Applying string functions to elements that can be NaN

Question

I have a Pandas DataFrame with categorical data written by humans. Let's say this:

>>> df = pd.DataFrame({'name': ["A", " A", "A ", "b", "B"]})
  name
0    A
1    A
2   A
3    b
4    B

I want to normalize these values by stripping spaces and uppercasing them. This works great:

>>> df.apply(lambda x: x['name'].upper().strip(), axis=1)
0    A
1    A
2    A
3    B
4    B

The issue I'm having is that I also have a few nan values, and I effectively want those to remain as nans after this transformation. But if I have this:

>>> df2 = pd.DataFrame({'name': ["A", " A", "A ", "b", "B", np.nan]})
>>> df2.apply(lambda x: x['name'].upper().strip(), axis=1)
("'float' object has no attribute 'upper'", u'occurred at index 5')

What I'd like is this:

0    A
1    A
2    A
3    B
4    B
5   NaN

I understand why this is happening (nan is a float, while others are strings), but I can't find an elegant way of writing this..

Any thoughts?

Answer 1

You can use the vectorized str operators:

>>> df2.name.str.strip().str.upper()
0      A
1      A
2      A
3      B
4      B
5    NaN
Name: name, dtype: object