typelevel / feral   0.1.0-M9

GitHub

Feral cats are homeless, feral functions are serverless

feral feral-core Scala version support Discord

feral is a framework for writing serverless functions in Scala with Cats Effect and deploying them to the cloud, targeting both JVM and JavaScript runtimes. By providing an idiomatic, purely functional interface, feral is both composable—integrations with natchez and http4s are provided out-of-the-box—and also highly customizable. The initial focus has been on supporting AWS Lambda and will expand to other serverless providers.

Getting started

Feral is published for Scala 2.13 and 3.1+ with artifacts for both JVM and Scala.js 1.8+.

// Scala.js setup
addSbtPlugin("org.typelevel" %% "sbt-feral-lambda" % "0.1.0-M1") // in plugins.sbt
enablePlugins(LambdaJSPlugin) // in build.sbt

// JVM setup
libraryDependencies += "org.typelevel" %% "feral-lambda" % "0.1.0-M1"

// Optional, specialized integrations, available for both JS and JVM
libraryDependencies += "org.typelevel" %%% "feral-lambda-http4s" % "0.1.0-M1"
libraryDependencies += "org.typelevel" %%% "feral-lambda-cloudformation-custom-resource" % "0.1.0-M1"

Next, implement your Lambda. Please refer to the examples for a tutorial.

There are several options to deploy your Lambda. For example you can use the Lambda console, the SAM CLI, or the serverless framework.

To deploy a Scala.js Lambda, you will need to know the following:

  1. The runtime for your Lambda is Node.js 14.
  2. The handler for your Lambda is index.yourLambdaName.
    • index refers to the index.js file containing the JavaScript sources for your Lambda.
    • yourLambdaName is the name of the Scala object you created that extends from IOLambda.
  3. Run sbt npmPackage to package your Lambda for deployment. Note that you can currently only have one Lambda per sbt (sub-)project. If you have multiple, you will need to select the one to deploy using Compile / mainClass := Some("my.lambda.handler").
  4. For the tooling of your choice, follow their instructions for deploying a Node.js Lambda using the contents of the target/scala-2.13/npm-package/ directory.

As the feral project develops, one of the goals is to provide an sbt plugin that simplifies and automates the deployment process. If this appeals to you, please contribute feature requests, ideas, and/or code!

Why go feral?

The premise that you can (and should!) write production-ready serverless functions in Scala targeting JavaScript may be a surprising one. This project—and the rapid maturity of the Typelevel.js ecosystem—is motivated by three ideas.

  1. JavaScript is the ideal compile target for serverless functions.

    There are a lot of reasons for this, cold-start being one of them, but more generally it's important to remember what the JVM is and is not good at. In particular, the JVM excels at long-lived multithreaded applications which are relatively memory-heavy and reply on medium-lifespan heap allocations. So in other words, persistent microservices.

    Serverless functions are, by definition, not this. They are not persistent, they are (generally) single-threaded, and they need to start very quickly with minimal warming. They do often apply moderate-to-significant heap pressure, but this factor is more than outweighed by the others.

    V8 (the JavaScript engine in Node.js) is a very good runtime for these kinds of use-cases. Realistically, it may be the best-optimized runtime in existence for these requirements, similar to how the JVM is likely the best-optimized runtime in existence for the persistent microservices case.

  2. Scala.js and Cats Effect work together to provide powerful, well-defined semantics for writing JavaScript applications.

    It hopefully should not take much convincing that Scala is a fantastic language to use, regardless of the ultimate compile target. But what might be unexpected by those new to Scala.js is how well it preserves Scala's JVM semantics in JavaScript. Save a few edge-cases, by and large Scala programs behave the same on JS as they do on the JVM.

    Cats Effect takes this a step further by establishing semantics for asynchronous programming (aka laws) and guaranteeing them across the JVM and JS. In fact, the initial testing of these semantics on Scala.js revealed a fairness issue that culminated in the deprecation of the default global ExecutionContext in Scala.js. As a replacement, the MacrotaskExecutor project was extracted from Cats Effect and is now the official recommendation for all Scala.js applications. Cats Effect IO is specifically optimized to take advantage of the MacrotaskExecutor's fairness properties while maximizing throughput and performance.

    IO also has features to enrich the observability and debuggability of your JavaScript applications during development. Tracing and enhanced exceptions capture the execution graph of a process in your program, even across asynchronous boundaries, while fiber dumps enable you to introspect the traces of all the concurrent processes in your program at any given time.

  3. Your favorite Typelevel libraries are already designed for Scala.js.

    Thanks to the platform-independent semantics, software built using abstractions from Cats Effect and other Typelevel libraries can often be easily cross-compiled for Scala.js. One spectacular example of this is skunk, a data access library for Postgres that was never intended to target JavaScript. However, due to its whole-hearted adoption of purely functional asynchronous programming, today it also runs on Node.js with virtually no changes to its source code.

    In practice, this means you can directly transfer your knowledge and experience writing Scala for the JVM to writing Scala.js and in many cases share code with your JVM applications. The following libraries offer identical APIs across the JVM and JS platforms: