Numpy and pandas not usable in Lambda

[ebasanez]

As AWS Lambda requires numpy and pandas are compiled in AWS compatible Linux images, and this action uses default GitHub action Ubuntu Linux image, adding numpy and pandas to requirements.txt results in non-compatible versions of those libraries, and the subsequent errors when executing the Lambda code.

This should usually be worked arround by removing both numpy and pandas form requirements.txt of the project and creating and applying an additional layer with compatible versions of the lilbraries. In this case this is not a valid solution, as github action method update_function_layers() only add the newly created layer to the new Lambda function version.

There are several ways to solve this:

• Execute github action in on AWS compatible Linux image (like lambci/lambda:python-build3.8), by defining a job and runs-on.
• Allow an optional parameter INPUT_LAMBDA_ADDITIONAL_LAYERS to be used in method update_function_layers() this way:

update_function_layers(){ echo "Using the layer in the function..." aws lambda update-function-configuration --function-name "${INPUT_LAMBDA_FUNCTION_NAME}" --layers "${INPUT_LAMBDA_LAYER_ARN}:${LAYER_VERSION} ${INPUT_LAMBDA_ADDITIONAL_LAYERS }" }

• In the former method, obtain LATEST Lambda function version layers and add those (all but the newly created one) to the new Lambda function version.

[magicyoda]

Hello @ebasanez , I'm exactly facing this issue with numpy and would like to follow your first 'way to solve it' : "Execute github action in on AWS compatible Linux image (like lambci/lambda:python-build3.8), by defining a job and runs-on.".

Unfortunately, I cannot figure out how to do so. How should I adapt the base .yml file to use the lambci/lambda image to run on? Thank you very much for your help.