Skip to content

Some useful decorators for any situation. Includes runtime type checking.

License

Notifications You must be signed in to change notification settings

LostInDarkMath/pedantic-python-decorators

Repository files navigation

pedantic-python-decorators Build Status Coverage Status PyPI version Conda Version Last Commit Stars Open Issues Open PRs

This packages includes many decorators that will make you write cleaner Python code.

Getting Started

This package requires Python 3.11 or later. There are multiple options for installing this package.

Option 1: Installing with pip from Pypi

Run pip install pedantic.

Option 2: Installing with conda from conda-forge

Run conda install -c conda-forge pedantic

Option 3: Installing with pip and git

  1. Install Git if you don't have it already.
  2. Run pip install git+https://github.com/LostInDarkMath/pedantic-python-decorators.git@master

Option 4: Offline installation using wheel

  1. Download the latest release here by clicking on pedantic-python-decorators-x.y.z-py-none-any.whl.
  2. Execute pip install pedantic-python-decorators-x.y.z-py3-none-any.whl.

The @pedantic decorator - Type checking at runtime

The @pedantic decorator does the following things:

  • The decorated function can only be called by using keyword arguments. Positional arguments are not accepted.
  • The decorated function must have type annotations.
  • Each time the decorated function is called, pedantic checks that the passed arguments and the return value of the function matches the given type annotations. As a consequence, the arguments are also checked for None, because None is only a valid argument, if it is annotated via typing.Optional.

In a nutshell: @pedantic raises an PedanticException if one of the following happened:

  • The decorated function is called with positional arguments.
  • The function has no type annotation for their return type or one or more parameters do not have type annotations.
  • A type annotation is incorrect.
  • A type annotation misses type arguments, e.g. typing.List instead of typing.List[int].

Minimal example

from pedantic import pedantic


@pedantic
def get_sum_of(values: list[int | float]) -> int:
    return sum(values)

get_sum_of(values=[0, 1.2, 3, 5.4])  # this raises the following runtime error: 
# Type hint of return value is incorrect: Expected type <class 'int'> but 10.0 of type <class 'float'> was the return value which does not match.

The @validate decorator

As the name suggests, with @validate you are able to validate the values that are passed to the decorated function. That is done in a highly customizable way. But the highest benefit of this decorator is that it makes it extremely easy to write decoupled easy testable, maintainable and scalable code. The following example shows the decoupled implementation of a configurable algorithm with the help of @validate:

import os
from dataclasses import dataclass

from pedantic import validate, ExternalParameter, overrides, Validator, Parameter, Min, ReturnAs


@dataclass(frozen=True)
class Configuration:
    iterations: int
    max_error: float


class ConfigurationValidator(Validator):
    @overrides(Validator)
    def validate(self, value: Configuration) -> Configuration:
        if value.iterations < 1 or value.max_error < 0:
            self.raise_exception(msg=f'Invalid configuration: {value}', value=value)

        return value


class ConfigFromEnvVar(ExternalParameter):
    """ Reads the configuration from environment variables. """

    @overrides(ExternalParameter)
    def has_value(self) -> bool:
        return 'iterations' in os.environ and 'max_error' in os.environ

    @overrides(ExternalParameter)
    def load_value(self) -> Configuration:
        return Configuration(
            iterations=int(os.environ['iterations']),
            max_error=float(os.environ['max_error']),
        )


class ConfigFromFile(ExternalParameter):
    """ Reads the configuration from a config file. """

    @overrides(ExternalParameter)
    def has_value(self) -> bool:
        return os.path.isfile('config.csv')

    @overrides(ExternalParameter)
    def load_value(self) -> Configuration:
        with open(file='config.csv', mode='r') as file:
            content = file.readlines()
            return Configuration(
                iterations=int(content[0].strip('\n')),
                max_error=float(content[1]),
            )


# choose your configuration source here:
@validate(ConfigFromEnvVar(name='config', validators=[ConfigurationValidator()]), strict=False, return_as=ReturnAs.KWARGS_WITH_NONE)
# @validate(ConfigFromFile(name='config', validators=[ConfigurationValidator()]), strict=False)

# with strict_mode = True (which is the default)
# you need to pass a Parameter for each parameter of the decorated function
# @validate(
#     Parameter(name='value', validators=[Min(5, include_boundary=False)]),
#     ConfigFromFile(name='config', validators=[ConfigurationValidator()]),
# )
def my_algorithm(value: float, config: Configuration) -> float:
    """
        This method calculates something that depends on the given value with considering the configuration.
        Note how well this small piece of code is designed:
            - Fhe function my_algorithm() need a Configuration but has no knowledge where this come from.
            - Furthermore, it doesn't care about parameter validation.
            - The ConfigurationValidator doesn't know anything about the creation of the data.
            - The @validate decorator is the only you need to change, if you want a different configuration source.
    """
    print(value)
    print(config)
    return value


if __name__ == '__main__':
    # we can call the function with a config like there is no decorator.
    # This makes testing extremely easy: no config files, no environment variables or stuff like that
    print(my_algorithm(value=2, config=Configuration(iterations=3, max_error=4.4)))

    os.environ['iterations'] = '12'
    os.environ['max_error'] = '3.1415'

    # but we also can omit the config and load it implicitly by our custom Parameters
    print(my_algorithm(value=42.0))

List of all decorators in this package

List of all mixins in this package

Dependencies

There are no hard dependencies. But if you want to use some advanced features you need to install the following packages:

  • multiprocess if you want to use the @in_subprocess decorator
  • flask if you want to you the request validators which are designed for Flask (see unit tests for examples):
    • FlaskParameter (abstract class)
    • FlaskJsonParameter
    • FlaskFormParameter
    • FlaskPathParameter
    • FlaskGetParameter
    • FlaskHeaderParameter
    • GenericFlaskDeserializer

Contributing

Feel free to contribute by submitting a pull request :)

Acknowledgments

Risks and side effects

The usage of decorators may affect the performance of your application. For this reason, I would highly recommend you to disable the decorators if your code runs in a productive environment. You can disable pedantic by set an environment variable:

export ENABLE_PEDANTIC=0

You can also disable or enable the environment variables in your project by calling a method:

from pedantic import enable_pedantic, disable_pedantic
enable_pedantic()

Issues with compiled Python code

This package is not compatible with compiled source code (e.g. with Nuitka). That's because it uses the inspect module from the standard library which will raise errors like OSError: could not get source code in case of compiled source code.

Don't forget to check out the documentation. Happy coding!