While using sklearn.preprocessing.OneHotEncoder, I encountered confusion regarding the use of sparse vs sparse_output. The documentation mentions that sparse is deprecated in favor of sparse_output since v1.2, but in practice, both parameters can still exist in some environments, and it's unclear which one takes precedence when both are passed.

from sklearn.preprocessing import OneHotEncoder

encoder = OneHotEncoder(sparse=False)  # No warning?

encoder = OneHotEncoder(sparse_output=False)  # Recommended?


encoder = OneHotEncoder(sparse=False, sparse_output=False)