Diy

I really like the Laravel Service Container. diy is my attempt at recreating something in Python. After working with FastAPI, I notices some shortcomings in their DI system. Other solutions in the Python ecosystem also did not satisfy me. Most of the issues I had are described in the guiding principles section of diys documentation, though no libraries are explicitly mentioned.

Quickstart

There are two important parts in diy. Specifications contain explicit instructions about how to resolve function parameters, construct types or map protocols/interfaces to a concrete implementation.

import diy

spec = diy.Specification()

This is our example class, which only takes one parameter

class ApiClient:
  def __init__(self, token: str):
    self.token = token

  def get(self, url: str):
    # imagine an actual implementation here using the token

This teaches our container how to build an ApiClient

import os

@spec.builders.decorate
def build_api_client() -> ApiClient:
  return ApiClient(token=os.environ["API_TOKEN"])

The container retrieves a spec, and uses the contained instructions to construct types

container = diy.RuntimeContainer(spec)

This is how you get an instance

api_client = diy.resolve(ApiClient)

You can also call functions by letting their parameters be resolved from the container

def get_data(api: ApiClient):
  return api.imaginary_method("")
result = diy.call(get_data)

Notes

• focus on good error messages
• dependency visualization
• autowiring
• validation