science

Allowing experiments in program logic to be studied and tested.

This library attempts to offer a solution for testing experimental code branches / paths in your real world applications with real world data and performance, without costing you the hassle of breaking existing data or code. It has been inspired by the github/scientist library.

Usage

Add the following to your project's build definition. Note: There is a 0.0.1 version published if you'd like to try it out right now.

resolvers += "bintray-adam-open-source-releases" at "https://dl.bintray.com/adamdecaf/open-source"
libraryDependencies ++= Seq(
  "org.decaf" %% "science-core" % "1.0.1"
)

import org.decaf.science._

// You'll want to pick a different storage
val storage = new InMemoryStorageStrategy[String]
val experiment = Experiment(storage)

// Your application logic
def old(): Int = ???
def new(): Double = ???

val result: Int = experiment(old(), new())

How it works

Two blocks of code are executed, which are a control and an "experiment". Each is saved off to a storage medium for later analysis. This allows you to compare directly the differences between the changes made in your application code. In order to maintain you existing types and flow the control block is returned so your application is able to continue on unchanged.

Storage Strategies

• fs (raw filesystem storage)
• In Memory (useful for basic testing) -- not threadsafe

Issues

Please make a github issue with your issue. I'll work with you to get a fix implemented and released. Thanks!

Contributing

1. Make a github issue or ping me on twitter (@adamdecaf) so we don't duplicate work.
2. Fork project to your organization / account
3. Submit PR with passing tests.
4. I'll merge it!

Contributors

• Adam Shannon (@adamdecaf)

License

Apache 2.0 - See LICENSE