[C API] Hide static types from the limited C API

@gvanrossum

BPO	40601
Nosy	@gvanrossum, @vstinner, @encukou, @shihai1991, @erlend-aasland, @h-vetinari, @junyixie
PRs	[WIP] bpo-40601: Add functions to get builtin types #24146

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2020-05-11.22:23:25.790>
labels = ['expert-subinterpreters', 'expert-C-API', '3.9']
title = '[C API] Hide static types from the limited C API'
updated_at = <Date 2022-01-05.17:59:11.136>
user = 'https://github.com/vstinner'

bugs.python.org fields:

activity = <Date 2022-01-05.17:59:11.136>
actor = 'erlendaasland'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['C API', 'Subinterpreters']
creation = <Date 2020-05-11.22:23:25.790>
creator = 'vstinner'
dependencies = []
files = []
hgrepos = []
issue_num = 40601
keywords = ['patch']
message_count = 11.0
messages = ['368667', '368668', '368676', '368731', '369999', '384539', '386140', '388541', '388545', '388546', '389474']
nosy_count = 7.0
nosy_names = ['gvanrossum', 'vstinner', 'petr.viktorin', 'shihai1991', 'erlendaasland', 'h-vetinari', 'JunyiXie']
pr_nums = ['24146']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = None
url = 'https://bugs.python.org/issue40601'
versions = ['Python 3.9']

"Statically allocated types" prevents to get per-interpreter GIL: bpo-40512. These types are currently shared by all interpreters.

Eric Snow proposed the idea of creating a heap allocated type in subintepreters. But we should take care of direct usage of the statically allocated type.

For example, Objects/longobject.c defines "PyTypeObject PyLong_Type = {...};". This type is exposed in the limited C API (!) in Include/longobject.c:

PyAPI_DATA(PyTypeObject) PyLong_Type;

It's used but such macro:

#define PyLong_CheckExact(op) Py_IS_TYPE(op, &PyLong_Type)

I don't think that these types are directly accessed in C extensions built with the limited C API. My expectation is that the type is only exposed for "CheckExact" macros.

Currently, 100 statically allocated types are declared in Python header files:

$ grep -F '(PyTypeObject)' Include/ -R
Include/cpython/fileobject.h:PyAPI_DATA(PyTypeObject) PyStdPrinter_Type;
(...)
Include/object.h:PyAPI_DATA(PyTypeObject) PySuper_Type; /* built-in 'super' */
Include/methodobject.h:PyAPI_DATA(PyTypeObject) PyCFunction_Type;

Most of them seem to be exposed in the limited C API.

I propose to break the limited C API backward compatibility on purpose by removing these type definitions form the limited C API.

For "CheckExact" macros, we can continue to provide them in the limited C API but as function calls. So a built C extension would no longer access directly the type, but only do function calls.

See also bpo-40077: "Convert static types to PyType_FromSpec()".

I propose to break the limited C API backward compatibility on purpose by removing these type definitions form the limited C API.

Hum. How would a C extension subclass the Python int type (PyLong_Type) if it's no longer exposed? One option is to add one function per type, like:

PyObject* Py_GetLongType(void);

It would return a *strong reference* to the type (PyLong_Type).

Another option is to get the type from builtins module or builtins dictionary (PyInterpreterState.builtins). But there is no simple C function to get a builtin object. It requires many calls, handle errors, etc. Maybe a generic helper like the following function would help:

PyObject *Py_GetBuiltin(const char *name);

Note: PyEval_GetBuiltins() exposes the builtins of the *current frame* which maybe not be what you may expect.

Currently, Py_GetBuiltin(name) is not needed since basically *all* Python builtins are *directly* exposed in the C API...

For example, Objects/longobject.c defines "PyTypeObject PyLong_Type = {...};". This type is exposed in the limited C API (!)

Technically, it is not, see https://www.python.org/dev/peps/pep-0384/#structures
Structures like PyLong_Type are *not* part of the limited API.

I propose to break the limited C API backward compatibility on purpose by removing these type definitions form the limited C API.

That could only be done in Python 4.0, or if we started C-API 4.0. But I don't think it's necessary here.

Technically, it is not, see https://www.python.org/dev/peps/pep-0384/#structures
Structures like PyLong_Type are *not* part of the limited API.

The symbol is exported by libpython:

$ objdump -T /lib64/libpython3.8.so.1.0|grep PyLong_Type
000000000030de00 g    DO .data	00000000000001a0  Base        PyLong_Type

A C extension can use a reference to PyLong_Type.

I don't think it's necessary here.

Did you read my rationale (first message)? Do you mean that per-interpreter GIL is not worth it?

--

A first step would be to expose "CheckExact" macros as function calls in the limited C API.

PC/python3dll.c exports 66 types in the stable ABI:

Py_GenericAliasType
PyObject_Type
_PyWeakref_CallableProxyType
_PyWeakref_ProxyType
_PyWeakref_RefType
PyBaseObject_Type
PyBool_Type
PyByteArray_Type
PyByteArrayIter_Type
PyBytes_Type
PyBytesIter_Type
PyCallIter_Type
PyCapsule_Type
PyCFunction_Type
PyClassMethodDescr_Type
PyComplex_Type
PyDict_Type
PyDictItems_Type
PyDictIterItem_Type
PyDictIterKey_Type
PyDictIterValue_Type
PyDictKeys_Type
PyDictProxy_Type
PyDictValues_Type
PyEllipsis_Type
PyEnum_Type
PyExc_TypeError
PyFilter_Type
PyFloat_Type
PyFrozenSet_Type
PyGetSetDescr_Type
PyList_Type
PyListIter_Type
PyListRevIter_Type
PyLong_Type
PyLongRangeIter_Type
PyMap_Type
PyMemberDescr_Type
PyMemoryView_Type
PyMethodDescr_Type
PyModule_Type
PyModuleDef_Type
PyNullImporter_Type
PyODict_Type
PyODictItems_Type
PyODictIter_Type
PyODictKeys_Type
PyODictValues_Type
PyProperty_Type
PyRange_Type
PyRangeIter_Type
PyReversed_Type
PySeqIter_Type
PySet_Type
PySetIter_Type
PySlice_Type
PySortWrapper_Type
PySuper_Type
PyTraceBack_Type
PyTuple_Type
PyTupleIter_Type
PyType_Type
PyUnicode_Type
PyUnicodeIter_Type
PyWrapperDescr_Type
PyZip_Type

Sorry, I lost this bug in my TODO list :(

> I don't think it's necessary here.

Did you read my rationale (first message)? Do you mean that per-interpreter GIL is not worth it?

Right, I mean that it it is not worth breaking the C-API for all existing modules.
Instead, I think that it can be done as an addition: only modules that don't use things like these static types would be allowed in subinterpreters that have their own GIL.

It seems that there is no continued progress for move static type in heap.This will make it impossible to continue to achieve sub interpreters parallel. Are there any plans to try other solutions to the problem?

In my project, i try to slove this problem, It can work, we verify on millions of devices.

In typeobject.c add lock to ensure that some functions that modification type are thread-safe
and make the PyCFunction and descri object of the Type will never be released. （Frequently used when load method/attributed, locking affects performance）

Can this change be submitted to cpython？

The Steering Council asked for a PEP to explain why static types should be converted to heap types.

I plan to write such PEP soon.

FWIW I have an idea that would allow code using e.g. &PyList_Type to continue to work, and even ABI compatible (though only in the main interpreter).

// In some header file

PyAPI_FUNC(PyHeapTypeObject *) PyList_GetType();

#define PyList_Type (PyList_GetType()->ht_type)

For the main interpreter we could make this return the address of PyList_Type.

I plan to write such PEP soon.

Sadly, I worked on other topics in the meanwhile and I left sub-interpreters aside. PEP 687 was accepted which is a step forward, even if it doesn't propose any solution for this specific problem.

Until I can come up with an idea which doesn't break the API, I prefer to close the issue. If someone wants to experiment Guido's idea, please go ahead and propose a PR or even write a PEP.

I'm not sure that going through PyTypeObject.ht_type is needed, the macro can just be: #define PyLong_Type (*PyLong_GetType()). Example:

#include <stdio.h>

typedef struct {
    const char *name;
} PyTypeObject;

PyTypeObject PyLong_Type = {.name = "int"};

PyTypeObject* PyLong_GetType(void) {
    return &PyLong_Type;
}

#define PyLong_Type_MACRO (*PyLong_GetType())

int main()
{
    printf("PyLong_Type.name = %s\n", PyLong_Type.name);
    // at the API level, PyLong_Type_MACRO type is a PyTypeObject instance
    printf("PyLong_Type.name = %s\n", PyLong_Type_MACRO.name);
}

Navigation Menu

Repository navigation

[C API] Hide static types from the limited C API #84781

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

[C API] Hide static types from the limited C API #84781

Description

Activity

vstinner commented on May 11, 2020

vstinner commented on May 11, 2020

vstinner commented on May 11, 2020

encukou commented on May 12, 2020

vstinner commented on May 26, 2020

vstinner commented on Jan 6, 2021

encukou commented on Feb 2, 2021

JunyiXie commented on Mar 12, 2021

vstinner commented on Mar 12, 2021

vstinner commented on Mar 12, 2021

gvanrossum commented on Mar 24, 2021

vstinner commented on Dec 13, 2022

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions