fix(Constant): Use object dtype with array creation (#375)

Author: eddiebergman

src/ConfigSpace/hyperparameters/hp_components.py

@@ -12,7 +12,7 @@
     normalize,
     scale,
 )
-from ConfigSpace.types import DType, f64, i64
+from ConfigSpace.types import DType, ObjectArray, f64, i64
 
 if TYPE_CHECKING:
     from ConfigSpace.types import Array, Mask
@@ -446,11 +446,14 @@ def ordinal_neighborhood(
     return np.array([seed.choice(neighbors)], dtype=f64)
 
 
+# HACK: Technically `Any` isn't an `np.number` that the Transformer expects
+# as it's type variable. However for a Constant, we can like with this typing
+# hack.
 @dataclass
-class TransformerConstant(Transformer[DType]):
+class TransformerConstant(Transformer[Any]):
     """Implementation of a transformer for a constant value."""
 
-    value: DType
+    value: Any
     """The constant value."""
 
     vector_value_yes: f64
@@ -479,23 +482,19 @@ def __post_init__(self) -> None:
         self.upper_vectorized = self.vector_value_yes
 
     @override
-    def to_vector(self, value: Array[DType]) -> Array[f64]:
+    def to_vector(self, value: ObjectArray) -> Array[f64]:
         return np.where(
             value == self.value,
             self.vector_value_yes,
             self.vector_value_no,
         )
 
     @override
-    def to_value(self, vector: Array[f64]) -> Array[DType]:
-        try:
-            return np.full_like(vector, self.value, dtype=type(self.value))
-        except TypeError:
-            # Let numpy figure it out
-            return np.array([self.value] * len(vector))
+    def to_value(self, vector: Array[f64]) -> ObjectArray:
+        return np.full_like(vector, self.value, dtype=object)
 
     @override
-    def legal_value(self, value: Array[DType]) -> Mask:
+    def legal_value(self, value: ObjectArray) -> Mask:
         return value == self.value  # type: ignore
 
     @override

src/ConfigSpace/types.py

@@ -27,6 +27,9 @@
 Array: TypeAlias = npt.NDArray[DType]
 """Array, a numpy array of a specific dtype."""
 
+ObjectArray: TypeAlias = npt.NDArray[np.object_]
+"""Object array, a numpy array of objects."""
+
 f64: TypeAlias = np.float64
 """64-bit floating point number."""