Ideas for another usecase

[gwdekker]

Hey! I would like to do the following: for a certain project, I want to do a strict dependency check. What i mean by that: i want to verify that all components in the projects' pyproject.toml are strictly needed by the bases that are imported in that same project. As I often see that after a refactoring, certain bricks are not needed anymore but AFAIK these are not being picked up by any of the poly commands so far.

I could imagine this being already a flag on poly check, something like poly check --strict-components. Is that something you would consider?

Second, related question: when trying to figure out how to build this myself i found poly deps --brick which I could already make use of. However, to make that easy I would like to be able to do the following:

1. run the cli commands directly from python, not from the command line
2. get the output of this command in a machine friendly format instead of a human friendly format.
   Any pointers on how to achieve that?

Thanks in advance!

[DavidVujic]

Great ideas @gwdekker!

When developing poly sync I considered to also remove seemingly unused bricks. But didn't want to risk removing something that should be there even if not used by other bricks. What do you think about me developing something like poly sync --strict (or a more suitable flag), that would add only used bricks into the pyproject.toml?

[DavidVujic]

About automation/an sdk, I'll give it some thought! Meanwhile, here's an approach that can be used already today:
#337

[gwdekker]

Thanks!
After playing around I think I will go for an approach like below:

Write a function (using poly internals) to find all components in a project that are not needed by its bases

from pathlib import Path
from polylith_cli.polylith import info, deps
from polylith_cli.polylith.commands.deps import get_bases

root_path = Path("xxx")
assert root_path.is_dir()


def poly_components_not_needed_by_bases_per_project(root: Path) -> dict:
    root = root_path
    ns: str = "xxx"
    projects_data = info.get_projects_data(root, ns)
    components_not_needed_by_bases = dict()
    for project in projects_data:
        if project["type"] != "project":
            continue
        bases = get_bases(root, ns, project)
        imports = flatten_imports(deps.get_brick_imports(root, ns, bases, components={}))
        components_not_needed_by_bases[project["name"]] = set(project["components"]) - imports
    return components_not_needed_by_bases

def flatten_imports(imports: dict) -> set:
    return set().union(*imports["bases"].values()) | set().union(*imports["components"].values())

print(poly_components_not_needed_by_bases_per_project(root_path))

Add an ignorelist option

in workspace.toml i will add a table where users can per project add a list of components they need in their projects that are not dependencies of their bases.

Something like this:

[tool.polylith.projects.extra_bricks_needed]
# Linting checks per project if the bricks listed are (transitive) dependencies of their bases. If you for some reason
# need to add a component that is not a dependency of the bases, you can add it here as an exception, with your project name as key and your component names as a list
project_name = ["component1", "component2"]

run this through either a unit test or a pre-commit hook

To be decided.

Curious what you think about this approach!

[DavidVujic]

Nice!

Looks like it will work, but I want to highlight that it is just below the public "API" of the CLI, and the code might change between versions (such as function input params and naming).

I will add something to the CLI to extract this kind of information later on this month (currently on vacay with the family).

[gwdekker]

Yes, thanks, that risk is understood! Thanks for the quick feedback, and enjoy your holidays!

[DavidVujic]

Ongoing work, extending poly check to identify unused bricks: #360

[DavidVujic]

Merged and published! 😄