rednafi / fastapi-nano Goto Github PK

(I hope I describe this correctly, it is possible that I have a misunderstanding of how imports are resolved)

Problem: The FastAPI documentation on debugging suggests to call uvicorn directly from main via the IDE/Debugger. However, in this case, fastapi-nano’s imports do not seem to work, as the folder of main.py is the root folder for the module imports with this method of executing the code. So from app.core import auth fails, , from core import auth would work. The imports work when calling uvicorn directly since uvicorn app.main:app (as documented in the readme) seems to make the module resolution uses the folder uvicorn was called from as its root (that is, the project’s main folder with the folder app in it)

Possible solutions:

• Assume calling from the /app folder and change paths accordingly
• Document another way to debug
• Dont fix/not a significant problem

Python 3.10 is now stable and the libraries have mostly caught up at this point. Update the docker container to use the latest Python version.

• Use fixtures to patch objects at the module level. Here's an example:

import random
from unittest.mock import patch

import pytest


@pytest.fixture(scope="module")
def mock_randint():
    """Mock random.randint function."""

    with patch("random.randint", return_value=42, auto=True) as m:
        yield m


@pytest.mark.parametrize(
    "seed, output",
    [(1, 42), (100, 42), (589, 42), (444, 42)],
)
def test_randint(mock_randint, seed, output):
    assert random.randint(0, seed) == output
    mock_randint.assert_called_with(0, seed)

• Use pytest.mark.parametrize to make the tests robust.

Hi, I was just browsing through this repo and noticed that the auth routes are declared using async def, in https://github.com/rednafi/fastapi-nano/blob/master/app/core/auth.py#L125 but the underlying database access in the function that is also used as a dependency on line https://github.com/rednafi/fastapi-nano/blob/master/app/core/auth.py#L117 is synchronous. My understanding of FastAPI is such that in this case, it would be better to declare the route and dependency using a simple def, so that the whole thing is handled in a thread and other requests can still be processed concurrently. The way it's currently implemented, I think the entire event loop might be blocked whenever the database is accessed in that route and the related dependency.

Just flagging out of interest, apologies if my understanding is incorrect here.

Would make sense to use starlette.config.Config instead of os.environ + load_dotenv?
Starlette Config does env validation and has builtin load dotenv which would remove the need for python-dotenv.
I could send a pull request if its fine.

fastapi | File "/usr/local/lib/python3.8/site-packages/gunicorn/config.py", line 611, in Workers
fastapi | default = int(os.environ.get("WEB_CONCURRENCY", 1))
fastapi | ValueError: invalid literal for int() with base 10: '1

Private registry credentials can bleed into the requirements file if you run

pip-compile requirements.in -o requirements.txt

Add the proper flags to stop this.

This project is kind of complete in a sense. I don't want to manually merge Dependabot PRs. Enable auto-merging via GitHub action.

Docker compose v2 has been stable for quite a while. Convert docker commands in the makefile to be v2 compatible.

Hi,
I am trying to use live reload for devlepement. I have added below code in Dockerfile as suggested in uvicorn-gunicorn-fastapi-docker. But when i check it still takes start.sh. Any idea? Thank you very much.

RUN pip install -r requirements.txt

# copy project
COPY . /code/

CMD [ "/start-reload.sh" ]  # <-----

# expose port
EXPOSE 5000

If we add how logging can be incorporated, when API calls are made, using middleware it will cover all facets.

Hi, I couldn't seem to find any documentation specifiying how to login with the access token instead of regular auth scheme?

Would fit contributing with helm chart from Kubernetes workloads?

I have a very simple one that is very portable.

What about a GitHub Action that publishes the Docker container to the GitHub repository registry?

• Upgrade to Python 3.12
• Drop Python 3.8 and 3.9 (Supporting only the last 3 Python versions)
• Upgrade to Pydantic v2
• Update FastAPI
• Fix outdated instructions in the README
• Build all the containers in the CI

how do I replace the fake Database with a real Database like the SQLite DB with the user models in it

https://github.com/charliermarsh/ruff