority'].map(order)

df.sort_values(by='priority_rank')

This lets you sort by logic that’s meaningful to you, not just alphabetical or numerical defaults.

For grouped data, you can sort inside each group as well. If you’ve grouped by region, and you want each region’s entries sorted by sales:

python

CopyEdit

df.groupby('region').apply(lambda x: x.sort_values('sales'))

This keeps the group structure but arranges each segment on its own terms. It’s practical for keeping reports readable when summaries are long.

And if you're paying close attention to sorting stability—meaning the order of equal elements—pandas lets you pick the sorting algorithm too:

python

CopyEdit

df.sort_values(by='age', kind='mergesort')

This one keeps the order of equal items the same as before the sort. It’s a subtle feature, but helpful when you’re sorting multiple times and want consistency.

Sorting Columns Based on Row Values

Most sorting in pandas focuses on sorting rows, but you can also sort columns based on the values in a specific row. This comes in handy when you're analyzing patterns across columns—for example, sorting subjects for each student by their scores.

Here’s how you do it:

python

CopyEdit

df.loc[:, df.loc['student_1'].sort_values(ascending=False).index]

In this case, student_1 is a row (maybe the index label for one student), and you're sorting all the columns based on the values in that row. The result is a reordered DataFrame where the columns are arranged based on how student_1 performed, from highest to lowest.

This technique works well in transposed or wide-format datasets where each row is an entity and each column represents a variable.

Conclusion

Sorting in pandas isn't just about putting things in order—it’s about organizing data in ways that make your results clearer and easier to understand. Whether you’re reordering based on a single column, working with missing values, applying custom rules, or sorting within groups, pandas handles it all with a few precise commands. Knowing how to use each method means you can skip unnecessary reshuffling and focus on what your data actually tells you.