Require Enum from polars.

python-package/xgboost/data.py

@@ -937,6 +937,18 @@ def _check_pyarrow_for_polars() -> None:
         raise ImportError("`pyarrow` is required for polars.")
 
 
+def _reject_polars_categorical(data: DataType) -> None:
+    pl = import_polars()
+
+    for name, dtype in zip(data.columns, data.dtypes):
+        if isinstance(dtype, pl.Categorical):
+            raise ValueError(
+                "XGBoost does not support `polars.Categorical` because its "
+                "encoding can be sparse. Use `polars.Enum` instead. "
+                f"Invalid column: {name}",
+            )
+
+
 def _transform_polars_df(
     data: DataType,
     enable_categorical: bool,
@@ -954,6 +966,7 @@ def _transform_polars_df(
         df = data
 
     _check_pyarrow_for_polars()
+    _reject_polars_categorical(df)
     table = df.to_arrow()
     return _transform_arrow_table(
         table, enable_categorical, feature_names, feature_types

src/common/hist_util.h

@@ -1,5 +1,5 @@
 /**
- * Copyright 2017-2024, XGBoost Contributors
+ * Copyright 2017-2026, XGBoost Contributors
  * \file hist_util.h
  * \brief Utility for fast histogram aggregation
  * \author Philip Cho, Tianqi Chen
@@ -8,7 +8,7 @@
 #define XGBOOST_COMMON_HIST_UTIL_H_
 
 #include <algorithm>
-#include <cmath>
+#include <cmath>    // for nextafter
 #include <cstdint>  // for uint32_t
 #include <limits>
 #include <map>

tests/python/test_with_polars.py

@@ -1,4 +1,4 @@
-"""Copyright 2024, XGBoost contributors"""
+"""Copyright 2024-2026, XGBoost contributors"""
 
 import json
 from pathlib import Path
@@ -38,15 +38,16 @@ def test_polars_basic(
     if isinstance(Xy, xgb.QuantileDMatrix):
         # skip min values in the cut.
         np.testing.assert_allclose(res[1:, :], res1[1:, :])
+        m = np.nextafter(np.float32(1), -np.inf)  # min_val
     else:
         np.testing.assert_allclose(res, res1)
+        m = np.float32(0)
 
     # boolean
     df = pl.DataFrame({"a": [True, False, False], "b": [False, False, True]})
     Xy = DMatrixT(df)
-    np.testing.assert_allclose(
-        Xy.get_data().data, np.array([1, 0, 0, 0, 0, 1]), atol=1e-5
-    )
+
+    np.testing.assert_equal(Xy.get_data().data, np.array([1, m, m, m, m, 1]))
 
 
 def test_polars_missing() -> None:
@@ -140,25 +141,23 @@ def test_regressor() -> None:
 
 
 def test_categorical() -> None:
-    import polars as pl
-
     cats = ["aa", "cc", "bb", "ee", "ee"]
     df = pl.DataFrame(
         {"f0": [1, 3, 2, 4, 4], "f1": cats},
-        schema=[("f0", pl.Int64()), ("f1", pl.Categorical(ordering="lexical"))],
+        schema=[("f0", pl.Int64()), ("f1", pl.Categorical())],
     )
 
-    data = xgb.DMatrix(df)
-    categories = data.get_categories(export_to_arrow=True)
-    assert dict(categories.to_arrow())["f0"] is None
-    f1 = dict(categories.to_arrow())["f1"]
-    assert f1 is not None
-    assert f1.to_pylist() == cats[:4]
+    with pytest.raises(ValueError, match="polars.Categorical.*polars.Enum"):
+        xgb.DMatrix(df)
 
     df = pl.DataFrame(
         {"f0": [1, 3, 2, 4, 4], "f1": cats},
         schema=[("f0", pl.Int64()), ("f1", pl.Enum(cats[:4]))],
     )
+    arr = df["f1"].to_arrow()
+    assert arr.dictionary.to_pylist() == cats[:4]
+    assert arr.indices.to_pylist() == [0, 1, 2, 3, 3]
+
     data = xgb.DMatrix(df)
     categories = data.get_categories(export_to_arrow=True)
     assert dict(categories.to_arrow())["f0"] is None