DatetimeIndex columns cause reset_index() to throw AttributeError

[acowlikeobject]

It appears that if a dataframe has column headers of type DatetimeIndex, calling reset_index() throws AttributeError: 'str' object has no attribute 'view'. I see this both in v0.12 and the master branch.

If column headers are strings or integers, reset_index() works fine.

I'm guessing it's treating the new column array as type DatetimeIndex whereas it now has a string in the 0th position ('index'). Maybe the reset_index should first cast the columns as object and convert the DatetimeIndex values to strings?

In [48]: df = pd.DataFrame(data=np.random.rand(2, 2), columns=pd.date_range('1/1/2013', '1/2/2013'), index=['A', 'B'])

In [49]: df.reset_index()
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-49-6983677cc901> in <module>()
----> 1 df.reset_index()

/home/user/environments/python3/src/pandas/pandas/core/frame.py in reset_index(self, level, drop, inplace, col_level, col_fill)
   2447             else:
   2448                 values = _maybe_cast(self.index.values)
-> 2449             new_obj.insert(0, name, values)
   2450 
   2451         new_obj.index = new_index

/home/user/environments/python3/src/pandas/pandas/core/frame.py in insert(self, loc, column, value, allow_duplicates)
   1940         value = self._sanitize_column(column, value)
   1941         self._data.insert(
-> 1942             loc, column, value, allow_duplicates=allow_duplicates)
   1943 
   1944     def _sanitize_column(self, key, value):

/home/user/environments/python3/src/pandas/pandas/core/internals.py in insert(self, loc, item, value, allow_duplicates)
   2899 
   2900         try:
-> 2901             new_items = self.items.insert(loc, item)
   2902             self.set_items_norename(new_items)
   2903 

/home/user/environments/python3/src/pandas/pandas/tseries/index.py in insert(self, loc, item)
   1539 
   1540         new_index = np.concatenate((self[:loc].asi8,
-> 1541                                     [item.view(np.int64)],
   1542                                     self[loc:].asi8))
   1543         return DatetimeIndex(new_index, freq='infer')

AttributeError: 'str' object has no attribute 'view'

In [50]: df = pd.DataFrame(data=np.random.rand(2, 2), columns=[1, 2], index=['A', 'B'])
In [51]: df.reset_index()
Out[51]: 
  index         1         2
0     A  0.947575  0.370406
1     B  0.664856  0.686524

[2 rows x 3 columns]

In [52]: df = pd.DataFrame(data=np.random.rand(2, 2), columns=['C', 'D'], index=['A', 'B'])

In [53]: df.reset_index()
Out[53]: 
  index         C         D
0     A  0.053455  0.599483
1     B  0.776364  0.680425

[2 rows x 3 columns]

[jtratner]

easy fix - DatetimeIndex's insert method wasn't handling non-datetimes correctly. I'll have a PR up shortly.