hi - im running some tests with a

python_tests

target. I would like to add some regression tests that make use of data files (eg json) is there a way to have these included in the test run ? I tried adding `files`/`resources` but

python_tests

doesnt have these params, and if i try to add a non-py file to

sources

that also fails

You need to add a separate target for the data files like this:

resources(
    name="data",
    sources=[
        "**/*.json",
    ],
)

ah, thanks

related to this, is there a way to write your python code such that pants can figure out these sort of dependenies?

We currently have [python-infer].string_imports for strings that look like modules, but it risks false positives (which mean less fine grained cache keys) We probably could do something more intelligent? like look for open() and the different resources APIs

i'm guessing it will have a hard time if you are programmatically constructing paths which we almost always are

True. And I do want to be aware of the risk of over engineering this here. Should stay something performant and predictable Probably no variable analysis 😂

would you consider allowing hints inline in the code? Like in a comment?

Definitely! We support that already for Shell by using Shellcheck comments (we use shellcheck as the import parser), and it works great Would you be interested in opening a GH issue with that idea? Maybe a suggested syntax?

sure

With

pyproject.toml

supported already, it seems like a good place when a project is leaning heavily on inference and wants minimal

BUILD

files.

What do you mean, like centralize config into pyproject.toml?

Yeah. Create a tool.pants section and put meta data that can't be inferred in their.

Given that we're multilingual, I think we'd want to stick with

pants.toml

rather than a Python-specific file like

pyproject.toml

. But definitely that's an option to move more config into

pants.toml

I bias towards it not being the best approach because it doesn't scale well. We generally believe metadata should live near the code it describes. But definitely still worth considering! What do you think are the benefits of putting in a centralized config file?

There is a

dependencies

section, but that is for pip installed dependencies.