…e vs str)This patch improves `_most_frequent` to gracefully handle arrays with mixed, incomparable types (e.g. `None`, `str`, `int`), which previously raised `TypeError` due to Python's `min()` not supporting heterogeneous comparisons.Fix:- Uses `type` + `hash` as a tie-breaking key to avoid direct comparison between incompatible types.- Keeps behaviour consistent with `scipy.stats.mode` tie-breaking logic.- Maintains deterministic and efficient output.Adds unit tests for edge cases:- Tie between `None` and `str`- Dominant `None`- Empty array handling- Tie involving extra_valueCloses:scikit-learn#31717

This update improves the _most_frequent function to gracefully handle arrays containing mixed, incomparable types (e.g., None, str, int). The patch uses a deterministic tie-breaking strategy based on type and hash to avoid TypeErrors, consistent with scipy.stats.mode behaviour.

…lintingThis update improves the _most_frequent function by:Ensuring safe, deterministic tie-breaking for mixed, incomparable types (e.g., None, str, int) using (type, hash) as a key.Fixing docstring formatting and style issues to meet project linting requirements.Aligning logic with scipy.stats.mode tie-breaking behavior.Improving code readability and PEP-8 compliance.

Ensure safe comparison in _most_frequent by using (str(type(x)), hash(x)) as a tie-breaking key.- Handles mixed types like None, str, int.- Prevents TypeError from Python's min() on incomparable values.- Aligns with scipy.stats.mode behaviour.- Ready for review and linted locally.

…akingThis patch improves the _most_frequent function to safely process arrays with mixed, incomparable types (e.g., None, str, int) without raising TypeError.

0xs1d left a comment