Imputations run on the reference of the value

[eltonlaw]

>>> random_normal(shape=(100, 100))
array([[        nan,         nan,         nan, ...,  2.29665783,            
                nan, -0.71412759],    
       [ 0.49046767, -0.31572648,         nan, ...,  1.39596078,            
        -1.14295358, -1.30660838],    
       [        nan, -2.0708818 ,  0.7393569 , ..., -0.52974738,            
         0.24717896, -2.37030327],    
       ...,        
       [ 0.01155974,  0.88793848, -0.04410631, ..., -1.28196955,            
         0.75566477, -0.39039914],    
       [ 0.23240304,         nan,  1.59899984, ..., -1.06248365,            
                nan,  0.65453688],    
       [-0.13855768, -0.00358682,         nan, ...,  1.29588659,            
        -0.20579175,  0.59610582]])   
>>> data = random_normal(shape=(100, 100))                                  
>>> import impyute as impy            
>>> impy.em(data)  
array([[ 0.5599156 , -0.24410474, -0.99875721, ..., -0.74595691,            
         0.25954462,  0.3936289 ],    
       [-0.5491675 ,  0.39810825,  0.15029102, ..., -0.99765863,            
        -0.98604735,  1.24321062],    
       [ 0.36389712,  1.56754062,  1.38492368, ..., -0.04457599,            
        -0.12098783,  0.98864098],    
       ...,        
       [ 1.37199931, -0.45710982, -1.30196092, ..., -0.38020366,            
         0.31780175, -0.08301059],    
       [ 0.52415   , -1.02749075,  2.03909177, ...,  0.66138282,            
         1.31679312, -0.41575647],    
       [-0.36272847,  0.65262579, -0.11336795, ..., -0.1538307 ,            
        -1.24756562, -0.27470951]])   
>>> impy.em(data)  
Traceback (most recent call last):    
  File "<stdin>", line 1, in <module> 
  File "/Users/elton.law/sandbox/github/impyute/impyute/utils/checks.py", line 34, in wrapper
    raise BadInputError("No NaN's in given data")                           
impyute.utils.errors.BadInputError: No NaN's in given data                  
>>> 

Need to make a copy and run each algorithm on that instead. This can be very expensive for big datasets. Keep an inplace keyword like in pandas so that we can use both behaviours