Make sense to have bricks that are not python packages ?

[ybenitezf]

Bear with me for a moment:

• Suppose we have an ongoing project with a git repository, on the main branch the component componentA - or a base - exists.
• In separate branch (branchY lest say) we make a refactor that erase this component (some of the functions or classes are moved to other components, just to put in a real live example) and we in this brach - branchY - remove all references to componentA.
• Change to branch main and merge branchY, all good, expect that git will not remove the folder for componentA, and normally it contains some ignore files, and if we run poly sync then componentA will be added again into the root pyproject.toml.

So i think that this function:

python-polylith/components/polylith/bricks/component.py

Lines 18 to 19 in 0c7b67a

	def is_brick_dir(p: Path) -> bool:
	return p.is_dir() and p.name not in {"__pycache__", ".venv", ".mypy_cache"}

should only return valid python packages, i. e. folder that contains a __init__.py directly or in a subfolder, this make sense ?

[DavidVujic]

Hi @ybenitezf!

Oh, so in the feature branch, the actual folder componentA isn't removed? I can imagine scenarios where there is json or other resource data in a component that should be included. But I agree that it would make sense to not include an empty folder in the root pyproject.toml file.

[ybenitezf]

Exactly, git do not remove the refactored component or base folder when you change branches. Maybe the removed folder only contains __pycache__ when branches are switched - that's enough for git to not delete the folder, but when you do poly sync is added to the pyproject.toml

[DavidVujic]

Maybe that scenario could be seen as a mistake in the feature branch? I am guessing that the folder should have been removed entirely if the code as been refactored and moved somewhere else.

I don't want to risk that the poly tool is too restrictive about the contents of a namespace package folder, and might miss some use cases where there is other things than python code in them - and that should be part of the deploys (json, csv etc.).