Please add a test in sklearn/datasets/tests/test_20news.py using check_return_X_y.

Please add simple test cases to ensure line coverage. If there are examples that use these fetch findings and only use data and target from them, please consider using this parameter in those examines. Thanks

you also need to update the Returns section
please add some tests

Thanks! Fetchers in the following files could also be changes,

• examples/model_selection/grid_search_text_feature_extraction.py
• examples/text/plot_document_clustering.py
• examples/text/plot_document_classification_20newsgroups.py
• examples/text/plot_hashing_vs_dict_vectorizer.py
• examples/bicluster/plot_bicluster_newsgroups.py
• examples/linear_model/plot_sparse_logistic_regression_20newsgroups.py
• examples/applications/plot_model_complexity_influence.py
• examples/plot_johnson_lindenstrauss_bound.py
• examples/ensemble/plot_forest_importances_faces.py
• examples/cluster/plot_dict_face_patches.py
  (but as you want, I'm fine with the current version as well)

I checked them and we are using target_names or other keys from the bunch and therefore we shall let the example as they are.

Fair enough. I don't think it's very critical anyway, having some examples using Bunch is also useful.

Minor nits

Whats new missing