SparseDataFrame empty slice coerces to loses dtype and fill_value

[TomAugspurger]

import pandas as pd
import numpy as np

In [59]: df = pd.SparseDataFrame({"A": pd.SparseSeries(np.array([1, 1, 0], dtype='uint8'), fill_value=0), "B": pd.SparseSeries(np.array([1, 1, 0], dtype='uint8'), fill_value=0)})

In [60]: df.dtypes
Out[60]:
A    uint8
B    uint8
dtype: object

In [61]: df.A.fill_value
Out[61]: 0

In [62]: df.iloc[0:0].dtypes
Out[62]:
A    float64
B    float64
dtype: object

In [63]: df.iloc[0:0].A.fill_value
Out[63]: nan

non-empty slices are fine.

[TomAugspurger]

Although... a DataFrame of sparse series handles this fine, so maybe we just ignore this?

[jorisvandenbossche]

This is no longer a problem with the new DataFrame[sparse]:

In [39]: df = pd.DataFrame({"A": pd.SparseArray(np.array([1, 1, 0], dtype='uint8'), fill_value=0), "B": pd.SparseArray(np.array([1, 1, 0], dtype='uint8'), fill_value=0)})                                         

In [40]: df                                                                                                                                                                                                        
Out[40]: 
   A  B
0  1  1
1  1  1
2  0  0

In [41]: df.dtypes                                                                                                                                                                                                 
Out[41]: 
A    Sparse[uint8, 0]
B    Sparse[uint8, 0]
dtype: object

In [42]: df.iloc[0:0].dtypes                                                                                                                                                                                       
Out[42]: 
A    Sparse[uint8, 0]
B    Sparse[uint8, 0]
dtype: object