PLEASE READ THIS :

The stewardship of Weaver is being moved to Typelevel. This repository will no longer receive any development, and will be archived after the first release by the Typelevel version. You can follow the progress here : https://github.com/typelevel/weaver-test/issues/1

Weaver-test

A test-framework built on cats-effect and fs2

Installation

Weaver-test is currently published for Scala 2.12, 2.13, and 3.0

SBT

Refer yourself to the releases page to know the latest released version, and add the following (or scoped equivalent) to your build.sbt file.

libraryDependencies += "com.disneystreaming" %% "weaver-cats" % "x.y.z" % Test
testFrameworks += new TestFramework("weaver.framework.CatsEffect")

// optionally (for Scalacheck usage)
libraryDependencies +=  "com.disneystreaming" %% "weaver-scalacheck" % "x.y.z" % Test

Motivation

Weaver aims at providing a nice experience when writing and running tests :

• tests within a suite are run in parallel by default for quickest results possible
• expectations (ie assertions) are composable values. This forces developers to separate the scenario of the test from the checks they perform, which generally keeps tests cleaner / clearer.
• failures are aggregated and reported at the end of the run. This prevents the developer from having to "scroll up" forever when trying to understand what failed.
• a lazy logger is provided for each test, and log statements are only displayed in case of a test failure. This lets the developer enrich their tests with clues and works perfectly well with parallel runs

API

Example suites (cats-effect)

SimpleIOSuite

The suite that is most familiar to developers :

import weaver.SimpleIOSuite
import cats.effect._

// Suites must be "objects" for them to be picked by the framework
object MySuite extends SimpleIOSuite {

  pureTest("non-effectful (pure) test"){
    expect("hello".size == 6)
  }

  private val random = IO(java.util.UUID.randomUUID())

  test("test with side-effects") {
    for {
      x <- random
      y <- random
    } yield expect(x != y)
  }

  loggedTest("test with side-effects and a logger"){ log =>
    for {
      x <- random
      _ <- log.info(s"x : $x")
      y <- random
      _ <- log.info(s"y : $y")
    } yield expect(x != y)
  }
}

IOSuite

The IOSuite constructs the given resource once for all tests in the suite.

import weaver.IOSuite
import cats.effect._

object MySuite extends IOSuite {

  type Res = Int

  def sharedResource : Resource[IO, Int] = Resource
    .make(
      IO(println("Making resource"))
        .as(123)
    )(n => IO(println(s"Closing resource $n")))

  test("test, but resource not visible"){
    IO(expect(123 == 123))
  }

  test("test with resource"){ n =>
    IO(expect(n == 123))
  }

  test("test with resource and a logger"){ (n, log) =>
    log.info("log was available") *>
    IO(expect(n == 123))
  }
}

Other suites

Weaver also includes support for

Expectations (assertions)

Building expectations

The various test functions have in common that they expect the developer to return a value of type Expectations, which is just a basic case class wrapping a cats.data.Validated value.

The most convenient way to build Expectations is to use the expect function. Based on Eugene Yokota's excellent expecty, it captures the boolean expression at compile time and provides useful feedback on what goes wrong when it does :

Nothing prevents the user from building their own expectations functions to resemble what they're used to.

Composing expectations

Something worth noting is that expectations are not throwing, and that if the user wants to perform several checks in the same test, he needs to compose the expectations via the and or the or methods they carry.

Filtering tests

When using the IOSuite variants, the user can call sbt's test command as such:

> testOnly -- -o *foo*

This filter will prevent the execution of any test that doesn't contain the string "foo" in is qualified name. For a test labeled "foo" in a "FooSuite" object, in the package "fooPackage", the qualified name of a test is:

fooPackage.FooSuite.foo

Running suites in standalone

It is possible to run suites outside of your build tool, via a good old main function. To do so, you can instantiate the weaver.Runner, create a fs2.Stream of the suites you want to run, and call runner.run(stream).

This is useful when you consider your tests (typically end-to-end ones) as a program of its own and want to avoid paying the cost of compiling them every time you run them.

Scalacheck (property-based testing)

Weaver comes with basic scalacheck integration.

import weaver._
import weaver.scalacheck._
import org.scalacheck.Gen

// Notice the Checkers mix-in
object ForallExamples extends SimpleIOSuite with Checkers {

  // CheckConfig can be overridden at the test suite level
  override def checkConfig: CheckConfig =
    super.checkConfig.copy(perPropertyParallelism = 100)

  test("Gen form") {
    // Takes an explicit "Gen" instance. There is only a single
    // version of this overload. If you want to pass several Gen instances
    // at once, just compose them monadically.
    forall(Gen.posNum[Int]) { a =>
      expect(a > 0)
    }
  }

  test("Arbitrary form") {
    // Takes a number of implicit "Arbitrary" instances. There are 6 overloads
    // to pass 1 to 6 parameters.
    forall { (a1: Int, a2: Int, a3: Int) =>
      expect(a1 * a2 * a3 == a3 * a2 * a1)
    }
  }

  test("foobar") {
    // CheckConfig can be overridden locally
    forall.withConfig(super.checkConfig.copy(perPropertyParallelism = 1,
                                             initialSeed = Some(7L))) {
      (x: Int) =>
        expect(x > 0)
    }
  }

}

Contributing

Contributions are most welcome !

Development requirements

:warning: This repository uses Git LFS. Having it installed is required for git checkout to work properly.

To install for Mac using brew:

brew install git-lfs
git lfs install

If you want to build and run the website then you will need yarn installed:

brew install yarn

Building the website

If you're changing documentation, here's how you can check your changes locally:

sbt docs/docusaurusCreateSite
cd website
yarn start

If you're only changing .md files, you can run sbt '~docs/mdoc'.

Note that the site will look a tiny bit different because to build a versioned website we have some machinery in the script running on CI - but you don't have to worry about that.

IntelliJ plugin

The code of the IntelliJ plugin lives there

PR Guidelines

Please:

• Sign the CLA
• Write positive and negative tests
• Include documentation

Inspiration

A HUGE thank you to Alexandru Nedelcu, author of Monix and contributor to cats-effect, as he wrote the minitest framework which got this framework started.