Simple benchmark for comparing different dataclass-like providers.

Fully works only up to Python 3.12. Does not work with Python 3.13 (yet), due to inaccuracies with Pympler for calculating the memory usage.

Adapted from https://github.com/mCodingLLC/VideosSampleCode/blob/master/videos/079_which_dataclass_is_best/dataclasses_comparison.py

But included comparisons to:

• msgspec
• SQLModel

Benchmarks

Creation Speeds (in ns)

Rank	Type	Time (ns)
0	tuple	17
1	msgspec	45
2	dict	53
3	dataclass (slots)	103
4	attr class (slots)	104
5	plain class (slots)	104
6	attr class	117
7	SimpleNamespace	118
8	dataclass	120
9	plain class	120
10	namedtuple	160
11	NamedTuple	162
12	pydantic	683
13	SQLModel (table=False)	1769
14	SQLAlchemyDataclass	2712
15	SQLAlchemy	3616

getattr Speeds (in ns)

Rank	Type	Time (ns)
0	tuple	8
1	pydantic	8
2	dataclass	8
3	plain class	8
4	plain class (slots)	8
5	msgspec	8
6	attr class (slots)	8
7	SQLModel (table=False)	8
8	dataclass (slots)	8
9	attr class	8
10	dict	9
11	namedtuple	11
12	NamedTuple	11
13	SimpleNamespace	16
14	SQLAlchemyDataclass	92
15	SQLAlchemy	94

setattr Speeds (in ns)

Rank	Type	Time (ns)
0	dataclass	8
1	plain class (slots)	8
2	dataclass (slots)	8
3	attr class (slots)	8
4	plain class	8
5	attr class	8
6	dict	12
7	msgspec	18
8	SimpleNamespace	19
9	SQLAlchemy	337
10	SQLAlchemyDataclass	346
11	pydantic	1233
12	SQLModel (table=False)	1497

Memory Usage (in bytes)

Rank	Type	Memory Usage (bytes)
0	dataclass (slots)	160
1	plain class (slots)	160
2	msgspec	160
3	tuple	168
4	namedtuple	168
5	NamedTuple	168
6	attr class (slots)	176
7	dict	432
8	SimpleNamespace	472
9	pydantic	504
10	SQLModel (table=False)	504
11	plain class	592
12	dataclass	592
13	attr class	592