bpo-41842: Add a unregister function in _codecs module

Doc/library/codecs.rst

@@ -163,11 +163,14 @@ function:
    :class:`CodecInfo` object. In case a search function cannot find
    a given encoding, it should return ``None``.
 
-   .. note::
 
-      Search function registration is not currently reversible,
-      which may cause problems in some cases, such as unit testing or
-      module reloading.
+.. function:: unregister(search_function)
+
+   Unregister a codec search function and clear the registry's cache.
+   If the search function is not registered, do nothing.
+
+   .. versionadded:: 3.10
+
 
 While the builtin :func:`open` and the associated :mod:`io` module are the
 recommended approach for working with encoded text files, this module

Doc/whatsnew/3.10.rst

@@ -109,6 +109,11 @@ base64
 Add :func:`base64.b32hexencode` and :func:`base64.b32hexdecode` to support the
 Base32 Encoding with Extended Hex Alphabet.
 
+codecs
+------
+
+Add :func:`codecs.unregister` to unregister a codec search function.
+
 curses
 ------
 

Include/codecs.h

@@ -27,6 +27,16 @@ PyAPI_FUNC(int) PyCodec_Register(
        PyObject *search_function
        );
 
+/* Unregister a codec search function.
+
+   Unregister a codec search function and clear the registry's cache.
+   If the search function is not registered, do nothing.
+   */
+
+PyAPI_FUNC(int) PyCodec_Unregister(
+       PyObject *search_function
+       );
+
 /* Codec registry lookup API.
 
    Looks up the given encoding and returns a CodecInfo object with

Lib/test/test_codecs.py

@@ -1641,6 +1641,13 @@ def test_register(self):
         self.assertRaises(TypeError, codecs.register)
         self.assertRaises(TypeError, codecs.register, 42)
 
+    def test_unregister(self):
+        search_function = mock.Mock(return_value=(1, 2, 3, 4))
+        codecs.register(search_function)
+        codecs.unregister(search_function)
+        self.assertRaises(LookupError, codecs.lookup, "test")
+        search_function.assert_not_called()
+
     def test_lookup(self):
         self.assertRaises(TypeError, codecs.lookup)
         self.assertRaises(LookupError, codecs.lookup, "__spam__")

Misc/NEWS.d/next/Library/2020-09-23-22-52-24.bpo-41842.lIuhC9.rst

@@ -0,0 +1 @@
+Add a :func:`codecs.unregister` to unregister a codec search function.

Modules/_codecsmodule.c

@@ -68,6 +68,27 @@ _codecs_register(PyObject *module, PyObject *search_function)
     Py_RETURN_NONE;
 }
 
+/*[clinic input]
+_codecs.unregister
+    search_function: object
+    /
+
+Unregister a codec search function and clear the registry's cache.
+
+If the search function is not registered, do nothing.
+[clinic start generated code]*/
+
+static PyObject *
+_codecs_unregister(PyObject *module, PyObject *search_function)
+/*[clinic end generated code: output=1f0edee9cf246399 input=dd7c004c652d345e]*/
+{
+    if (PyCodec_Unregister(search_function) < 0) {
+        return NULL;
+    }
+
+    Py_RETURN_NONE;
+}
+
 /*[clinic input]
 _codecs.lookup
     encoding: str
@@ -992,6 +1013,7 @@ _codecs_lookup_error_impl(PyObject *module, const char *name)
 
 static PyMethodDef _codecs_functions[] = {
     _CODECS_REGISTER_METHODDEF
+    _CODECS_UNREGISTER_METHODDEF
     _CODECS_LOOKUP_METHODDEF
     _CODECS_ENCODE_METHODDEF
     _CODECS_DECODE_METHODDEF

Python/codecs.c

@@ -50,6 +50,26 @@ int PyCodec_Register(PyObject *search_function)
     return -1;
 }
 
+int
+PyCodec_Unregister(PyObject *search_function)
+{
+    PyInterpreterState *interp = PyInterpreterState_Get();
+    /* Do nothing if codec_search_path is not created yet or was created. */
+    if (interp->codec_search_path == NULL) {
+        return 0;
+    }
+
+    Py_ssize_t n = PyList_Size(interp->codec_search_path);
+    for (Py_ssize_t i = 0; i < n; i++) {
+        PyObject *item = PyList_GetItem(interp->codec_search_path, i);
+        if (item == search_function) {
+            PyDict_Clear(interp->codec_search_cache);
+            return PyList_SetSlice(interp->codec_search_path, i, i+1, NULL);
+        }
+    }
+    return 0;
+}
+
 extern int _Py_normalize_encoding(const char *, char *, size_t);
 
 /* Convert a string to a normalized Python string(decoded from UTF-8): all characters are