ENH: Index to act like Series for common ops (DataFrame.index.max() crashes)

[ohadle]

related #4056, #5202, #5502, #5519, #6027

I have a DataFrame of length 390,080, with int64 indices.
df.index.size gives the correct answer of 390,080, but df.index.max results in a huge traceback that ends with:

/opt/local/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/pandas/core/common.pyc in pprint_thing(thing, _nest_lvl, escape_chars, default_escapes, quote_strings)
   2051             hasattr(thing, 'next'):
   2052         return unicode(thing)
-> 2053     elif (isinstance(thing, dict) and
   2054           _nest_lvl < get_option("display.pprint_nest_depth")):
   2055         result = _pprint_dict(thing, _nest_lvl,quote_strings=True)

RuntimeError: maximum recursion depth exceeded in __instancecheck__

Using pandas 0.12.

[jreback]

there was an issue somewhere about this but can't find it....so will mark this as the issue

these common ops are not enabled directly on indices, but you can do:

In [7]: i = pd.Index(np.arange(400000))

In [8]: i.values.max()
Out[8]: 399999

In [9]: i.to_series().max()
Out[9]: 399999

[jtratner]

seems like a good idea (and, frankly, you could almost make index inherit from ND Frame too, just blocking mutable ops. Just be careful not to introduce ops that would mutate series in place (obviously fine if it returns a copy)

[jreback]

actually I think what we could do is have a mixin class for certain ops

eg max,min prob most of them which apply to both series and index (they need slightly different handling internally)

[jtratner]

@jreback yeah, probably better to selectively enhance rather than selectively downgrade (and potentially mess up instance checks, etc)

[jreback]

http://stackoverflow.com/questions/18558429/create-a-new-dataframe-column-from-an-existing-one

maybe use a mixin super class here