validating 0.0.1

Provides a lightweight `attr()` descriptor factory that adds runtime validation to dataclass fields.

validating

validating is a lightweight runtime validation library focused on making dataclass fields and function arguments safer and easier to validate.

It exposes three main entry points:

• attr(...): declare validated fields (type checks, bounds, allow/deny lists, custom validators)
• @dataclass(...): a compatible enhancement of dataclasses.dataclass with automatic validation integration
• @validate: a decorator that validates function arguments using type annotations

Installation

pip install validating

Quick Start

from validating import attr, dataclass

@dataclass
class UserConfig:
    age: int = attr(lb=0)
    role: str = attr(allowlist=["admin", "user"])

cfg = UserConfig(age=18, role="admin")
cfg.age = 20          # ✅
cfg.role = "user"    # ✅
# cfg.age = -1        # ❌ ValueError
# cfg.role = "guest" # ❌ ValueError

Core API

attr(...)

Declares a descriptor-backed field that validates values on initialization and assignment.

Common parameters:

• default: default value
• default_factory: lazy default factory (mutually exclusive with default)
• allowlist: whitelist of allowed values
• denylist: blacklist of forbidden values
• lb / ub: inclusive lower / upper bounds
• slb / sub: exclusive lower / upper bounds
• validator: custom validator function with signature Callable[[Any], bool]
• init / repr / hash / compare / kw_only: forwarded dataclass field behavior controls

Error semantics:

• Misconfiguration at class-definition time (for example, default type mismatch) raises ValidatorError
• Invalid runtime values raise TypeError or ValueError

---

@dataclass(...)

validating.dataclass is a compatible enhanced wrapper around dataclasses.dataclass.

Enhancements:

1. Automatic default promotion: x: int = 1 is promoted to attr(default=1)
2. Method argument validation: when validate_methods=True (by default False), all public methods are wrapped with validate (including @staticmethod and @classmethod)

from validating import dataclass

@dataclass(validate_methods=True)
class Service:
    retries: int = 3

    def run(self, timeout: int) -> int:
        return timeout + self.retries

svc = Service()
svc.run(1)       # ✅
# svc.run("1")   # ❌ TypeError

---

@validate

Enables call-time argument validation based on type annotations.

from typing import Literal
from validating import validate

@validate
def configure(mode: Literal["dev", "prod"], workers: int):
    return mode, workers

configure("dev", 4)    # ✅
# configure("test", 4)  # ❌ TypeError

@validate also handles assert statements in validated functions:

• assert <expr>, "custom message" keeps the original AssertionError with your message.
• assert <expr> (without message) is converted to a ValueError when <expr> is a direct assertion on function arguments, with a message like expected <expr>, got <value> instead.
• Assertions that are not direct checks on function arguments are left as regular AssertionError.

from validating import validate

@validate
def check_score(score: int) -> int:
    assert score >= 60
    return score

check_score(80)   # ✅
# check_score(59) # ❌ ValueError: expected score >= 60, got 59 instead

More Examples

1) default_factory

from validating import attr, dataclass

@dataclass
class Cache:
    items: list[int] = attr(default_factory=list)

2) Complex type hints

from typing import Literal
from validating import attr, dataclass

@dataclass
class AppConfig:
    mode: Literal["dev", "prod"] = attr()
    token: int | str = attr()
    mapping: dict[str, int] = attr()

3) Custom validator

from validating import attr, dataclass

@dataclass
class EvenNumber:
    value: int = attr(validator=lambda x: x % 2 == 0)

Notes

• Dataclasses with slots=True are currently not supported.
• This project focuses on runtime validation and does not replace static type checking.

See Also

• GitHub: https://github.com/Chitaoji/validating/
• PyPI: https://pypi.org/project/validating/

License

This project falls under the BSD 3-Clause License.

History

v0.0.1

• Initial release.