FEAT: Implementing PyArray_ArrFuncs_nonzero

src/csrc/dtype.c

@@ -389,6 +389,25 @@ quadprec_fromstr(char *s, void *dptr, char **endptr, PyArray_Descr *descr_generi
     return 0;
 }
 
+static npy_bool
+quadprec_nonzero(void *data, void *arr)
+{
+    PyArrayObject *arr_obj = (PyArrayObject *)arr;
+    QuadPrecDTypeObject *descr = (QuadPrecDTypeObject *)PyArray_DESCR(arr_obj);
+    QuadBackendType backend = descr->backend;
+
+    if (backend == BACKEND_SLEEF) {
+        Sleef_quad val;
+        memcpy(&val, data, sizeof(Sleef_quad));
+        return !Sleef_icmpeqq1(val, QUAD_PRECISION_ZERO);
+    }
+    else {
+        long double val;
+        memcpy(&val, data, sizeof(long double));
+        return val != 0.0L;
+    }
+}
+
 static PyType_Slot QuadPrecDType_Slots[] = {
         {NPY_DT_ensure_canonical, &ensure_canonical},
         {NPY_DT_common_instance, &common_instance},
@@ -398,6 +417,7 @@ static PyType_Slot QuadPrecDType_Slots[] = {
         {NPY_DT_getitem, &quadprec_getitem},
         {NPY_DT_default_descr, &quadprec_default_descr},
         {NPY_DT_get_constant, &quadprec_get_constant},
+        {NPY_DT_PyArray_ArrFuncs_nonzero, &quadprec_nonzero},
         {NPY_DT_PyArray_ArrFuncs_fill, &quadprec_fill},
         {NPY_DT_PyArray_ArrFuncs_scanfunc, &quadprec_scanfunc},
         {NPY_DT_PyArray_ArrFuncs_fromstr, &quadprec_fromstr},

tests/test_quaddtype.py

@@ -1531,6 +1531,30 @@ def test_unary_logical_not(x):
     assert isinstance(quad_result, (bool, np.bool_)), f"Result should be bool, got {type(quad_result)}"
 
 
+@pytest.mark.parametrize("val,expected", [
+    ("1.0", True),
+    ("0.5", True),
+    ("1e-100", True),
+    ("-1.0", True),
+    ("0.0", False),
+    ("-0.0", False),
+])
+def test_bool_0d_array(val, expected):
+    """
+    Test boolean conversion on 0-d QuadPrecision arrays.
+
+    This tests that bool(np.array(QuadPrecision(x))) works correctly
+    and doesn't segfault due to missing dtype nonzero function.
+    """
+    quad_scalar = QuadPrecision(val)
+    arr_0d = np.array(quad_scalar)
+    # Ensure it's actually a 0-d array
+    assert arr_0d.ndim == 0, f"Expected 0-d array, got {arr_0d.ndim}-d"
+
+    # This should not segfault
+    result = bool(arr_0d)
+    assert result == expected, f"bool(np.array(QuadPrecision({val}))) should be {expected}, got {result}"
+
 @pytest.mark.parametrize("op", ["amin", "amax", "nanmin", "nanmax"])
 @pytest.mark.parametrize("a", ["3.0", "12.5", "100.0", "0.0", "-0.0", "inf", "-inf", "nan", "-nan"])
 @pytest.mark.parametrize("b", ["3.0", "12.5", "100.0", "0.0", "-0.0", "inf", "-inf", "nan", "-nan"])