Optimize vector-number math by avoiding RealNumber_Check

PyFloat_AsDouble does basically the same checks internally, so since this is the last case checked in the generic_math functions in math.c, we can just try to convert directly and handle if if the conversion fails.

Author: Starbuck5

src_c/math.c

@@ -688,7 +688,7 @@ vector_generic_math(PyObject *o1, PyObject *o2, int op)
     Py_ssize_t i, dim;
     double *vec_coords;
     double other_coords[VECTOR_MAX_SIZE] = {0};
-    double tmp;
+    double tmp = 0.0;
     PyObject *other;
     pgVector *vec, *ret = NULL;
     if (pgVector_Check(o1)) {
@@ -712,11 +712,15 @@ vector_generic_math(PyObject *o1, PyObject *o2, int op)
     if (pg_VectorCoordsFromObj(other, dim, other_coords)) {
         op |= OP_ARG_VECTOR;
     }
-    else if (RealNumber_Check(other)) {
-        op |= OP_ARG_NUMBER;
-    }
     else {
-        op |= OP_ARG_UNKNOWN;
+        tmp = PyFloat_AsDouble(other);
+        if (tmp == -1.0 && PyErr_Occurred()) {
+            PyErr_Clear();
+            op |= OP_ARG_UNKNOWN;
+        }
+        else {
+            op |= OP_ARG_NUMBER;
+        }
     }
 
     if (op & OP_INPLACE) {
@@ -761,14 +765,12 @@ vector_generic_math(PyObject *o1, PyObject *o2, int op)
         case OP_MUL | OP_ARG_NUMBER:
         case OP_MUL | OP_ARG_NUMBER | OP_ARG_REVERSE:
         case OP_MUL | OP_ARG_NUMBER | OP_INPLACE:
-            tmp = PyFloat_AsDouble(other);
             for (i = 0; i < dim; i++) {
                 ret->coords[i] = vec_coords[i] * tmp;
             }
             break;
         case OP_DIV | OP_ARG_NUMBER:
         case OP_DIV | OP_ARG_NUMBER | OP_INPLACE:
-            tmp = PyFloat_AsDouble(other);
             if (tmp == 0.) {
                 PyErr_SetString(PyExc_ZeroDivisionError, "division by zero");
                 Py_DECREF(ret);
@@ -781,7 +783,6 @@ vector_generic_math(PyObject *o1, PyObject *o2, int op)
             break;
         case OP_FLOOR_DIV | OP_ARG_NUMBER:
         case OP_FLOOR_DIV | OP_ARG_NUMBER | OP_INPLACE:
-            tmp = PyFloat_AsDouble(other);
             if (tmp == 0.) {
                 PyErr_SetString(PyExc_ZeroDivisionError, "division by zero");
                 Py_DECREF(ret);
@@ -3971,12 +3972,15 @@ vector_elementwiseproxy_generic_math(PyObject *o1, PyObject *o2, int op)
             return NULL;
         }
     }
-    else if (RealNumber_Check(other)) {
-        op |= OP_ARG_NUMBER;
-        other_value = PyFloat_AsDouble(other);
-    }
     else {
-        op |= OP_ARG_UNKNOWN;
+        other_value = PyFloat_AsDouble(other);
+        if (other_value == -1.0 && PyErr_Occurred()) {
+            PyErr_Clear();
+            op |= OP_ARG_UNKNOWN;
+        }
+        else {
+            op |= OP_ARG_NUMBER;
+        }
     }
 
     ret = _vector_subtype_new(vec);