Optimise `inspect.getattr_static`

[AlexWaygood]

Feature or enhancement

Following 6d59c9e, typing._ProtocolMeta.__instancecheck__ uses inspect.getattr_static in a tight loop, where it had previously used hasattr. This improves semantics in several highly desirable ways, but causes a considerable slowdown for _ProtocolMeta.__instancecheck__, as inspect.getattr_static is much slower than hasattr.

The performance hit to _ProtocolMeta.__instancecheck__ has already been mostly mitigated through several typing-specific optimisations that are tracked in this issue:

• Performance of typing._ProtocolMeta._get_protocol_attrs and isinstance #74690

However, it would be good to also see if we can improve the performance of inspect.getattr_static. This will not only improve the performance of isinstance() checks against classes subclassing typing.Protocol. It will also improve the performance of all other tools that use inspect.getattr_static for introspection without side effects.

Linked PRs

• gh-103193: Micro-optimise helper functions for inspect.getattr_static #103195
• gh-103193: Use LBYL idioms rather than EAFP in inspect.getattr_static #103318
• gh-103193: Speedup and inline inspect._is_type #103321
• gh-103193: Celebrate performance improvements to inspect.getattr_static in 'What's New in Python 3.12' #103349
• gh-103193: cache calls to inspect._shadowed_dict in inspect.getattr_static #104267
• gh-103193: Improve getattr_static test coverage #104286
• [3.11] gh-103193: Improve getattr_static test coverage (GH-104286) #104290
• gh-103193: Fix refleaks in test_inspect and test_typing #104320

[AlexWaygood]

I can't see anything further that can be easily optimised here, but happy to reopen if somebody else can!

[sobolevn]

I've measured my ideas, but the results are close to noise.

Initial numbers:

type[Foo]                :  793 ±  1 ns 
Foo                      : 1852 ±  3 ns 
type[Bar]                :  791 ±  1 ns 
Bar                      : 1843 ±  2 ns 
WithParentClassX         : 2768 ±  2 ns 
Baz                      : 2712 ±  8 ns 
WithParentX              : 3611 ±  9 ns 
type[Missing]            : 3364 ± 13 ns 
Missing                  : 2749 ± 22 ns 
Slotted                  : 2363 ±  7 ns 
Method                   : 1863 ± 11 ns 
StMethod                 : 1863 ±  8 ns 
ClsMethod                : 1847 ±  4 ns 

After all changes:

type[Foo]                :  793 ±  2 ns 
Foo                      : 1833 ±  2 ns 
type[Bar]                :  793 ±  3 ns 
Bar                      : 1833 ±  5 ns 
WithParentClassX         : 2756 ±  4 ns 
Baz                      : 2701 ±  8 ns 
WithParentX              : 3696 ± 28 ns 
type[Missing]            : 3390 ± 41 ns 
Missing                  : 2703 ±  9 ns 
Slotted                  : 2312 ± 11 ns 
Method                   : 1860 ± 19 ns 
StMethod                 : 1828 ±  5 ns 
ClsMethod                : 1859 ± 23 ns

Code:

from timeit import timeit
from statistics import mean, stdev
from inspect import getattr_static

class Foo:
    @property
    def x(self) -> int:
        return 42

class Bar:
    x = 42

class WithParentClassX(Bar): ...

class Baz:
    def __init__(self):
        self.x = 42

class WithParentX(Baz): ...

class Missing: ...

class Slotted:
    __slots__ = ('x',)
    def __init__(self):
        self.x = 42

class Method:
    def x(self): ...

class ClsMethod:
    @classmethod
    def x(cls): ...

class StMethod:
    @staticmethod
    def x(): ...

import gc
gc.disable()

times = []
def stats():
    ts = [t * 1e8 for t in sorted(times)[:5]]
    return f'{round(mean(ts)):4} ± {round(stdev(ts)):2} ns '

def bench(obj):
    # Warmup:
    for _ in range(5):
        number = 100
        timeit(lambda: getattr_static(obj, 'x', None), number=number)

    # Actual bench:
    for _ in range(50):
        number = 1000
        t = timeit(lambda: getattr_static(obj, 'x', None), number=number) / number
        times.append(t)

    bench_name = (
        f'type[{obj.__name__}]'
        if isinstance(obj, type)
        else obj.__class__.__name__
    )
    print(f"{bench_name: <25}: {stats()}")
    times.clear()


bench(Foo)
bench(Foo())
bench(Bar)
bench(Bar())
bench(WithParentClassX())
bench(Baz())
bench(WithParentX())
bench(Missing)
bench(Missing())
bench(Slotted())
bench(Method())
bench(StMethod())
bench(ClsMethod())