python · vstinner · Feb 21, 2025 · Feb 1, 2025 · Feb 2, 2025 · Feb 2, 2025
diff --git a/Misc/NEWS.d/next/Library/2025-02-02-16-30-27.gh-issue-129569.i0kPOG.rst b/Misc/NEWS.d/next/Library/2025-02-02-16-30-27.gh-issue-129569.i0kPOG.rst
@@ -0,0 +1 @@
+Fix an issue in :func:`unicodedata.normalize` where the return value's type is inconsistent when the function receives an instance of a subclass of :class:`str` due to incorrect optimization. Now, the function always returns an instance of the built-in :class:`str` type.
diff --git a/Modules/unicodedata.c b/Modules/unicodedata.c
@@ -933,34 +933,34 @@ unicodedata_UCD_normalize_impl(PyObject *self, PyObject *form,
     if (PyUnicode_GET_LENGTH(input) == 0) {
         /* Special case empty input strings, since resizing
            them  later would cause internal errors. */
-        return Py_NewRef(input);
+        return PyUnicode_FromObject(input);
     }
 
     if (PyUnicode_CompareWithASCIIString(form, "NFC") == 0) {
         if (is_normalized_quickcheck(self, input,
                                      true,  false, true) == YES) {
-            return Py_NewRef(input);
+            return PyUnicode_FromObject(input);
         }
         return nfc_nfkc(self, input, 0);
     }
     if (PyUnicode_CompareWithASCIIString(form, "NFKC") == 0) {
         if (is_normalized_quickcheck(self, input,
                                      true,  true,  true) == YES) {
-            return Py_NewRef(input);
+            return PyUnicode_FromObject(input);
         }
         return nfc_nfkc(self, input, 1);
     }
     if (PyUnicode_CompareWithASCIIString(form, "NFD") == 0) {
         if (is_normalized_quickcheck(self, input,
                                      false, false, true) == YES) {
-            return Py_NewRef(input);
+            return PyUnicode_FromObject(input);
         }
         return nfd_nfkd(self, input, 0);
     }
     if (PyUnicode_CompareWithASCIIString(form, "NFKD") == 0) {
         if (is_normalized_quickcheck(self, input,
                                      false, true,  true) == YES) {
-            return Py_NewRef(input);
+            return PyUnicode_FromObject(input);
         }
         return nfd_nfkd(self, input, 1);
     }
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1 @@
		Fix an issue in :func:`unicodedata.normalize` where the return value's type is inconsistent when the function receives an instance of a subclass of :class:`str` due to incorrect optimization. Now, the function always returns an instance of the built-in :class:`str` type.
Copy link Member vstinner Feb 4, 2025 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others. Learn more. I suggest: Change :func:`unicodedata.normalize` return type to :class:`str` when the string is already normalized and the input is a :class:`str` subclass. Copy link Contributor Author Hizuru3 Feb 7, 2025 Choose a reason for hiding this comment The reason will be displayed to describe this comment to others. Learn more. I'd like to retain the word "fix" since the issue is considered to be a bug to fix, not merely to change. How about this? Fix :func:`unicodedata.normalize` to always return a built-in :class:`str` object when given an input of a :class:`str` subclass, regardless of whether the string is already normalized.