Extendable

About

Extendable is a module that aims to provide a way to define extensible python classes. It is designed to provide developers with a convenient and flexible way to extend the functionality of their Python applications. By leveraging the "extendable" library, developers can easily create modular and customizable components that can be added or modified without modifying the core codebase. This library utilizes Python's object-oriented programming features and provides a simple and intuitive interface for extending and customizing applications. It aims to enhance code reusability, maintainability, and overall development efficiency. It implements the extension by inheritance and composition pattern. It's inspired by the way odoo implements its models. Others patterns can be used to make your code pluggable and this library doesn't replace them.

Quick start

Let's define a first python class.

from extendable import ExtendableMeta

class Person(metaclass=ExtendableMeta):

    def __init__(self, name: str):
        self.name = name

    def __repr__(self) -> str:
        return self.name

Someone using the module where the class is defined would need to extend the person definition with a firstname field.

from extendable import ExtendableMeta

class PersonExt(Person, extends=Person):
    def __init__(self, name: str):
        super().__init__(name)
        self._firstname = None

    @property
    def firstname(self) -> str:
        return self._firstname

    @firstname.setter
    def firstname(self, value:str) -> None:
        self._firstname = value

    def __repr__(self) -> str:
        res = super().__repr__()
        return f"{res}, {self.firstname or ''}"

At this time we have defined that PersonExt extends the initial definition of Person. To finalyse the process, we need to instruct the runtime that all our class declarations are done by building the final class definitions and making it available into the current execution context.

from extendable import context, registry

_registry = registry.ExtendableClassesRegistry()
context.extendable_registry.set(_registry)
_registry.init_registry()

Once it's done the Person and PersonExt classes can be used interchangeably into your code since both represent the same class...

p = Person("Mignon")
p.firstname = "Laurent"
print (p)
#> Mignon, Laurent

⚠️ This way of extending a predefined behaviour must be used carefully and in accordance with the Liskov substitution principle It doesn't replace others design patterns that can be used to make your code pluggable.

How it works

Behind the scenes, the "extendable" library utilizes several key concepts and mechanisms to enable its functionality. Overall, the "extendable" library leverages metaclasses, registry initialization, and dynamic loading to provide a flexible and modular approach to extending Python classes. By utilizing these mechanisms, developers can easily enhance the functionality of their applications without the need for extensive modifications to the core codebase.

Metaclasses

The metaclass do 2 things.

• It collects the definitions of the declared class and gathers information about its attributes, methods, and other characteristics. These definitions are stored in a global registry by module. This registry is a map of module name to a list of class definitions.
• This information is then used to build a class object that acts as a proxy or blueprint for the actual concrete class that will be created later when the registry is initialized based on the aggregated definition of all the classes declared to extend the initial class...

Registry initialization

The registry initialization is the process that build the final class definition. To make your blueprint class work, you need to initialize the registry. This is done by calling the init_registry method of the registry object. This method will build the final class definition by aggregating all the definitions of the classes declared to extend the initial class through a class hierarchy. The order of the classes in the hierarchy is important. This order is by default the order in which the classes are loaded by the python interpreter. For advanced usage, you can change this order or even the list of definitions used to build the final class definition. This is done by calling the init_registry method with the list of modules¹ to load as argument.

from extendable import registry

_registry = registry.ExtendableClassesRegistry()
_registry.init_registry(["module1", "module2.*"])

Once the registry is initialized, it must be made available into the current execution context so the blueprint class can use it. To do so you must set the registry into the extendable_registry context variable. This is done by calling the set method of the extendable_registry context variable.

from extendable import context, registry

_registry = registry.ExtendableClassesRegistry()
context.extendable_registry.set(_registry)
_registry.init_registry()

Dynamic loading

All of this is made possible by the dynamic loading capabilities of Python. The concept of dynamic loading in Python refers to the ability to load and execute code at runtime, rather than during the initial compilation or execution phase. It allows developers to dynamically import and use modules, classes, functions or variables based on certain conditions or user input. The dynamic loading can also be applied at class instantiation time. This is the mechanism used by the "extendable" library to instantiate the final class definition when you call the constructor of the blueprint class. This is done by implementing the __call__ method into the metaclass to return the final class definition instead of the blueprint class itself. The same applies to pretend that the blueprint class is the final class definition through the implementation of the __subclasscheck__ method into the metaclass.

Development

To run tests, use tox. You will get a test coverage report in htmlcov/index.html. An easy way to install tox is pipx install tox.

This project uses pre-commit to enforce linting (among which black for code formating, isort for sorting imports, and mypy for type checking).

To make sure linters run locally on each of your commits, install pre-commit (pipx install pre-commit is recommended), and run pre-commit install in your local clone of the extendable repository.

To release:

• run ``bumpversion patch|minor|major --list
• Check the new_version value returned by the previous command
• run towncrier build.
• Inspect and commit the updated HISTORY.rst.
• git tag {new_version} ; git push --tags.

Contributing

All kind of contributions are welcome.

Footnotes

1. When you specify a module into the list of modules to load, the wildcart character * is allowed at the end of the module name to load all the submodules of the module. Otherwise, only the module itself is loaded. ↩