Jot.jl

Streamlines the creation and management of AWS Lambda functions written in Julia

Amazon Web Services does not provide native support for Julia, so functions must be put into docker containers which implement AWS's Lambda API, and uploaded to AWS Elastic Container Registry (ECR). Jot aims to abstract these complexities away, allowing both julia packages and scripts to be turned into low-latency Lambda functions.

Introduction

1. From the JULIA REPL, create a simple script to use as a lambda function...

open("increment_vector.jl", "w") do f
  write(f, "increment_vector(v::Vector{Int}) = map(x -> x + 1, v)")
end

2. ...and turn it into a Responder

increment_responder = get_responder("./increment_vector.jl", :increment_vector, Vector{Int})

3. Create a local docker image that will implement the responder

local_image = create_local_image(increment_responder; image_suffix="increment-vector")

4. Push this local docker image to AWS ECR

remote_image = push_to_ecr!(local_image)

5. Create a lambda function from this remote_image...

increment_vector_lambda = create_lambda_function(remote_image)

6. ... and test it to see if it's working OK

@test run_lambda_function_test(increment_vector_lambda, [2,3,4], [3,4,5]; check_function_state=true) |> first

Alternatively, there is the LambdaComponents type, which collects together local docker images, remote docker images and lambda functions that share the same underlying function code:

lambda_components = create_lambda_components(increment_responder; image_suffix="increment-vector")
lambda_components |> with_remote_image! |> with_lambda_function! |> run_test

Package Features

• Easily create AWS Lambda functions from Julia packages or scripts
• PackageCompiler.jl is used to maximize performance and minimize cold start times
• JSON read/write and error handling is handled by Jot - you just write standard Julia
• Allows easy checking for version consistency - ie, is a given Lambda Function using the correct code?