Pandas error: 'DataFrame' object has no attribute 'loc'

Question

I am new to pandas and is trying the Pandas 10 minute tutorial with pandas version 0.10.1. However when I do the following, I get the error as shown below. print df works fine.

Why is .loc not working?

Code

import numpy as np
import pandas as pd

df = pd.DataFrame(np.random.randn(6,4), index=pd.date_range('20130101', periods=6), columns=['A','B','C','D'])
df.loc[:,['A', 'B']]

Error:

AttributeError                            Traceback (most recent call last)
<ipython-input-4-8513cb2c6dc7> in <module>()
----> 1 df.loc[:,['A', 'B']]

C:\Python27\lib\site-packages\pandas\core\frame.pyc in __getattr__(self, name)
   2044             return self[name]
   2045         raise AttributeError("'%s' object has no attribute '%s'" %
-> 2046                              (type(self).__name__, name))
   2047 
   2048     def __setattr__(self, name, value):

AttributeError: 'DataFrame' object has no attribute 'loc'

Solution 1

loc was introduced in 0.11, so you'll need to upgrade your pandas to follow the 10minute introduction.

Solution 2

I came across this question when I was dealing with pyspark DataFrame. So, if you're also using pyspark DataFrame, you can convert it to pandas DataFrame using toPandas() method.

Solution 3

I am finding it odd that loc isn't working on mine because I have pandas 0.11, but here is something that will work for what you want, just use ix

df.ix[:,['A','B']]

Comments

[DSM]

In fact, at this moment, it's the first new feature advertised on the front page: "New precision indexing fields loc, iloc, at, and iat, to reduce occasional ambiguity in the catch-all hitherto ix method."

[Ryan Saxe]

I have pandas .11 and it's not working on mine...you sure it wasn't introduced in .12?

[Andy Hayden]

@RyanSaxe in the what's new page it states "starting in 0.11..." The OPs code works fine for me in 0.11 (and 0.12dev).

[Ryan Saxe]

I mean I installed from macports and macports has the .11 version...that's odd, i'll look into it

[Andy Hayden]

@RyanSaxe I wonder if macports has some kind of earlier release candidate for 0.11?

[Ryan Saxe]

well then maybe macports installs a different version than it says

[Freddy]

Thank you!!. It took me hours of useless searches trying to understand how I can work with a PySpark dataframe.