pythongh-116146: Add C-API to create module from spec and initfunc (pythonGH-139196)

Co-authored-by: Kumar Aditya <[email protected]>
Co-authored-by: Petr Viktorin <[email protected]>
Co-authored-by: Victor Stinner <[email protected]>

Author: 4 people

Doc/c-api/import.rst

@@ -333,3 +333,24 @@ Importing Modules
    strings instead of Python :class:`str` objects.
 
    .. versionadded:: 3.14
+
+.. c:function:: PyObject* PyImport_CreateModuleFromInitfunc(PyObject *spec, PyObject* (*initfunc)(void))
+
+   This function is a building block that enables embedders to implement
+   the :py:meth:`~importlib.abc.Loader.create_module` step of custom
+   static extension importers (e.g. of statically-linked extensions).
+
+   *spec* must be a :class:`~importlib.machinery.ModuleSpec` object.
+
+   *initfunc* must be an :ref:`initialization function <extension-export-hook>`,
+   the same as for :c:func:`PyImport_AppendInittab`.
+
+   On success, create and return a module object.
+   This module will not be initialized; call :c:func:`!PyModule_Exec`
+   to initialize it.
+   (Custom importers should do this in their
+   :py:meth:`~importlib.abc.Loader.exec_module` method.)
+
+   On error, return NULL with an exception set.
+
+   .. versionadded:: next

Doc/whatsnew/3.15.rst

@@ -1080,6 +1080,10 @@ New features
   thread state.
   (Contributed by Victor Stinner in :gh:`139653`.)
 
+* Add a new :c:func:`PyImport_CreateModuleFromInitfunc` C-API for creating
+  a module from a *spec* and *initfunc*.
+  (Contributed by Itamar Oren in :gh:`116146`.)
+
 
 Changed C APIs
 --------------

Include/cpython/import.h

@@ -10,6 +10,13 @@ struct _inittab {
 PyAPI_DATA(struct _inittab *) PyImport_Inittab;
 PyAPI_FUNC(int) PyImport_ExtendInittab(struct _inittab *newtab);
 
+// Custom importers may use this API to initialize statically linked
+// extension modules directly from a spec and init function,
+// without needing to go through inittab
+PyAPI_FUNC(PyObject *) PyImport_CreateModuleFromInitfunc(
+    PyObject *spec,
+    PyObject *(*initfunc)(void));
+
 struct _frozen {
     const char *name;                 /* ASCII encoded string */
     const unsigned char *code;

Lib/test/test_embed.py

@@ -239,6 +239,31 @@ def test_repeated_init_and_inittab(self):
         lines = "\n".join(lines) + "\n"
         self.assertEqual(out, lines)
 
+    def test_create_module_from_initfunc(self):
+        out, err = self.run_embedded_interpreter("test_create_module_from_initfunc")
+        if support.Py_GIL_DISABLED:
+            # the test imports a singlephase init extension, so it emits a warning
+            # under the free-threaded build
+            expected_runtime_warning = (
+                "RuntimeWarning: The global interpreter lock (GIL)"
+                " has been enabled to load module 'embedded_ext'"
+            )
+            filtered_err_lines = [
+                line
+                for line in err.strip().splitlines()
+                if expected_runtime_warning not in line
+            ]
+            self.assertEqual(filtered_err_lines, [])
+        else:
+            self.assertEqual(err, "")
+        self.assertEqual(out,
+                         "<module 'my_test_extension' (static-extension)>\n"
+                         "my_test_extension.executed='yes'\n"
+                         "my_test_extension.exec_slot_ran='yes'\n"
+                         "<module 'embedded_ext' (static-extension)>\n"
+                         "embedded_ext.executed='yes'\n"
+                         )
+
     def test_forced_io_encoding(self):
         # Checks forced configuration of embedded interpreter IO streams
         env = dict(os.environ, PYTHONIOENCODING="utf-8:surrogateescape")

Misc/NEWS.d/next/C_API/2025-11-08-10-51-50.gh-issue-116146.pCmx6L.rst

@@ -0,0 +1,2 @@
+Add a new :c:func:`PyImport_CreateModuleFromInitfunc` C-API for creating a
+module from a *spec* and *initfunc*. Patch by Itamar Oren.

Programs/_testembed.c

@@ -166,6 +166,8 @@ static PyModuleDef embedded_ext = {
 static PyObject*
 PyInit_embedded_ext(void)
 {
+    // keep this as a single-phase initialization module;
+    // see test_create_module_from_initfunc
     return PyModule_Create(&embedded_ext);
 }
 
@@ -1894,8 +1896,16 @@ static int test_initconfig_exit(void)
 }
 
 
+int
+extension_module_exec(PyObject *mod)
+{
+    return PyModule_AddStringConstant(mod, "exec_slot_ran", "yes");
+}
+
+
 static PyModuleDef_Slot extension_slots[] = {
     {Py_mod_gil, Py_MOD_GIL_NOT_USED},
+    {Py_mod_exec, extension_module_exec},
     {0, NULL}
 };
 
@@ -2213,6 +2223,106 @@ static int test_repeated_init_and_inittab(void)
     return 0;
 }
 
+static PyObject*
+create_module(PyObject* self, PyObject* spec)
+{
+    PyObject *name = PyObject_GetAttrString(spec, "name");
+    if (!name) {
+        return NULL;
+    }
+    if (PyUnicode_EqualToUTF8(name, "my_test_extension")) {
+        Py_DECREF(name);
+        return PyImport_CreateModuleFromInitfunc(spec, init_my_test_extension);
+    }
+    if (PyUnicode_EqualToUTF8(name, "embedded_ext")) {
+        Py_DECREF(name);
+        return PyImport_CreateModuleFromInitfunc(spec, PyInit_embedded_ext);
+    }
+    PyErr_Format(PyExc_LookupError, "static module %R not found", name);
+    Py_DECREF(name);
+    return NULL;
+}
+
+static PyObject*
+exec_module(PyObject* self, PyObject* mod)
+{
+    if (PyModule_Exec(mod) < 0) {
+        return NULL;
+    }
+    Py_RETURN_NONE;
+}
+
+static PyMethodDef create_static_module_methods[] = {
+    {"create_module", create_module, METH_O, NULL},
+    {"exec_module", exec_module, METH_O, NULL},
+    {}
+};
+
+static struct PyModuleDef create_static_module_def = {
+    PyModuleDef_HEAD_INIT,
+    .m_name = "create_static_module",
+    .m_size = 0,
+    .m_methods = create_static_module_methods,
+    .m_slots = extension_slots,
+};
+
+PyMODINIT_FUNC PyInit_create_static_module(void) {
+    return PyModuleDef_Init(&create_static_module_def);
+}
+
+static int
+test_create_module_from_initfunc(void)
+{
+    wchar_t* argv[] = {
+        PROGRAM_NAME,
+        L"-c",
+        // Multi-phase initialization
+        L"import my_test_extension;"
+        L"print(my_test_extension);"
+        L"print(f'{my_test_extension.executed=}');"
+        L"print(f'{my_test_extension.exec_slot_ran=}');"
+        // Single-phase initialization
+        L"import embedded_ext;"
+        L"print(embedded_ext);"
+        L"print(f'{embedded_ext.executed=}');"
+    };
+    PyConfig config;
+    if (PyImport_AppendInittab("create_static_module",
+                               &PyInit_create_static_module) != 0) {
+        fprintf(stderr, "PyImport_AppendInittab() failed\n");
+        return 1;
+    }
+    PyConfig_InitPythonConfig(&config);
+    config.isolated = 1;
+    config_set_argv(&config, Py_ARRAY_LENGTH(argv), argv);
+    init_from_config_clear(&config);
+    int result = PyRun_SimpleString(
+        "import sys\n"
+        "from importlib.util import spec_from_loader\n"
+        "import create_static_module\n"
+        "class StaticExtensionImporter:\n"
+        "   _ORIGIN = \"static-extension\"\n"
+        "   @classmethod\n"
+        "   def find_spec(cls, fullname, path, target=None):\n"
+        "       if fullname in {'my_test_extension', 'embedded_ext'}:\n"
+        "           return spec_from_loader(fullname, cls, origin=cls._ORIGIN)\n"
+        "       return None\n"
+        "   @staticmethod\n"
+        "   def create_module(spec):\n"
+        "       return create_static_module.create_module(spec)\n"
+        "   @staticmethod\n"
+        "   def exec_module(module):\n"
+        "       create_static_module.exec_module(module)\n"
+        "       module.executed = 'yes'\n"
+        "sys.meta_path.append(StaticExtensionImporter)\n"
+    );
+    if (result < 0) {
+        fprintf(stderr, "PyRun_SimpleString() failed\n");
+        return 1;
+    }
+    return Py_RunMain();
+}
+
 static void wrap_allocator(PyMemAllocatorEx *allocator);
 static void unwrap_allocator(PyMemAllocatorEx *allocator);
 
@@ -2396,6 +2506,7 @@ static struct TestCase TestCases[] = {
 #endif
     {"test_get_incomplete_frame", test_get_incomplete_frame},
     {"test_gilstate_after_finalization", test_gilstate_after_finalization},
+    {"test_create_module_from_initfunc", test_create_module_from_initfunc},
     {NULL, NULL}
 };
 

Python/import.c

@@ -2364,8 +2364,23 @@ is_builtin(PyObject *name)
     return 0;
 }
 
+static PyModInitFunction
+lookup_inittab_initfunc(const struct _Py_ext_module_loader_info* info)
+{
+    for (struct _inittab *p = INITTAB; p->name != NULL; p++) {
+        if (_PyUnicode_EqualToASCIIString(info->name, p->name)) {
+            return (PyModInitFunction)p->initfunc;
+        }
+    }
+    // not found
+    return NULL;
+}
+
 static PyObject*
-create_builtin(PyThreadState *tstate, PyObject *name, PyObject *spec)
+create_builtin(
+    PyThreadState *tstate, PyObject *name,
+    PyObject *spec,
+    PyModInitFunction initfunc)
 {
     struct _Py_ext_module_loader_info info;
     if (_Py_ext_module_loader_info_init_for_builtin(&info, name) < 0) {
@@ -2396,25 +2411,15 @@ create_builtin(PyThreadState *tstate, PyObject *name, PyObject *spec)
         _extensions_cache_delete(info.path, info.name);
     }
 
-    struct _inittab *found = NULL;
-    for (struct _inittab *p = INITTAB; p->name != NULL; p++) {
-        if (_PyUnicode_EqualToASCIIString(info.name, p->name)) {
-            found = p;
-            break;
-        }
-    }
-    if (found == NULL) {
-        // not found
-        mod = Py_NewRef(Py_None);
-        goto finally;
-    }
-
-    PyModInitFunction p0 = (PyModInitFunction)found->initfunc;
+    PyModInitFunction p0 = initfunc;
     if (p0 == NULL) {
-        /* Cannot re-init internal module ("sys" or "builtins") */
-        assert(is_core_module(tstate->interp, info.name, info.path));
-        mod = import_add_module(tstate, info.name);
-        goto finally;
+        p0 = lookup_inittab_initfunc(&info);
+        if (p0 == NULL) {
+            /* Cannot re-init internal module ("sys" or "builtins") */
+            assert(is_core_module(tstate->interp, info.name, info.path));
+            mod = import_add_module(tstate, info.name);
+            goto finally;
+        }
     }
 
 #ifdef Py_GIL_DISABLED
@@ -2440,6 +2445,33 @@ create_builtin(PyThreadState *tstate, PyObject *name, PyObject *spec)
     return mod;
 }
 
+PyObject*
+PyImport_CreateModuleFromInitfunc(
+    PyObject *spec, PyObject *(*initfunc)(void))
+{
+    if (initfunc == NULL) {
+        PyErr_BadInternalCall();
+        return NULL;
+    }
+
+    PyThreadState *tstate = _PyThreadState_GET();
+
+    PyObject *name = PyObject_GetAttr(spec, &_Py_ID(name));
+    if (name == NULL) {
+        return NULL;
+    }
+
+    if (!PyUnicode_Check(name)) {
+        PyErr_Format(PyExc_TypeError,
+                     "spec name must be string, not %T", name);
+        Py_DECREF(name);
+        return NULL;
+    }
+
+    PyObject *mod = create_builtin(tstate, name, spec, initfunc);
+    Py_DECREF(name);
+    return mod;
+}
 
 /*****************************/
 /* the builtin modules table */
@@ -3209,7 +3241,7 @@ bootstrap_imp(PyThreadState *tstate)
     }
 
     // Create the _imp module from its definition.
-    PyObject *mod = create_builtin(tstate, name, spec);
+    PyObject *mod = create_builtin(tstate, name, spec, NULL);
     Py_CLEAR(name);
     Py_DECREF(spec);
     if (mod == NULL) {
@@ -4369,7 +4401,7 @@ _imp_create_builtin(PyObject *module, PyObject *spec)
         return NULL;
     }
 
-    PyObject *mod = create_builtin(tstate, name, spec);
+    PyObject *mod = create_builtin(tstate, name, spec, NULL);
     Py_DECREF(name);
     return mod;
 }