Hi, I'm investigating using pants with a Python monorepo and have started by testing pants packaging for AWS lambda and deployment. However, I am encountering an issue when deploying due to the artifact ZIP created by pants being too large. Previously, I used serverless to package for lambda, and had settings to reduce the deployment size. How can I apply these similar settings to the pants packaging step?

Hi! Presumably it's the no-deploying of pandas and numpy part that matters for your case?

Hi Ben, thanks for responding so quickly! Numpy and pandas are being provided with a custom lambda layer

Aha, so Pants 2.18 will support building lambda layers. See here for how this works: https://www.pantsbuild.org/v2.18/docs/awslambda-python

Hi Ben, thanks for your help. I've tried to excluding numpy and pandas in the dependencies field:

python_aws_lambda_function(
    name="x",
    output_path='api/x.zip',
    handler="handlers.py:*",
    dependencies=['api/x/y.py', '!!3rdparty/python:numpy', '!!3rdparty/python:pandas'],
    complete_platforms=["api/lambda_config:aws_lambda_python_3_10"]
)

But I get this error:

InvalidFieldException: api/analytics/BUILD:3: Failed to get dependencies for api/analytics:x: The file or directory '3rdparty/python' does not exist on disk in the workspace, so the address '3rdparty/python:pandas' from the `dependencies` field from the target api/analytics:x cannot be resolved.

Cool