Enable categorical feature support by default.

[trivialfis]

close #12014

I kept the XGBoost parameter to avoid a breaking change to the API, and users can still disable categorical feature support if they wish.

[Copilot]

Pull request overview

This PR enables categorical feature support by default in XGBoost by changing the default value of the enable_categorical parameter from False to True. This aligns with the goal stated in issue #12014 to make categorical support the default behavior while maintaining backward compatibility by keeping the parameter available for users who wish to disable it.

Changes:

• Changed default value of enable_categorical from False to True in core DMatrix classes and sklearn estimators
• Removed explicit enable_categorical=True from test code throughout the codebase
• Removed redundant Python-side validation for categorical features with exact tree method (now handled in C++)
• Applied code formatting improvements including import ordering standardization

Reviewed changes

Copilot reviewed 17 out of 17 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
python-package/xgboost/core.py	Changed default value of enable_categorical to True in DMatrix, QuantileDMatrix, and ExtMemQuantileDMatrix classes; improved docstring formatting
python-package/xgboost/sklearn.py	Changed default value of enable_categorical to True in sklearn estimator base class; removed redundant validation logic for exact tree method with categorical features; improved docstring formatting
tests/test_distributed/test_with_dask/test_with_dask.py	Removed explicit enable_categorical=True from DaskDMatrix and DaskXGBRegressor instantiations; improved import ordering
tests/test_distributed/test_gpu_with_dask/test_gpu_with_dask.py	Removed explicit enable_categorical=True from DaskQuantileDMatrix instantiations; improved import ordering
tests/python/test_with_sklearn.py	Removed explicit enable_categorical=True from XGBRegressor instantiations; applied extensive code formatting improvements including quote normalization and line breaking
tests/python/test_with_polars.py	Removed explicit enable_categorical=True from DMatrix and QuantileDMatrix instantiations; removed test checking for error when enable_categorical not set; improved import ordering
tests/python/test_with_pandas.py	Removed explicit enable_categorical=True from DMatrix instantiations; removed test checking for error when enable_categorical not set; improved import ordering
tests/python/test_quantile_dmatrix.py	Removed explicit enable_categorical=True from XGBRegressor, DMatrix, and QuantileDMatrix instantiations; improved import ordering
tests/python/test_predict.py	Removed explicit enable_categorical=True from DMatrix instantiation; improved import ordering
tests/python/test_model_io.py	Removed explicit enable_categorical=True from DMatrix instantiation; improved code formatting
tests/python/test_demos.py	Removed explicit enable_categorical=True from XGBRegressor instantiation
tests/python/test_data_iterator.py	Removed explicit enable_categorical=True from DMatrix and QuantileDMatrix instantiations; improved import ordering
tests/python/generate_models.py	Removed explicit enable_categorical=True from DMatrix, XGBRegressor, and QuantileDMatrix instantiations; improved import ordering
tests/python-gpu/test_gpu_with_sklearn.py	Removed explicit enable_categorical=True from XGBClassifier and XGBRegressor instantiations; improved import ordering
tests/python-gpu/test_gpu_prediction.py	Removed explicit enable_categorical=True from DMatrix instantiations; improved import ordering
tests/python-gpu/test_gpu_ordinal.py	Removed explicit enable_categorical=True from DMatrix instantiations; improved import ordering
tests/python-gpu/test_from_cudf.py	Removed explicit enable_categorical=True from DMatrix and QuantileDMatrix instantiations; removed tests checking for error when enable_categorical not set; improved import ordering

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[trivialfis]

I don't know how to fix the pylint error without major refactoring.

[RAMitchell]

@trivialfis let me have a go in another PR.