Python dataclasses vs attrs vs Pydantic: When to Use Each

Python has multiple data class libraries. Choosing the wrong one for the use case leads to missing validation, performance issues, or excessive boilerplate.

Choose based on your primary need: stdlib simplicity, performance, or validation.

# dataclasses — stdlib, simple, no validation
from dataclasses import dataclass, field
from typing import List

@dataclass
class User:
    name: str
    email: str
    tags: List[str] = field(default_factory=list)
    # No runtime validation — name could be an int

# attrs — performance, validators, converters
import attr

@attr.s(auto_attribs=True, slots=True)
class User:
    name: str = attr.ib(validator=attr.validators.instance_of(str))
    email: str = attr.ib()
    tags: list = attr.ib(factory=list)
    # 2-5x faster than dataclasses for large-scale use

# Pydantic v2 — validation + serialization (FastAPI standard)
from pydantic import BaseModel, EmailStr, field_validator

class User(BaseModel):
    name: str
    email: EmailStr  # Validates format
    tags: list[str] = []

    @field_validator('name')
    @classmethod
    def name_not_empty(cls, v: str) -> str:
        if not v.strip():
            raise ValueError('Name cannot be empty')
        return v.strip()

user = User(name='Alice', email='alice@example.com')
user.model_dump()  # {'name': 'Alice', 'email': 'alice@example.com', 'tags': []}

dataclasses are built-in but have no validation. attrs has validators and is faster (optional slots). Pydantic v2 (Rust core) has runtime type coercion and validation, essential for API boundaries.