Nightjar

Description

This project is a Python package that provides a simple way to create objects of different types based on a configuration object. The package is inspired by how huggingface/transformers package creates different types of models based on a configuration object. The package provides a base class BaseModule that can be subclassed to create different class types. Each class type is defined by a configuration class that inherits from BaseConfig. The AutoModule class is used to automatically create instances of the correct object type based on the configuration. The dispatch attribute of the configuration class is used to specify the static attribute that determines the object type.

Installation

To install this package, run the following command:

pip install nightjar

Usage

Example

Let's see the usage of this package with an example.

from typing import ClassVar

from nightjar import AutoModule, BaseConifg, BaseModule


class VehicleConfig(BaseConifg, dispatch=["type"]):
    type: ClassVar[str]


class Vehicle(BaseModule):
    config: VehicleConfig


class AutoVehicle(AutoModule):
    def __new__(cls, config: VehicleConfig) -> Vehicle:
        return super().__new__(cls, config)


class CarConfig(VehicleConfig):
    type: ClassVar[str] = "car"


class Car(Vehicle):
    config: CarConfig


class VanConfig(VehicleConfig):
    type: ClassVar[str] = "van"


class Van(Vehicle):
    config: VanConfig

Explanation

classDiagram
    class BaseConfig {
        <<abstract>>
        +dispatch: ClassVar[str | List[str]]
        +from_dict(cls, data: Dict[str, Any]) -> BaseConfig
        +to_dict(self) -> Dict[str, Any]
    }
    class BaseModule {
        <<abstract>>
        +config: BaseConfig
    }
    class AutoModule {
        +__new__(cls, config: BaseConfig) -> BaseModule
    }
    class VehicleConfig {
        +type: ClassVar[str]
    }
    class Vehicle {
        +config: VehicleConfig
    }
    class AutoVehicle {
        +__new__(cls, config: VehicleConfig) -> Vehicle
    }
    class CarConfig {
        +type: ClassVar[str] = "car"
    }
    class Car {
        +config: CarConfig
    }
    class VanConfig {
        +type: ClassVar[str] = "van"
    }
    class Van {
        +config: VanConfig
    }
    BaseConfig <|-- VehicleConfig
    BaseModule <|-- Vehicle
    AutoModule <|-- AutoVehicle
    VehicleConfig <|-- CarConfig
    VehicleConfig <|-- VanConfig
    Vehicle <|-- Car
    Vehicle <|-- Van

This package provides a base class BaseModule that can be subclassed to create different types of objects. Each object type is defined by a configuration class that inherits from BaseConfig. The AutoModule class is used to automatically create instances of the correct object type based on the configuration. The dispatch attribute of the configuration class is used to specify the static attribute that determines the object type.

AutoModule is a generic class that takes the configuration object as init argument and returns an instance of the correct object type. The __new__ method is used to create the correct object type based on the configuration. It is not necessary to define the __new__ method in the subclass of AutoModule or to subclass AutoModule at all. The AutoModule class can be used directly to create instances of the correct object type. However, subclassing AutoModule can be useful to add additional functionality or to customize the creation of objects and for type hinting.

# use from_dict method to create a configuration object from a dictionary this will automatically create the correct jar config.
config = VehicleConfig.from_dict({"type": "car"})
# Now you can create a car object using the configuration object with Auto* object
car = AutoVehicle(config)
# Now you can access the config object
assert car.config.type == "car", f"expected 'car', got '{car.config.type}'"

You could use AutoJar, Jar, JarConfig instead of AutoModule, BaseModule, BaseConfig respectively if you prefer.

License

Please see the MIT License file for details.