tidi

Tidi

A small dependency injection Python library.

Inspired by Kent Tong's disl and FastAPI's Depends.

Motivation

I found myself wanting to learn more about how dependency injection can be done in a pythonic way, with type-hinting, so had the itch to develop (yet another) library for it and share it as open source 🧑‍💻✌️

Read more about the motivation in this Medium article.

Primary API

The top level import of tidi provides everything needed it's primary intended use.

• @tidi.inject - a decorator that will replace certain keyword arguments with dependencies, based on their type & if they haven't been passed in
• tidi.Injected[DependencyClass] - a type alias, wrapping typing.Annotated, that indicates that a keyword argument should be injected
• tidi.register(dependency_instance) - a function that registers an object to be available for injection as a dependency
• tidi.Provider(get_dependency_function) - a wrapper class around a function that will be called to provide a dependency
• tidi.UNSET - a sentinel object to indicate that a dependency should be loaded from the registry
• tidi.field_factory(DependencyClass) - a helper function for injecting dependencies into dataclass fields

Example of use

Consider a micro-sized interactive CLI that lets a user choose a handbag then search through it,

# search-handbag.py

import tidi

from handbags import Handbag, HandbagItem, load_handbag


@tidi.inject
def dig_through_handbag(
    item_type: str,
    handbag: tidi.Injected[Handbag] = tidi.UNSET,
) -> HandbagItem | None:
    return handbag.get_items_by_type(item_type).first(default=None)


def init_handbag():
    selected_handbag = input("Select a handbag: ")
    tidi.register(load_handbag(selected_handbag))


def run_search():
    item_type = input("What are you looking for? ")
    # ⬇️ registered `Handbag` instance gets injected ✨
    item = dig_through_handbag(item_type)
    if item is None:
        print("Uh oh, can't find it 🤷‍♀️, try again")
        run_search()
    else:
        print(f"We're in luck! Here's your {item.name} 😎")


if __name__ == "__main__":
    init_handbag()
    run_search()

Running it looks something like this,

$ python search-handbag
Select a handbag: BCBGMAXAZRIA
What are you looking for? Nail file
Uh oh, can't find it 🤷‍♀️, try again
What are you looking for? Lip balm
We're in luck! Here's your Blistex 😎

We can see Dependency Injection happening hear to achieve Inversion of Control and obey the Law of Demeter.

• dig_through_handbag isn't responsible for creating a Handbag and doesn't require its caller to know about it, rather a Handbag is injected ✨
• init_handbag creates the Handbag, but doesn't need to return it. An example of separating the app initialisation from the main logic.
• run_search doesn't need to know about anything that it doesn't use (in this case the Handbag), obeying the Law of Demeter.

When testing,

• a mock Handbag could be passed in as a keyword argument to test dig_through_handbag, and
• patching dig_through_handbag with a stub could be done to test run_search with no requirement for a mock Handbag.

More examples

You can find some executable examples in the demo/ directory of the repo.

Also see the Usage documentation for more examples.

Interested in contributing?

Feel free to create an issue or author a PR 😊

For the latter, check out the CONTRIBUTING guide for a quick start on development.