I really have a weird issue I'm trying to better understand. I have a monorepo with a bunch of modules that are packaged up using python_distribution targets. Those targets have pretty broad sources, i.e.

resources("dist_files", sources=["module_name/**/*", "README.md"])

python_distribution(
    name="dist",
    dependencies=[":dist_files"],
    ...
)

As a result, when a developer creates a new sub-directory somewhere in the module, when they run

pants tailor ::

, it isn't generating the

python_sources

for what I assume is because pants is detecting the files being owned by a target already. So the question is, is this a bug or is there some feature I'm unaware of and should be doing this differently? I'm working around it by commenting out the resources, re-running pants tailor and then uncommenting.

I'm pretty sure it's a feature that it won't tailor for already owned targets.

I know prescribed thing to do here is use actual existing python sources in the deps for the python_dist, but for various reasons that's a relatively high cost. I wish there was a macro that could dynamically find targets and then provide those, i.e.

python_distribution(
    name="dist",
    dependencies=recursive_python_source_targets(),
    ...
)

I suppose you could do

python_sources(sources="module_name/**/*.py")

and depend on that only, and then, if you still want your resources(), do an ignore selector that removes .py from it.

yeah, I think you're right. Would that have any negative side-effects, i.e. worse test caching, etc.?

If you have set https://www.pantsbuild.org/2.19/reference/subsystems/python-infer#imports to

false

it would mean that all sources depend on all other sources grouped together by that generator. But the default is

true

which means pants looks at imports to build the dependency graph.