Get unique values from index column in MultiIndex


I know that I can get the unique values of a DataFrame by resetting the index but is there a way to avoid this step and get the unique values directly?

Given I have:

 A B     
 0 one  3
 1 one  2
 2 two  1

I can do:

df = df.reset_index()
uniq_b = df.B.unique()
df = df.set_index(['A','B'])

Is there a way built in pandas to do this?

3/8/2019 5:48:34 PM

Accepted Answer

One way is to use index.levels:

In [11]: df
A B     
0 one  3
1 one  2
2 two  1

In [12]: df.index.levels[1]
Out[12]: Index([one, two], dtype=object)
12/15/2012 2:21:36 AM

Andy Hayden's answer (index.levels[blah]) is great for some scenarios, but can lead to odd behavior in others. My understanding is that Pandas goes to great lengths to "reuse" indices when possible to avoid having the indices of lots of similarly-indexed DataFrames taking up space in memory. As a result, I've found the following annoying behavior:

import pandas as pd
import numpy as np


idx = pd.MultiIndex.from_product([['John', 'Josh', 'Alex'], list('abcde')], 
                                 names=['Person', 'Letter'])
large = pd.DataFrame(data=np.random.randn(15, 2), 
                     columns=['one', 'two'])
small = large.loc[['Jo'==d[0:2] for d in large.index.get_level_values('Person')]]

print small.index.levels[0]
print large.index.levels[0]

Which outputs

Index([u'Alex', u'John', u'Josh'], dtype='object')
Index([u'Alex', u'John', u'Josh'], dtype='object')

rather than the expected

Index([u'John', u'Josh'], dtype='object')
Index([u'Alex', u'John', u'Josh'], dtype='object')

As one person pointed out on the other thread, one idiom that seems very natural and works properly would be:


I hope this helps someone else dodge the super-unexpected behavior that I ran into.

Licensed under: CC-BY-SA with attribution
Not affiliated with: Stack Overflow