Fix mypy errors. (#10907)

python-package/xgboost/_typing.py

@@ -5,6 +5,7 @@
 from typing import (
     TYPE_CHECKING,
     Any,
+    AnyStr,
     Callable,
     Dict,
     List,
@@ -28,7 +29,10 @@
 BoosterParam = Union[List, Dict[str, Any]]  # better be sequence
 
 ArrayLike = Any
-PathLike = Union[str, os.PathLike]
+if TYPE_CHECKING:
+    PathLike = Union[str, os.PathLike[str]]
+else:
+    PathLike = Union[str, os.PathLike]
 CupyT = ArrayLike  # maybe need a stub for cupy arrays
 NumpyOrCupy = Any
 NumpyDType = Union[str, Type[np.number]]  # pylint: disable=invalid-name
@@ -48,7 +52,7 @@
 # c_bst_ulong corresponds to bst_ulong defined in xgboost/c_api.h
 c_bst_ulong = ctypes.c_uint64  # pylint: disable=C0103
 
-ModelIn = Union[str, bytearray, os.PathLike]
+ModelIn = Union[os.PathLike[AnyStr], bytearray, str]
 
 CTypeT = TypeVar(
     "CTypeT",

python-package/xgboost/core.py

@@ -26,6 +26,7 @@
     Sequence,
     Tuple,
     Type,
+    TypeGuard,
     TypeVar,
     Union,
     cast,
@@ -53,6 +54,7 @@
     IterationRange,
     ModelIn,
     NumpyOrCupy,
+    PathLike,
     TransformedData,
     c_bst_ulong,
 )
@@ -1112,7 +1114,7 @@ def set_uint_info(self, field: str, data: ArrayLike) -> None:
 
         dispatch_meta_backend(self, data, field, "uint32")
 
-    def save_binary(self, fname: Union[str, os.PathLike], silent: bool = True) -> None:
+    def save_binary(self, fname: PathLike, silent: bool = True) -> None:
         """Save DMatrix to an XGBoost buffer.  Saved binary can be later loaded
         by providing the path to :py:func:`xgboost.DMatrix` as input.
 
@@ -2735,7 +2737,7 @@ def inplace_predict(
             "Data type:" + str(type(data)) + " not supported by inplace prediction."
         )
 
-    def save_model(self, fname: Union[str, os.PathLike]) -> None:
+    def save_model(self, fname: PathLike) -> None:
         """Save the model to a file.
 
         The model is saved in an XGBoost internal format which is universal among the
@@ -2808,9 +2810,12 @@ def load_model(self, fname: ModelIn) -> None:
             Input file name or memory buffer(see also save_raw)
 
         """
-        if isinstance(fname, (str, os.PathLike)):
-            # assume file name, cannot use os.path.exist to check, file can be
-            # from URL.
+
+        def is_pathlike(path: ModelIn) -> TypeGuard[os.PathLike[str]]:
+            return isinstance(path, os.PathLike)
+
+        if isinstance(fname, str) or is_pathlike(fname):
+            # assume file name, cannot use os.path.exist to check, file can be from URL.
             fname = os.fspath(os.path.expanduser(fname))
             _check_call(_LIB.XGBoosterLoadModel(self.handle, c_str(fname)))
         elif isinstance(fname, bytearray):
@@ -2870,8 +2875,8 @@ def num_features(self) -> int:
 
     def dump_model(
         self,
-        fout: Union[str, os.PathLike],
-        fmap: Union[str, os.PathLike] = "",
+        fout: PathLike,
+        fmap: PathLike = "",
         with_stats: bool = False,
         dump_format: str = "text",
     ) -> None:
@@ -2915,7 +2920,7 @@ def dump_model(
 
     def get_dump(
         self,
-        fmap: Union[str, os.PathLike] = "",
+        fmap: PathLike = "",
         with_stats: bool = False,
         dump_format: str = "text",
     ) -> List[str]:
@@ -2968,7 +2973,7 @@ def get_fscore(
         return self.get_score(fmap, importance_type="weight")
 
     def get_score(
-        self, fmap: Union[str, os.PathLike] = "", importance_type: str = "weight"
+        self, fmap: PathLike = "", importance_type: str = "weight"
     ) -> Dict[str, Union[float, List[float]]]:
         """Get feature importance of each feature.
         For tree model Importance type can be defined as:
@@ -3033,7 +3038,7 @@ def get_score(
         return results
 
     # pylint: disable=too-many-statements
-    def trees_to_dataframe(self, fmap: Union[str, os.PathLike] = "") -> DataFrame:
+    def trees_to_dataframe(self, fmap: PathLike = "") -> DataFrame:
         """Parse a boosted tree model text dump into a pandas DataFrame structure.
 
         This feature is only defined when the decision tree model is chosen as base
@@ -3199,7 +3204,7 @@ def _validate_features(self, feature_names: Optional[FeatureNames]) -> None:
     def get_split_value_histogram(
         self,
         feature: str,
-        fmap: Union[os.PathLike, str] = "",
+        fmap: PathLike = "",
         bins: Optional[int] = None,
         as_pandas: bool = True,
     ) -> Union[np.ndarray, DataFrame]:

python-package/xgboost/data.py

@@ -5,17 +5,7 @@
 import json
 import os
 import warnings
-from typing import (
-    Any,
-    Callable,
-    List,
-    Optional,
-    Sequence,
-    Tuple,
-    TypeGuard,
-    Union,
-    cast,
-)
+from typing import Any, Callable, List, Optional, Sequence, Tuple, TypeGuard, cast
 
 import numpy as np
 
@@ -27,6 +17,7 @@
     FloatCompatible,
     NumpyDType,
     PandasDType,
+    PathLike,
     TransformedData,
     c_bst_ulong,
 )
@@ -1078,12 +1069,12 @@ def _from_dlpack(
     return _from_cupy_array(data, missing, nthread, feature_names, feature_types)
 
 
-def _is_uri(data: DataType) -> TypeGuard[Union[str, os.PathLike]]:
+def _is_uri(data: DataType) -> TypeGuard[PathLike]:
     return isinstance(data, (str, os.PathLike))
 
 
 def _from_uri(
-    data: Union[str, os.PathLike],
+    data: PathLike,
     missing: Optional[FloatCompatible],
     feature_names: Optional[FeatureNames],
     feature_types: Optional[FeatureTypes],

python-package/xgboost/testing/__init__.py

@@ -49,6 +49,8 @@
     memory,
 )
 
+from .._typing import PathLike
+
 hypothesis = pytest.importorskip("hypothesis")
 
 # pylint:disable=wrong-import-position,wrong-import-order
@@ -746,7 +748,7 @@ class DirectoryExcursion:
 
     """
 
-    def __init__(self, path: Union[os.PathLike, str], cleanup: bool = False):
+    def __init__(self, path: PathLike, cleanup: bool = False):
         self.path = path
         self.curdir = os.path.normpath(os.path.abspath(os.path.curdir))
         self.cleanup = cleanup