idxmax and idxmin do not work with strings

[cancan101]

Against master:

In [8]: pandas.DataFrame({'A':["A","B","C", "A"], 'B':list("ABCD")}).groupby("A").B.idxmax()
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-8-6aa5f42cc358> in <module>()
----> 1 pandas.DataFrame({'A':["A","B","C", "A"], 'B':list("ABCD")}).groupby("A").B.idxmax()

/home/apprun/gitrepos/pandas/pandas/core/groupby.py in wrapper(*args, **kwargs)
    341                 return self.apply(curried_with_axis)
    342             except Exception:
--> 343                 return self.apply(curried)
    344
    345         return wrapper

/home/apprun/gitrepos/pandas/pandas/core/groupby.py in apply(self, func, *args, **kwargs)
    422             return func(g, *args, **kwargs)
    423
--> 424         return self._python_apply_general(f)
    425
    426     def _python_apply_general(self, f):

/home/apprun/gitrepos/pandas/pandas/core/groupby.py in _python_apply_general(self, f)
    425
    426     def _python_apply_general(self, f):
--> 427         keys, values, mutated = self.grouper.apply(f, self.obj, self.axis)
    428
    429         return self._wrap_applied_output(keys, values,

/home/apprun/gitrepos/pandas/pandas/core/groupby.py in apply(self, f, data, axis)
    881             # group might be modified
    882             group_axes = _get_axes(group)
--> 883             res = f(group)
    884             if not _is_indexed_like(res, group_axes):
    885                 mutated = True

/home/apprun/gitrepos/pandas/pandas/core/groupby.py in f(g)
    420         @wraps(func)
    421         def f(g):
--> 422             return func(g, *args, **kwargs)
    423
    424         return self._python_apply_general(f)

/home/apprun/gitrepos/pandas/pandas/core/groupby.py in curried(x)
    327
    328             def curried(x):
--> 329                 return f(x, *args, **kwargs)
    330
    331             # preserve the name so we can detect it when calling plot methods,

/home/apprun/gitrepos/pandas/pandas/core/series.py in idxmax(self, axis, out, skipna)
   1284         DataFrame.idxmax
   1285         """
-> 1286         i = nanops.nanargmax(_values_from_object(self), skipna=skipna)
   1287         if i == -1:
   1288             return pa.NA

/home/apprun/gitrepos/pandas/pandas/core/nanops.py in nanargmax(values, axis, skipna)
    387     """
    388     values, mask, dtype = _get_values(values, skipna, fill_value_typ='-inf',
--> 389                                       isfinite=True)
    390     result = values.argmax(axis)
    391     result = _maybe_arg_null_out(result, axis, mask, skipna)

/home/apprun/gitrepos/pandas/pandas/core/nanops.py in _get_values(values, skipna, fill_value, fill_value_typ, isfinite, copy)
    139     values = _values_from_object(values)
    140     if isfinite:
--> 141         mask = _isfinite(values)
    142     else:
    143         mask = isnull(values)

/home/apprun/gitrepos/pandas/pandas/core/nanops.py in _isfinite(values)
    173         return isnull(values)
    174     elif isinstance(values.dtype, object):
--> 175         return -np.isfinite(values.astype('float64'))
    176
    177     return -np.isfinite(values)

ValueError: could not convert string to float: D

but this works:

In [9]: pandas.DataFrame({'A':["A","B","C", "A"], 'B':list("ABCD")}).groupby("A").B.max()
Out[9]:
A
A    D
B    B
C    C
Name: B, dtype: object

[cancan101]

This seems related: #4279 although that issue seems to indicate a numpy problem which this does not seem to be,

[jreback]

its the same issue; it could be made to work

[jreback]

closing this and I'll put a reference