Fix casting of boolean/unsigned integers to complex in C-backend

pytensor/scalar/basic.py

@@ -511,6 +511,10 @@ class ScalarType(CType, HasDataType, HasShape):
                 "npy_int16",
                 "npy_int32",
                 "npy_int64",
+                "npy_uint8",  # also covers npy_bool
+                "npy_uint16",
+                "npy_uint32",
+                "npy_uint64",
                 "npy_float32",
                 "npy_float64",
             ]

tests/scalar/test_basic.py

@@ -3,7 +3,7 @@ import pytest
 
 import pytensor
 import pytensor.tensor as pt
-from pytensor.compile.mode import Mode
+from pytensor.compile.mode import Mode, get_default_mode
 from pytensor.graph.fg import FunctionGraph
 from pytensor.link.c.basic import DualLinker
 from pytensor.link.numba import NumbaLinker
@@ -15,6 +15,7 @@ from pytensor.scalar.basic import (
     ScalarType,
     TrueDiv,
     add,
+    all_types,
     and_,
     arccos,
     arccosh,
@@ -55,6 +56,7 @@ from pytensor.scalar.basic import (
     tanh,
     true_div,
 )
+from pytensor.tensor import tensor_from_scalar
 from pytensor.tensor.type import fscalar, imatrix, matrix
 from tests.link.test_link import make_function
 
@@ -515,3 +517,16 @@ def test_rfloordiv():
     assert isinstance(y.owner.op, IntDiv)
     assert isinstance(y.type, ScalarType)
     assert y.eval({x: 2.0}) == 2.0
+
+
+@pytest.mark.parametrize("inp_type", all_types, ids=lambda x: x.dtype)
+def test_cast_to_complex(inp_type):
+    if inp_type.dtype == "float16":
+        if isinstance(get_default_mode().linker, NumbaLinker):
+            pytest.skip("Numba doesn't support float16")
+    x = inp_type("x")
+    # Output as tensor to sidestep numba issue with numpy scalar outputs
+    y = tensor_from_scalar(x.astype("complex64"))
+    res_y = y.eval({x: np.array(1.0, dtype=inp_type.dtype)})
+    assert res_y == 1
+    assert res_y.dtype == "complex64"