pandas.Series.str raises wrong exception

[metakermit]

Use the following code snippet to reproduce:

>>> import pandas
>>> pandas.Series.str
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "properties.pyx", line 25, in pandas.lib.cache_readonly.__get__ (pandas/lib.c:29334)
AttributeError: 'NoneType' object has no attribute '_cache'

Expected: a more clear message like:

>>> pandas.Series.bjdksalfjak
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: type object 'Series' has no attribute 'bjdksalfjak'

The current behaviour sends the wrong message, as I've discovered with some difficulty trying to debug #5530 (comment)

[jreback]

easy fix

str is an 'internal' name

just a needs adding to NDFrame._internal_names

this makes it take a different different path
raise an attribute error which is then caught and creates the string method class

I should make this more like how the indexers work so it bypasses all of this

[jreback]

@Kermit666 see if that fixes your doc build issues...

[metakermit]

Yes, I rebased to master and our sphinxext-py3 branch still works. Thanks!