groupby looses dtype on empty columns

[paolini]

In this snippet:

import pandas as pd
import numpy as np

print 'numpy', np.__version__
print 'pandas', pd.__version__

df = pd.DataFrame({'x': [],'range': np.arange(0)})

t = df.sort('x').groupby('x').first()

print 'types', df['range'].dtype, t['range'].dtype

it is shown that after the groupby, an empty "int" column becomes an empty "float". This is the output on my machine:

numpy 1.6.2
pandas 0.13.1
types int32 float64

[jreback]

this actually is very subtle....thanks for the report