FIX - fixing get_feature_names_out in DatetimeEncoder

CHANGES.rst

@@ -47,7 +47,8 @@ Bugfixes
   (not into their subclasses). If you need the items evaluated (ie if they
   contain DataOps or Choices), store them in one of the builtin collections.
   :pr:`1612` by :user:`Jérôme Dockès <jeromedockes>`.
-
+- Fixed a bug that happened when ``get_feature_names_out`` was called on instances
+  of the :class:`DatetimeEncoder`. :pr:`1622` by :user:`Riccardo Cappuzzo<rcap107>`.
 
 Release 0.6.1
 ===================

skrub/_datetime_encoder.py

@@ -434,7 +434,7 @@ def transform(self, column):
         transformed : DataFrame
             The extracted features.
         """
-        check_is_fitted(self, "extracted_features_")
+        check_is_fitted(self, "all_outputs_")
         name = sbd.name(column)
 
         # Checking again which values are null if calling only transform
@@ -460,6 +460,8 @@ def transform(self, column):
         X_out = sbd.copy_index(column, sbd.make_dataframe_like(column, all_extracted))
         X_out = sbd.concat(X_out, *new_features, axis=1)
 
+        self.all_outputs_ = sbd.column_names(X_out)
+
         # Censoring all the null features
         X_out = sbd.where_row(X_out, not_nulls, null_mask)
 
@@ -492,6 +494,22 @@ def __sklearn_tags__(self):
         tags.transformer_tags = TransformerTags(preserves_dtype=[])
         return tags
 
+    def get_feature_names_out(self, input_features=None):
+        """Get output feature names for transformation.
+
+        Parameters
+        ----------
+        input_features : array-like of str or None, default=None
+            Ignored.
+
+        Returns
+        -------
+        feature_names_out : ndarray of str objects
+            Transformed feature names.
+        """
+        check_is_fitted(self, "all_outputs_")
+        return self.all_outputs_
+
 
 class _SplineEncoder(SingleColumnTransformer):
     """Generate univariate B-spline bases for features.

skrub/tests/test_datetime_encoder.py

@@ -299,6 +299,7 @@ def test_all_outputs_choice(datetime_cols, params, all_outputs):
     enc = DatetimeEncoder(**params)
     res = enc.fit_transform(datetime_cols.datetime)
     assert enc.all_outputs_ == [f"when_{f}" for f in all_outputs]
+    assert enc.get_feature_names_out() == [f"when_{f}" for f in all_outputs]
     assert sbd.column_names(res) == [f"{f}" for f in enc.all_outputs_]