Get the minimum and maximum value of a Numpy MaskedArray raises AttributeError

Question

I have a masked array containing some real values and some duds. Via ipython, this is what I have set up:

In [31]: x                                                                      
Out[31]:                                                                        
masked_array(data=[--, --, --, 2.718281828459045, 3.141592653589793],           
             mask=[ True,  True,  True, False, False],                          
       fill_value='?',                                                          
            dtype=object)                                                       

Note: the reason dtype=object is because this is an actual ring_buffer containing datetimes and float values. However, for this it should not matter.

I want to find the maximum and minimum of that array. The documentation points to the .min(…) and .max(…) methods to do just that. However, when I call them, I get this exception:

In [32]: x.max()                                                                
---------------------------------------------------------------------------     
AttributeError                            Traceback (most recent call last)     
<ipython-input-32-031604a175b1> in <module>()                                   
----> 1 x.max()                                                                 

/home/usr/repos/proj/.tox/develop/lib/python2.7/site-packages/numpy/ma/core.pyc in max(self, axis, out, fill_value, keepdims) 
   5707         if out is None:                                                 
   5708             result = self.filled(fill_value).max(                       
-> 5709                 axis=axis, out=out, **kwargs).view(type(self))          
   5710             if result.ndim:                                             
   5711                 # Set the mask                                          

AttributeError: 'str' object has no attribute 'view'         

What am I doing wrong?

Answer 1

So the fact you're using dtype=object does matter. The masked array doesn't know how to properly handle these so just uses the standard numpy.array methods. The default fill value for dtype=object is '?', hence the error you get.

You should not be mixing datatypes in a numpy array. It defeats every advantage you get from them. Use lists (or pandas if dealing with different columns of different types) instead. max will fail on a numpy array of floats and datetimes anyway.

x = array([datetime.datetime(2018, 8, 28, 8, 52, 2, 107691), 3.14, 2.78],
           dtype=object)

x.max()
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-152-031604a175b1> in <module>()
----> 1 x.max()

C:\Programs\Anaconda3\lib\site-packages\numpy\core\_methods.py in _amax(a, axis, out, keepdims)
     24 # small reductions
     25 def _amax(a, axis=None, out=None, keepdims=False):
---> 26     return umr_maximum(a, axis, None, out, keepdims)
     27 
     28 def _amin(a, axis=None, out=None, keepdims=False):

TypeError: '>=' not supported between instances of 'datetime.datetime' and 'float'