When Pants packages python dependencies (specifically using

python_awslambda

target) Is there a difference from importing from a modules init vs using a relative import? For example, lets say I have a utils module that I need a single function from. It's listed in utils

__init__.py

so I can just import it like

from utils import helper_function

. But my problem is that the

__init__.py

also imports some other large helper functions with dependencies that I absolutely don't need. (maybe

other_helper_function

requires pandas) I tried to import

helper_function

as a relative import

from ..utils.helper_file import helper_function

but my

python_awslambda

zip file will include

other_helper_function

Is there any way to only import

helper_function

into the zip? Or do I need to explicitly exclude

other_helper_function

in my BUILD file?

Can you remove the reexport of the large function from the init file? Potentially even make it empty If not, you’ll need to do some sort of manual exclusion, but you’ll also need to carefully adjust when the imports happen to avoid failing at runtime with import errors, because the relevant files are being loaded at runtime

I can empty out the init file, just forces me to update everywhere that imports any of those functions. It's probably worth the effort though

Yeah, I’ve personally come to the conclusion that I should just tolerate large numbers of imports from long-ish module paths (relying on editor support to make this easier), in order to better leverage pants dependency inference, which is all file based. It’s not great “locally” for any individual bit of code, but it’s definitely worth it “globally” for the result on the final packages

We mostly do that anyways. Just have a few straggling init files with stuff dumped in there.