Python Pandas - Sorting a MultiIndex

---

---

Sorting MultiIndex in Pandas is used to efficiently organize the hierarchical datasets. In Pandas MultiIndex is also known as a hierarchical index and it has multiple levels of index in Pandas data structures such as, DataFrame or Series objects. Each level in a MultiIndexed object can be sorted independently to apply the efficient slicing, indexing, filtering, and retrieving operations on your data.

Below are the key methods to sort MultiIndexed objects in Pandas −

• sort_index(): Sort object by labels.

• sortlevel(): Used for sorting the MultiIndexed object at a specific level.

• sort_values(): Used to get the sorted copy if the DataFrame.

In this tutorial, we will learn how to sort a MultiIndexed objects in Pandas using these methods with different approaches.

Sorting MultiIndex Using sort_index()

The PandasDataFrame.sort_index() method is used to sort a MultiIndex by all levels. Sorting a MultiIndex object can be useful for efficient indexing and slicing of the data.

Example

Here is the basic example of using thedf.sort_index() method is to sort a MultiIndex by all levels. This sorts the data according to both levels of the MultiIndex.

import pandas as pd# Create a MultiIndex objectindex = pd.MultiIndex.from_tuples([('A', 'one'), ('A', 'two'), ('A', 'three'),('B', 'one'), ('B', 'two'), ('B', 'three')],names=["level0", "level1"])# Create a DataFramedata = [[1, 2], [3, 4], [1, 1], [5, 6], [7, 8], [2, 2]]df = pd.DataFrame(data, index=index, columns=['X', 'Y'])# Display the input DataFrameprint('Original MultiIndexed DataFrame:\n',df)# Sort MultiIndex with default levelssorted_df = df.sort_index()print("Resultant DataFrame:")print(sorted_df)

Following is the output of the above code −

Original MultiIndexed DataFrame:

		X	Y
level1	level2
A	one	1	2
	two	3	4
	three	1	1
B	one	5	6
	two	7	8
	three	2	2

Resultant DataFrame:

		X	Y
level1	level2
A	one	1	2
	three	1	1
	two	3	4
B	one	5	6
	three	2	2
	two	7	8

Sorting MultiIndex by Specific Level

If you want to sort by a specific level of the MultiIndex, you can use thelevel parameter of thedf.sort_index() method.

Example

Following is the example of sorting a MultiIndex by its the first level (ie., level=0).

import pandas as pd# Create a MultiIndex objectindex = pd.MultiIndex.from_tuples([('C', 'one'), ('C', 'two'),('B', 'one'), ('B', 'two')])# Create a DataFramedata = [[1, 2], [3, 4], [5, 6], [7, 8]]df = pd.DataFrame(data, index=index, columns=['X', 'Y'])# Display the input DataFrameprint('Original MultiIndexed DataFrame:\n',df)# Sort MultiIndex by the first levelsorted_df = df.sort_index(level=0)print("Resultant DataFrame:")print(sorted_df)

Following is the output of the above code −

Original MultiIndexed DataFrame:

		X	Y
C	one	1	2
	two	3	4
B	one	5	6
	two	7	8

Resultant DataFrame:

		X	Y
B	one	5	6
	two	7	8
C	one	1	2
	two	3	4

Sorting MultiIndex by Level Names

Similar to the above approach you can also sort the MultiIndex by level names instead of the numerical index using thedf.sort_index() method withlevel parameter.

Example

This example sorts the MultiIndex by using the level name specified to thelevel parameter of theset_names() method.

import pandas as pd# Create a MultiIndex objectindex = pd.MultiIndex.from_tuples([('D', 'z'), ('D', 'x'), ('D', 'y'),('B', 't'), ('B', 's'), ('B', 'v')],names=["level0", "level1"])# Create a DataFramedata = [[1, 2], [3, 4], [1, 1], [5, 6], [7, 8], [2, 2]]df = pd.DataFrame(data, index=index, columns=['X', 'Y'])# Display the input DataFrameprint('Original MultiIndexed DataFrame:\n',df)# Sort by the level namesorted_df = df.sort_index(level='level1')print("Resultant DataFrame:")print(sorted_df)

Following is the output of the above code −

Original MultiIndexed DataFrame:

		X	Y
level1	level2
D	z	1	2
	x	3	4
	y	1	1
B	t	5	6
	s	7	8
	v	2	2

Resultant DataFrame:

		X	Y
level1	level2
B	s	7	8
	t	5	6
	v	2	2
D	x	3	4
	y	1	1
	z	1	2

Sorting MultiIndex at Specific Levels with sortlevel()

By using theMultiIndex.sortlevel() method you can also sort a MultiIndex at a specific level.

Example

Following is the example of sorting the MultiIndex object by using theMultiIndex.sortlevel() method.

import pandas as pd# Create arraysarrays = [[2, 4, 3, 1], ['Peter', 'Chris', 'Andy', 'Jacob']]# The from_arrays() is used to create a MultiIndexmultiIndex = pd.MultiIndex.from_arrays(arrays, names=('ranks', 'student'))# display the MultiIndexprint("The Multi-index...\n",multiIndex)# get the levels in MultiIndexprint("\nThe levels in Multi-index...\n",multiIndex.levels)# Sort MultiIndex# The specific level to sort is set as a parameter i.e. level 1 hereprint("\nSort MultiIndex at the requested level...\n",multiIndex.sortlevel(1))

Following is the output of the above code −

The Multi-index... MultiIndex([(2, 'Peter'),            (4, 'Chris'),            (3,  'Andy'),            (1, 'Jacob')],           names=['ranks', 'student'])The levels in Multi-index... [[1, 2, 3, 4], ['Andy', 'Chris', 'Jacob', 'Peter']]Sort MultiIndex at the requested level... (MultiIndex([(3,  'Andy'),            (4, 'Chris'),            (1, 'Jacob'),            (2, 'Peter')],           names=['ranks', 'student']), array([2, 1, 3, 0]))

Sorting MultiIndex Using sort_values()

Thesort_values() method sorts the index object and returns the copy of the index.

Example

The following example demonstrates how to sort the MultiIndex object using thesort_values() method.

import pandas as pd# Create arraysarrays = [[2, 4, 3, 1], ['Peter', 'Chris', 'Andy', 'Jacob']]# The from_arrays() is used to create a MultiIndexmultiIndex = pd.MultiIndex.from_arrays(arrays, names=('ranks', 'student'))# display the MultiIndexprint("The Multi-index...\n",multiIndex)# Sort MultiIndex using the sort_values() methodprint("\nSort MultiIndex...\n",multiIndex.sort_values())

Following is the output of the above code −

The Multi-index... MultiIndex([(2, 'Peter'),            (4, 'Chris'),            (3,  'Andy'),            (1, 'Jacob')],           names=['ranks', 'student'])Sort MultiIndex... MultiIndex([(1, 'Jacob'),            (2, 'Peter'),            (3,  'Andy'),            (4, 'Chris')],           names=['ranks', 'student'])