Where should schemas live in Python packages?

[t-mart]

I'm writing a CLI tool that will read a configuration file that should conform to a prescribed schema. I've written this schema to a file, and I am trying to figure out:

• Where should it live in my project's directory structure?
• How can I locate it from within code? I'm hoping this would not be a __file__ relative path because that would depend on the path of the locating module. Maybe there is some semblance of a "package root"?
• Since it is not code, will my package manager (I use Poetry) need to be aware of this file to include it for distribution?
• And finally, will such a solution work in both development (e.g. testing things inside the source code directory) and in installed (e.g. user's pip install) use cases?

Thanks!

[sbrunner]

Personally, I put it in the code and use pkgutil.get_data("<module>", "<file>") to get it.
With Poetry I add:

include = [
    "<module>/*.json",
]

In the [tool.poetry] of the pyproject.toml.

I didn't find any issue with dev or production mode with this solution :-)

[Julian]

This sounds like decent advice to me too. The only thing I'd add is importlib-resources is technically the newer replacement for that pkgutil API.

You can see an example of using it for precisely this use case in jsonschema itself, where it's used to load (meta-)schemas in a private API.