[MRG] FIX: raise error with inconsistent dtype X and missing_values

sklearn/impute.py

@@ -40,6 +40,17 @@
 ]
 
 
+def _check_inputs_dtype(X, missing_values):
+    """Check that the dtype of X is in accordance with the one of
+    missing_values."""
+    if (X.dtype.kind in ("f", "i", "u") and
+            not isinstance(missing_values, numbers.Real)):
+        raise TypeError("The data type of 'missing_values' and 'X' are "
+                        "not compatible. 'missing_values' data type is "
+                        "{} and 'X' is {}."
+                        .format(type(missing_values), X.dtype))
+
+
 def _get_mask(X, value_to_mask):
     """Compute the boolean mask X == missing_values."""
     if value_to_mask is np.nan:
@@ -51,7 +62,6 @@ def _get_mask(X, value_to_mask):
         else:
             # np.isnan does not work on object dtypes.
             return _object_dtype_isnan(X)
-
     else:
         # X == value_to_mask with object dytpes does not always perform
         # element-wise for old versions of numpy
@@ -183,6 +193,7 @@ def _validate_input(self, X):
             else:
                 raise ve
 
+        _check_inputs_dtype(X, self.missing_values)
         if X.dtype.kind not in ("i", "u", "f", "O"):
             raise ValueError("SimpleImputer does not support data with dtype "
                              "{0}. Please provide either a numeric array (with"
@@ -788,6 +799,7 @@ def _initial_imputation(self, X):
 
         X = check_array(X, dtype=FLOAT_DTYPES, order="F",
                         force_all_finite=force_all_finite)
+        _check_inputs_dtype(X, self.missing_values)
 
         mask_missing_values = _get_mask(X, self.missing_values)
         if self.initial_imputer_ is None:

sklearn/tests/test_impute.py

@@ -705,3 +705,23 @@ def test_chained_imputer_additive_matrix():
                              random_state=rng).fit(X_train)
     X_test_est = imputer.transform(X_test)
     assert_allclose(X_test_filled, X_test_est, atol=0.01)
+
+
+@pytest.mark.parametrize("imputer_constructor",
+                         [SimpleImputer, ChainedImputer])
+@pytest.mark.parametrize(
+    "missing_values, X_missing_value, err_type, err_msg",
+    [("NaN", np.nan, ValueError, "contains"),
+     ("-1", -1, TypeError, "not compatible")])
+def test_inconsistent_dtype_X_missing_values(imputer_constructor,
+                                             missing_values, X_missing_value,
+                                             err_type, err_msg):
+    # regression test for issue #11390. Comparison between incoherent dtype
+    # for X and missing_values was not raising a proper error.
+    X = np.random.randn(1000, 10)
+    X[0, 0] = X_missing_value
+
+    imputer = imputer_constructor(missing_values=missing_values)
+
+    with pytest.raises(err_type, match=err_msg):
+        imputer.fit_transform(X)