Cryptic error in imputers due to missing checking in _get_mask

[glemaitre]

By working on the MissingIndicator it seems that there is a missing checking between X and missing_values dtype:

import numpy as np
X = np.array([[1.6464405 , 2.145568  , 1.80829   , 1.6346495 , 1.2709644 ],
              [1.3127615 , 2.675319  , 2.8906 , 2.1489816 , 0.8682183 ],
              [0.5495741 , 1.7595388 , 0.06032264, 2.4868202 , 0.01408643]],
             dtype=np.float32)
from sklearn.impute import SimpleImputer
trans = SimpleImputer(missing_values="NaN")
trans.fit_transform(X)

/home/lemaitre/Documents/code/toolbox/scikit-learn/sklearn/impute.py:59: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
  return np.equal(X, value_to_mask)
/home/lemaitre/Documents/code/toolbox/scikit-learn/sklearn/impute.py:59: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
  return np.equal(X, value_to_mask)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
~/miniconda3/envs/dev/lib/python3.6/site-packages/numpy/core/fromnumeric.py in _wrapfunc(obj, method, *args, **kwds)
     51     try:
---> 52         return getattr(obj, method)(*args, **kwds)
     53 

TypeError: int() argument must be a string, a bytes-like object or a number, not 'NotImplementedType'

During handling of the above exception, another exception occurred:

TypeError                                 Traceback (most recent call last)
<ipython-input-18-151141bb4b39> in <module>()
----> 1 trans.fit_transform(X)

~/Documents/code/toolbox/scikit-learn/sklearn/base.py in fit_transform(self, X, y, **fit_params)
    457         if y is None:
    458             # fit method of arity 1 (unsupervised transformation)
--> 459             return self.fit(X, **fit_params).transform(X)
    460         else:
    461             # fit method of arity 2 (supervised transformation)

~/Documents/code/toolbox/scikit-learn/sklearn/impute.py in transform(self, X)
    417             mask = _get_mask(X, self.missing_values)
    418             n_missing = np.sum(mask, axis=0)
--> 419             values = np.repeat(valid_statistics, n_missing)
    420             coordinates = np.where(mask.transpose())[::-1]
    421 

~/miniconda3/envs/dev/lib/python3.6/site-packages/numpy/core/fromnumeric.py in repeat(a, repeats, axis)
    421 
    422     """
--> 423     return _wrapfunc(a, 'repeat', repeats, axis=axis)
    424 
    425 

~/miniconda3/envs/dev/lib/python3.6/site-packages/numpy/core/fromnumeric.py in _wrapfunc(obj, method, *args, **kwds)
     60     # a downstream library like 'pandas'.
     61     except (AttributeError, TypeError):
---> 62         return _wrapit(obj, method, *args, **kwds)
     63 
     64 

~/miniconda3/envs/dev/lib/python3.6/site-packages/numpy/core/fromnumeric.py in _wrapit(obj, method, *args, **kwds)
     40     except AttributeError:
     41         wrap = None
---> 42     result = getattr(asarray(obj), method)(*args, **kwds)
     43     if wrap:
     44         if not isinstance(result, mu.ndarray):

TypeError: int() argument must be a string, a bytes-like object or a number, not 'NotImplementedType'

In short, NotImplement is raised by equal in case of numeric and string mixed dtype. We should put a check in _get_mask which is shared across classes.