gh-144140: Optimize len for string constants in optimizer (GH-144142)

Author: y1yang0

Lib/test/test_capi/test_opt.py

@@ -2419,6 +2419,21 @@ def testfunc(n):
         self.assertNotIn("_GUARD_TOS_INT", uops)
         self.assertIn("_POP_TOP_NOP", uops)
 
+    def test_call_len_string(self):
+        def testfunc(n):
+            for _ in range(n):
+                _ = len("abc")
+                d = ''
+                _ = len(d)
+                _ = len(b"def")
+                _ = len(b"")
+
+        _, ex = self._run_with_optimizer(testfunc, TIER2_THRESHOLD)
+        self.assertIsNotNone(ex)
+        uops = get_opnames(ex)
+        self.assertNotIn("_CALL_LEN", uops)
+        self.assertEqual(count_ops(ex, "_SHUFFLE_3_LOAD_CONST_INLINE_BORROW"), 4)
+
     def test_call_len_known_length_small_int(self):
         # Make sure that len(t) is optimized for a tuple of length 5.
         # See https://github.com/python/cpython/issues/139393.

Python/optimizer_bytecodes.c

@@ -1455,15 +1455,28 @@ dummy_func(void) {
 
     op(_CALL_LEN, (callable, null, arg -- res, a, c)) {
         res = sym_new_type(ctx, &PyLong_Type);
-        Py_ssize_t tuple_length = sym_tuple_length(arg);
-        if (tuple_length >= 0) {
-            PyObject *temp = PyLong_FromSsize_t(tuple_length);
+        Py_ssize_t length = sym_tuple_length(arg);
+
+        // Not a tuple, check if it's a const string
+        if (length < 0 && sym_is_const(ctx, arg)) {
+            PyObject *const_val = sym_get_const(ctx, arg);
+            if (const_val != NULL) {
+                if (PyUnicode_CheckExact(const_val)) {
+                    length = PyUnicode_GET_LENGTH(const_val);
+                }
+                else if (PyBytes_CheckExact(const_val)) {
+                    length = PyBytes_GET_SIZE(const_val);
+                }
+            }
+        }
+
+        if (length >= 0) {
+            PyObject *temp = PyLong_FromSsize_t(length);
             if (temp == NULL) {
                 goto error;
             }
             if (_Py_IsImmortal(temp)) {
-                ADD_OP(_SHUFFLE_3_LOAD_CONST_INLINE_BORROW,
-                           0, (uintptr_t)temp);
+                ADD_OP(_SHUFFLE_3_LOAD_CONST_INLINE_BORROW, 0, (uintptr_t)temp);
             }
             res = sym_new_const(ctx, temp);
             Py_DECREF(temp);