Make Class Label Retrieval More Lenient

openml/datasets/dataset.py

@@ -908,8 +908,18 @@ def retrieve_class_labels(self, target_name: str = "class") -> None | list[str]:
         list
         """
         for feature in self.features.values():
-            if (feature.name == target_name) and (feature.data_type == "nominal"):
-                return feature.nominal_values
+            if feature.name == target_name:
+                if feature.data_type == "nominal":
+                    return feature.nominal_values
+
+                if feature.data_type == "string":
+                    # Rel.: #1311
+                    # The target is invalid for a classification task if the feature type is string
+                    # and not nominal. For such miss-configured tasks, we silently fix it here as
+                    # we can safely interpreter string as nominal.
+                    df, *_ = self.get_data()
+                    return list(df[feature.name].unique())
+
         return None
 
     def get_features_by_type(  # noqa: C901

tests/test_datasets/test_dataset_functions.py

@@ -626,11 +626,18 @@ def test__retrieve_class_labels(self):
         openml.config.set_root_cache_directory(self.static_cache_dir)
         labels = openml.datasets.get_dataset(2, download_data=False).retrieve_class_labels()
         assert labels == ["1", "2", "3", "4", "5", "U"]
+
         labels = openml.datasets.get_dataset(2, download_data=False).retrieve_class_labels(
             target_name="product-type",
         )
         assert labels == ["C", "H", "G"]
 
+        # Test workaround for string-typed class labels
+        custom_ds = openml.datasets.get_dataset(2, download_data=False)
+        custom_ds.features[31].data_type = "string"
+        labels = custom_ds.retrieve_class_labels(target_name=custom_ds.features[31].name)
+        assert labels == ["COIL", "SHEET"]
+
     def test_upload_dataset_with_url(self):
         dataset = OpenMLDataset(
             "%s-UploadTestWithURL" % self._get_sentinel(),