Behavior of mapper with no transformers

[micahjsmith]

A DataFrameMapper can be initialized and fit with an empty transformer list without errors. But then an error is thrown when transformation is attempted.

• I would expect (and would find it convenient) that a null mapper would be acceptable, and would transform any X to a np.empty((np.size(X, 0), 0)) or to None. Otherwise this ends up being a fairly annoying special case in application logic.
• Alternately, if the desired behavior is to throw an error in this situation, it seems that the error is "destined" as soon as the DataFrameMapper is initialized with empty transformer list. So might as well throw the error right away.

In [1]: import pandas as pd
   ...: from sklearn_pandas import DataFrameMapper
   ...:                                                                    
   ...:                                                                    
                                                                                                                                                      
In [2]: X_df = pd.util.testing.makeDataFrame()     
   ...: mapper = DataFrameMapper([], input_df=True)
   ...:                                           
   ...: 
                               
In [3]: mapper.fit(X_df)
Out[3]:                                                                                                                                               
DataFrameMapper(default=False, df_out=False, features=[], input_df=True,
        sparse=False)
                                               
In [4]: mapper.transform(X_df)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-4-0de258be511f> in <module>()      
----> 1 mapper.transform(X_df)
         
/usr/local/anaconda3/lib/python3.7/site-packages/sklearn_pandas/dataframe_mapper.py in transform(self, X)
    326                 stacked = stacked.toarray()
    327         else:
--> 328             stacked = np.hstack(extracted)
    329
    330         if self.df_out:

/usr/local/anaconda3/lib/python3.7/site-packages/numpy/core/shape_base.py in hstack(tup)
    286         return _nx.concatenate(arrs, 0)
    287     else:
--> 288         return _nx.concatenate(arrs, 1)
    289
    290

ValueError: need at least one array to concatenate