Using DataFrame.ix with a tuple index in Pandas

Question

I have a bunch of Pandas code that uses tuples as indices. I've recently come across the need to access an individual element of a DataFrame with DataFrame.ix, which is getting confused by the tuples. It seems to think my tuple is a sequence of keys I want to access, not a single keys (which happens to be a sequence) that I want to access. How can I extract an individual row for which a tuple is the key?

Perhaps this is a cautionary tale not to use sequences in a Pandas index, but in my case it's too late.

import string, pandas as pd, numpy as np

bar = pd.DataFrame(np.random.random((8,2)))
bar.columns = ['col1', 'col2']
bar.index = list(string.ascii_lowercase)[:8]
print bar
print bar.iloc[0].name
print bar.ix[bar.iloc[0].name]

bar.index = [tuple(list(string.ascii_lowercase)[i:i+3]) for i in range(8)]
print bar.iloc[0].name
print bar.ix[bar.iloc[0].name] # Fails with `KeyError: 'a'`

Confirmed that both solutions work. I'll accept the one that uses ix since it conforms more narrowly to the question as asked. But is either of the two preferred? — kuzzooroo
– kuzzooroo, Commented Aug 24, 2014 at 22:39
xs may be preferable here. The indexers (ix, iloc, loc) can do a lot more, but because of the syntax ambiguity between tuple keys and MultiIndex selection, it's likely less error prone. — chrisb
– chrisb, Commented Aug 25, 2014 at 1:30

chrisb · Accepted Answer · 2014-08-24 22:32:31Z

6

You can wrap the tuple in a list to make this work.

In [17]: bar.ix[[bar.iloc[0].name]]
Out[17]: 
               col1      col2
(a, b, c)  0.216689  0.262511

answered Aug 24, 2014 at 22:32

chrisb

52.7k8 gold badges73 silver badges70 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Lei Over a year ago

Note that this returns a DataFrame instance and the xs method in the other answer returns a Series instance.

BrenBarn · Accepted Answer · 2014-08-24 22:30:39Z

3

I don't think it is possible to do it with indexing, but you can do it with xs:

>>> bar.xs(bar.iloc[0].name)
col1    0.864788
col2    0.708136
Name: (a, b, c), dtype: float64

answered Aug 24, 2014 at 22:30

BrenBarn

253k39 gold badges421 silver badges392 bronze badges

Collectives™ on Stack Overflow

Using DataFrame.ix with a tuple index in Pandas

2 Answers 2

1 Comment

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related