Improve the Efficiency of Python3.11.1 __getattr__

[wangxiang-hz]

Feature or enhancement

__getattr__ in Python3.11.1 is much slower than @Property and visiting an object's attribute. It's even slower than Python3.10.4.
_PyObject_GenericGetAttrWithDict is the key reason. If Python3 fails finding an attribute in normal ways, it will return NULL and raise an exception. But raising an exception has performance cost. Python3.11.1 add set_attribute_error_context to support Fine Grained Error Locations in Tracebacks. It makes things worser.

Pitch

We can use this test code:

import time
import sys

class A:
    def foo(self):
        print("Call A.foo!")

    def __getattr__(self, name):
        return 2

    @property
    def ppp(self):
        return 3

class B(A):
    def foo(self):
        print("Call B.foo!")


class C(B):

    def __init__(self) -> None:
        self.pps = 1

    def foo(self):
        print("Call C.foo!")


def main():
    start = time.time()
    for i in range(1, 1000000):
        pass
    end = time.time()
    peer = end - start
    c = C()
    print(f"Python version of {sys.version}")
    start = time.time()
    for i in range(1, 1000000):
        s = c.pps
    end = time.time()
    print(f"Normal getattr spend time: {end - start - peer}")
    start = time.time()
    for i in range(1, 1000000):
        s = c.ppa
    end = time.time()
    print(f"Call __getattr__ spend time: {end - start - peer}")
    start = time.time()
    for i in range(1, 1000000):
        s = c.ppp
    end = time.time()
    print(f"Call property spend time: {end - start - peer}")

if __name__ == "__main__":
    main()

The result shows how slow __getattr__ is:

Python version of 3.11.1 (main, Dec 26 2022, 16:32:50) [GCC 8.3.0]
Normal getattr spend time: 0.03204226493835449
Call __getattr__ spend time: 0.4767305850982666
Call property spend time: 0.06345891952514648

When we define __getattr__, failed to find an attribute is what we expected. If we can get this result and then call __getattr__ without exception handling, it will be faster.
I tried to modify Python3.11.1 like this:

1. add a new function in object.c:

PyObject *
PyObject_GenericTryGetAttr(PyObject *obj, PyObject *name)
{
    return _PyObject_GenericGetAttrWithDict(obj, name, NULL, 1);
}

2. change typeobject.c :

if (getattribute == NULL ||
        (Py_IS_TYPE(getattribute, &PyWrapperDescr_Type) &&
         ((PyWrapperDescrObject *)getattribute)->d_wrapped ==
         (void *)PyObject_GenericGetAttr))
        // res = PyObject_GenericGetAttr(self, name);
        res = PyObject_GenericTryGetAttr(self, name);
    else {
        Py_INCREF(getattribute);
        res = call_attribute(self, getattribute, name);
        Py_DECREF(getattribute);
    }
    if (res == NULL) {
        if (PyErr_ExceptionMatches(PyExc_AttributeError))
            PyErr_Clear();
        res = call_attribute(self, getattr, name);
    }
    Py_DECREF(getattr);
    return res;

Rebuild python, it really become faster: spend time: 0.13772845268249512.

Previous discussion

getattr is much slower in Python3.11

Linked PRs

• gh-102213: Optimizing the Performance of __getattr__ #102213 #102248
• Revert "gh-102213: Optimize the performance of __getattr__ (GH-102248)" #103332
• gh-102213: Optimize the performance of __getattr__ #103761

[Fidget-Spinner]

Could you submit a pull request please? I'll try to review it.

[wangxiang-hz]

Could you submit a pull request please? I'll try to review it.

I have already submitted a pull request, but the default assigned reviewer is too busy to review it. Could you please review it and provide some suggestions when you are available? Thank you very much.

[Fidget-Spinner]

Sorry. I didn't see the pull request (pull requests linked to an issue don't ping me even though I'm subscribe to the issue). I'll review it. Thanks!

[wangxiang-hz]

Sorry. I didn't see the pull request (pull requests linked to an issue don't ping me even though I'm subscribe to the issue). I'll review it. Thanks!

Thank you. I know that contributors in open source communities are very busy, so I didn't want to bother you with a “polite ping” :) I really appreciate your review.