The Annotations trait provides access to an annotation map that tests can add arbitrary annotations to. Each annotation consists of a string identifier, an initial value, and a function for combining two values. Annotations form monoids and you can think of Annotations as a more structured logging service or as a super polymorphic version of the writer monad effect.

An Assertion[A] is capable of producing assertion results on an A. As a proposition, assertions compose using logical conjunction and disjunction, and can be negated.

An AssertionM[A] is capable of producing assertion results on an A. As a proposition, assertions compose using logical conjunction and disjunction, and can be negated.

An AssertionValue keeps track of a assertion and a value, existentially hiding the type. This is used internally by the library to provide useful error messages in the event of test failures.

CustomAssertion allows users to create their own custom assertions for use in assertTrue. They are constructed with CustomAssertion.make.

// Definition
sealed trait Pet
case class Dog(hasBone: Boolean) extends Pet
case class Fish(bubbles: Double) extends Pet
case class Cat(livesRemaining: Int) extends Color

val lives =
  CustomAssertion.make[Pet] {
    case Cat(livesRemaining) => Right(livesRemaining)
    case other => Left(s"Expected $$other to be Cat")
  }

// Usage
suite("custom assertions")(
  test("as even") {
    val pet: Option[Pet] = Some(Cat(8))
    assertTrue(pet.is(_.some.custom(lives)) == 8)
  }
)

A value of type Eql[A, B] provides implicit evidence that two values with types A and B could potentially be equal, that is, that A is a subtype of B or B is a subtype of A.

A Gen[R, A] represents a generator of values of type A, which requires an environment R. Generators may be random or deterministic.

The Live trait provides access to the "live" environment from within the test environment for effects such as printing test results to the console or timing out tests where it is necessary to access the real environment.

The easiest way to access the "live" environment is to use the live method with an effect that would otherwise access the test environment.

import zio.Clock
import zio.test._

val realTime = live(Clock.nanoTime)

The withLive method can be used to apply a transformation to an effect with the live environment while ensuring that the effect itself still runs with the test environment, for example to time out a test. Both of these methods are re-exported in the environment package for easy availability.

A Spec[R, E, T] is the backbone of _ZIO Test_. Every spec is either a suite, which contains other specs, or a test of type T. All specs require an environment of type R and may potentially fail with an error of type E.

A TestAspect is an aspect that can be weaved into specs. You can think of an aspect as a polymorphic function, capable of transforming one test into another, possibly enlarging the environment or error type.

TestClock makes it easy to deterministically and efficiently test effects involving the passage of time.

Instead of waiting for actual time to pass, sleep and methods implemented in terms of it schedule effects to take place at a given clock time. Users can adjust the clock time using the adjust and setTime methods, and all effects scheduled to take place on or before that time will automatically be run in order.

For example, here is how we can test ZIO#timeout using TestClock:

import zio.ZIO
import zio.test.TestClock

for {
  fiber  <- ZIO.sleep(5.minutes).timeout(1.minute).fork
  _      <- TestClock.adjust(1.minute)
  result <- fiber.join
} yield result == None

Note how we forked the fiber that sleep was invoked on. Calls to sleep and methods derived from it will semantically block until the time is set to on or after the time they are scheduled to run. If we didn't fork the fiber on which we called sleep we would never get to set the time on the line below. Thus, a useful pattern when using TestClock is to fork the effect being tested, then adjust the clock time, and finally verify that the expected effects have been performed.

For example, here is how we can test an effect that recurs with a fixed delay:

import zio.Queue
import zio.test.TestClock

for {
  q <- Queue.unbounded[Unit]
  _ <- q.offer(()).delay(60.minutes).forever.fork
  a <- q.poll.map(_.isEmpty)
  _ <- TestClock.adjust(60.minutes)
  b <- q.take.as(true)
  c <- q.poll.map(_.isEmpty)
  _ <- TestClock.adjust(60.minutes)
  d <- q.take.as(true)
  e <- q.poll.map(_.isEmpty)
} yield a && b && c && d && e

Here we verify that no effect is performed before the recurrence period, that an effect is performed after the recurrence period, and that the effect is performed exactly once. The key thing to note here is that after each recurrence the next recurrence is scheduled to occur at the appropriate time in the future, so when we adjust the clock by 60 minutes exactly one value is placed in the queue, and when we adjust the clock by another 60 minutes exactly one more value is placed in the queue.

TestConsole provides a testable interface for programs interacting with the console by modeling input and output as reading from and writing to input and output buffers maintained by TestConsole and backed by a Ref.

All calls to print and printLine using the TestConsole will write the string to the output buffer and all calls to readLine will take a string from the input buffer. To facilitate debugging, by default output will also be rendered to standard output. You can enable or disable this for a scope using debug, silent, or the corresponding test aspects.

TestConsole has several methods to access and manipulate the content of these buffers including feedLines to feed strings to the input buffer that will then be returned by calls to readLine, output to get the content of the output buffer from calls to print and printLine, and clearInput and clearOutput to clear the respective buffers.

Together, these functions make it easy to test programs interacting with the console.

import zio.Console._
import zio.test.TestConsole
import zio.ZIO

val sayHello = for {
  name <- readLine
  _    <- printLine("Hello, " + name + "!")
} yield ()

for {
  _ <- TestConsole.feedLines("John", "Jane", "Sally")
  _ <- ZIO.collectAll(List.fill(3)(sayHello))
  result <- TestConsole.output
} yield result == Vector("Hello, John!\n", "Hello, Jane!\n", "Hello, Sally!\n")

TestRandom allows for deterministically testing effects involving randomness.

TestRandom operates in two modes. In the first mode, TestRandom is a purely functional pseudo-random number generator. It will generate pseudo-random values just like scala.util.Random except that no internal state is mutated. Instead, methods like nextInt describe state transitions from one random state to another that are automatically composed together through methods like flatMap. The random seed can be set using setSeed and TestRandom is guaranteed to return the same sequence of values for any given seed. This is useful for deterministically generating a sequence of pseudo-random values and powers the property based testing functionality in ZIO Test.

In the second mode, TestRandom maintains an internal buffer of values that can be "fed" with methods such as feedInts and then when random values of that type are generated they will first be taken from the buffer. This is useful for verifying that functions produce the expected output for a given sequence of "random" inputs.

import zio.Random
import zio.test.TestRandom

for {
  _ <- TestRandom.feedInts(4, 5, 2)
  x <- Random.nextIntBounded(6)
  y <- Random.nextIntBounded(6)
  z <- Random.nextIntBounded(6)
} yield x + y + z == 11

TestRandom will automatically take values from the buffer if a value of the appropriate type is available and otherwise generate a pseudo-random value, so there is nothing you need to do to switch between the two modes. Just generate random values as you normally would to get pseudo-random values, or feed in values of your own to get those values back. You can also use methods like clearInts to clear the buffer of values of a given type so you can fill the buffer with new values or go back to pseudo-random number generation.

A TestRunner[R, E] encapsulates all the logic necessary to run specs that require an environment R and may fail with an error E. Test runners require a test executor, a runtime configuration, and a reporter.

TestSystem supports deterministic testing of effects involving system properties. Internally, TestSystem maintains mappings of environment variables and system properties that can be set and accessed. No actual environment variables or system properties will be accessed or set as a result of these actions.

import zio.system
import zio.test.TestSystem

for {
  _      <- TestSystem.putProperty("java.vm.name", "VM")
  result <- system.property("java.vm.name")
} yield result == Some("VM")

A ZSpec[R, E] is the canonical spec for testing ZIO programs. The spec's test type is a ZIO effect that requires an R and might fail with an E.

Syntax for writing test like

object MySpec extends DefaultMutableRunnableSpec {
  suite("foo") {
    test("name") {
    } @@ ignore

    test("name 2")
  }
  suite("another suite") {
    test("name 3")
  }
}

Syntax for writing test like

object MySpec extends MutableRunnableSpec(layer, aspect) {
  suite("foo") {
    test("name") {
    } @@ ignore

    test("name 2")
  }
  suite("another suite") {
    test("name 3")
  }
}