TypedDict instead of dataclass?

[nsmith-]

In a microbenchmark it is slightly faster to use typed dictionaries:

import timeit
from dataclasses import dataclass
from typing import TypedDict


@dataclass
class TestClass:
    a: int
    b: bytes
    c: int


@dataclass
class TestDerived(TestClass):
    d: int
    e: bytes
    f: int


class TestClass2(TypedDict):
    a: int
    b: bytes
    c: int


class TestDerived2(TestClass2):
    d: int
    e: bytes
    f: int


ARGS = (1, b"2", 3, 4, b"5", 6)
KWARGS = {"a": 1, "b": b"2", "c": 3, "d": 4, "e": b"5", "f": 6}


@timeit.timeit
def make_from_args():
    return TestDerived(*ARGS)


@timeit.timeit
def make_from_dict():
    return TestDerived(**KWARGS)


@timeit.timeit
def make_from_dict_typed():
    return TestDerived2(KWARGS)  # type: ignore[call-arg]


if __name__ == "__main__":
    print("From args (us):", make_from_args)
    print("From dict (us):", make_from_dict)
    print("From typed dict (us):", make_from_dict_typed)

returns for me

From args (us): 0.09821387499687262
From dict (us): 0.1693307499808725
From typed dict (us): 0.06846924999263138