TST: cover reflected scalar operators

array_api_tests/test_operators_and_elementwise_functions.py

@@ -2104,6 +2104,59 @@ def test_subtract(ctx, data):
         raise
 
 
+def make_reflected_scalar_params():
+    cases = (
+        ("__radd__", "add", operator.add, dh.numeric_dtypes),
+        ("__rsub__", "subtract", operator.sub, dh.numeric_dtypes),
+        ("__rmul__", "multiply", operator.mul, dh.numeric_dtypes),
+        ("__rtruediv__", "divide", operator.truediv, dh.all_float_dtypes),
+        ("__rfloordiv__", "floor_divide", operator.floordiv, dh.real_dtypes),
+        ("__rpow__", "pow", operator.pow, dh.numeric_dtypes),
+        ("__rmod__", "remainder", operator.mod, dh.real_dtypes),
+        ("__rand__", "bitwise_and", operator.and_, dh.bool_and_all_int_dtypes),
+        ("__ror__", "bitwise_or", operator.or_, dh.bool_and_all_int_dtypes),
+        ("__rxor__", "bitwise_xor", operator.xor, dh.bool_and_all_int_dtypes),
+        ("__rlshift__", "bitwise_left_shift", operator.lshift, dh.all_int_dtypes),
+        ("__rrshift__", "bitwise_right_shift", operator.rshift, dh.all_int_dtypes),
+    )
+    for method, func_name, op, dtypes in cases:
+        for dtype in dtypes:
+            if isinstance(dtype, xp._UndefinedStub):
+                continue
+            yield pytest.param(
+                method, func_name, op, dtype,
+                id=f"{method}-{dh.dtype_to_name[dtype]}",
+            )
+
+
+@pytest.mark.parametrize(
+    "method, func_name, op, dtype", list(make_reflected_scalar_params())
+)
+@given(data=st.data())
+def test_reflected_scalar_operator(method, func_name, op, dtype, data):
+    # Small positive values avoid division by zero, invalid powers, and overflow.
+    if dtype == xp.bool:
+        elements = st.booleans()
+    else:
+        elements = st.integers(1, 3).map(dh.get_scalar_type(dtype))
+    scalar_strategy = elements
+    if func_name == "subtract" and dtype in dh.uint_dtypes:
+        scalar_strategy = st.integers(3, 7)
+    scalar = data.draw(scalar_strategy, label="scalar")
+    x = data.draw(
+        hh.arrays(
+            dtype=dtype, shape=hh.shapes(min_dims=1, min_side=1), elements=elements
+        ),
+        label="x",
+    )
+    out = op(scalar, x)
+    expected = getattr(xp, func_name)(xp.asarray(scalar, dtype=dtype), x)
+
+    ph.assert_dtype(method, in_dtype=dtype, out_dtype=out.dtype)
+    ph.assert_result_shape(method, in_shapes=[x.shape], out_shape=out.shape)
+    ph.assert_array_elements(method, out=out, expected=expected)
+
+
 @given(hh.arrays(dtype=hh.all_floating_dtypes(), shape=hh.shapes()))
 def test_tan(x):
     repro_snippet = ph.format_snippet(f"xp.tan({x!r})")