Merge pull request #20 from ubc-systopia/sklearn-compatibility-updates

Changes to fulfill the sklearn API changes that came with version 1.6.1

Author: ilias-karimalis

src/gosdt/_binarizer.py

@@ -30,7 +30,7 @@ def _halfway_points(values):
     """
     return [(values[i] + values[i + 1]) / 2 for i in range(len(values) - 1)]
 
-class NumericBinarizer(BaseEstimator, TransformerMixin):
+class NumericBinarizer(TransformerMixin, BaseEstimator):
     f"""
     Encode numerical features as a one-hot numeric array.
 
@@ -212,8 +212,7 @@ def transform(self, X):
         # Check that the input is of the same shape as the one passed
         # during fit.
         if X.shape[1] != self.n_features_in_:
-            raise ValueError('Shape of input is different form what was seen'
-                             'in `fit`')
+            raise ValueError(f"X has {X.shape[1]} features, but NumericBinarizer is expecting {self.n_features_in_} features as input")
 
         # Return a binarization of the input samples based on halfway point splits
         Xb_lists = [[X[:, i] <= val for val in _halfway_points(self.column_values_[i])] for i in range(self.n_features_in_)]
@@ -243,8 +242,7 @@ def inverse_transform(self, Xt):
         # Check that the input is of the same shape as the one passed
         # during fit.
         if Xt.shape[1] != self.n_features_out_:
-            raise ValueError('Shape of input is different from what was seen '
-                             'in `fit`')
+            raise ValueError(f"X has {Xt.shape[1]} features, but NumericBinarizer.inverse_transform is expecting {self.n_features_out_} features as input")
 
         # Initialize an empty array for inverse transformed data
         X = np.empty((Xt.shape[0], self.n_features_in_))

src/gosdt/_threshold_guessing.py

@@ -6,7 +6,7 @@
 from sklearn.ensemble import GradientBoostingClassifier
 from sklearn.utils.validation import check_array, check_is_fitted
 
-class ThresholdGuessBinarizer(BaseEstimator, TransformerMixin):
+class ThresholdGuessBinarizer(TransformerMixin, BaseEstimator):
     f"""
     Encode numerical features as a one-hot numeric array. 
     The encoding is based on the thresholds found by a Gradient Boosting Classifier.
@@ -216,8 +216,7 @@ def transform(self, X):
         # Check that the input is of the same shape as the one passed
         # during fit.
         if X.shape[1] != self.n_features_in_:
-            raise ValueError('Shape of input is different from what was seen'
-                             'in `fit`')
+            raise ValueError(f"X has {X.shape[1]} features, but ThresholdBinarizer is expecting {self.n_features_in_} features as input")
 
         return np.concatenate([np.atleast_2d(X[:, j] <= thresh).T for j, thresh in self.thresholds_], axis=1).astype(float)