Sharing fixtures

A test fixture is composed of the objects and other artifacts (files, sockets, database connections, etc.) tests use to do their work. When multiple tests need to work with the same fixtures, it is important to try and avoid duplicating the fixture code across those tests. The more code duplication you have in your tests, the greater drag the tests will have on refactoring the actual production code. ScalaTest recommends several techniques to eliminate such code duplication, and provides several traits to help. Each technique is geared towards helping you reduce code duplication without introducing instance vars, shared mutable objects, or other dependencies between tests. Eliminating shared mutable state across tests will make your test code easier to reason about and more amenable for parallel test execution.

The following sections describe these techniques, including explaining the recommended usage for each. But first, here's a table summarizing the options:

Note: Although the techniques described on this page are shown with FlatSpec examples, the techniques can be used in exactly the same way with any style trait. The Scaladoc for each style trait gives corresponding examples for the other style traits.

Calling get-fixture methods

If you need to create the same mutable fixture objects in multiple tests, and don't need to clean them up after using them, the simplest approach is to write one or more get-fixture methods. A get-fixture method returns a new instance of a needed fixture object (or an holder object containing multiple fixture objects) each time it is called. You can call a get-fixture method at the beginning of each test that needs the fixture, storing the returned object or objects in local variables. Here's an example:

package org.scalatest.examples.flatspec.getfixture

import org.scalatest.FlatSpec
import collection.mutable.ListBuffer

class ExampleSpec extends FlatSpec {

  def fixture =
    new {
      val builder = new StringBuilder("ScalaTest is ")
      val buffer = new ListBuffer[String]
    }

  "Testing" should "be easy" in {
    val f = fixture
    f.builder.append("easy!")
    assert(f.builder.toString === "ScalaTest is easy!")
    assert(f.buffer.isEmpty)
    f.buffer += "sweet"
  }

  it should "be fun" in {
    val f = fixture
    f.builder.append("fun!")
    assert(f.builder.toString === "ScalaTest is fun!")
    assert(f.buffer.isEmpty)
  }
}

The “f.” in front of each use of a fixture object provides a visual indication of which objects are part of the fixture, but if you prefer, you can import the the members with “import f._” and use the names directly.

If you need to configure fixture objects differently in different tests, you can pass configuration into the get-fixture method. For example, if you could pass in an initial value for a mutable fixture object as a parameter to the get-fixture method.

Instantiating fixture-context objects

A alternate technique that is especially useful when different tests need different combinations of fixture objects is to define the fixture objects as instance variables of fixture-context objects whose instantiation forms the body of tests. Like get-fixture methods, fixture-context objects are anly appropriate if you don't need to clean up the fixtures after using them.

To use this technique, you define instance variables intialized with fixture objects in traits and/or classes, then in each test instantiate an object that contains just the fixture objects needed by the test. Keep in mind that traits allow you to mix together just the fixture objects needed by each test, whereas classes allow you to pass data in via a constructor to configure the fixture objects. Here's an example in which fixture objects are partitioned into two traits and each test just mixes together the traits it needs:

package org.scalatest.examples.flatspec.fixturecontext

import collection.mutable.ListBuffer
import org.scalatest.FlatSpec

class ExampleSpec extends FlatSpec {

  trait Builder {
    val builder = new StringBuilder("ScalaTest is ")
  }

  trait Buffer {
    val buffer = ListBuffer("ScalaTest", "is")
  }

  // This test needs the StringBuilder fixture
  "Testing" should "be productive" in new Builder {
    builder.append("productive!")
    assert(builder.toString === "ScalaTest is productive!")
  }

  // This test needs the ListBuffer[String] fixture
  "Test code" should "be readable" in new Buffer {
    buffer += ("readable!")
    assert(buffer === List("ScalaTest", "is", "readable!"))
  }

  // This test needs both the StringBuilder and ListBuffer
  it should "be clear and concise" in new Builder with Buffer {
    builder.append("clear!")
    buffer += ("concise!")
    assert(builder.toString === "ScalaTest is clear!")
    assert(buffer === List("ScalaTest", "is", "concise!"))
  }
}

Mixing in OneInstancePerTest

If every test requires the same set of mutable fixture objects, and none require cleanup, one other approach you can take is make them simply vals and mix in trait OneInstancePerTest. If you mix in OneInstancePerTest, each test will be run in its own instance of the Suite, similar to the way JUnit tests are executed. Here's an example:

package org.scalatest.examples.flatspec.oneinstancepertest

import org.scalatest._
import collection.mutable.ListBuffer

class ExampleSuite extends FlatSpec with OneInstancePerTest {

  val builder = new StringBuilder("ScalaTest is ")
  val buffer = new ListBuffer[String]

  "Testing" should "be easy" in {
    builder.append("easy!")
    assert(builder.toString === "ScalaTest is easy!")
    assert(buffer.isEmpty)
    buffer += "sweet"
  }

  it should "be fun" in {
    builder.append("fun!")
    assert(builder.toString === "ScalaTest is fun!")
    assert(buffer.isEmpty)
  }
}

One way to think of OneInstancePerTest is that the entire Suite instance is like a fixture-context object, but with the difference that the test code doesn't run during construction as it does with the real fixture-context object technique. Because this trait emulates JUnit's manner of running tests, this trait can be helpful when porting JUnit tests to ScalaTest. The primary intended use of OneInstancePerTest is to serve as a supertrait for ParallelTestExecution and the path traits, but you can also mix it in directly to help you port JUnit tests to ScalaTest or if you prefer JUnit's approach to test isolation.

Overriding withFixture(NoArgTest)

Although the get-fixture method, fixture-context object, and OneInstancePerTest approaches take care of setting up a fixture at the beginning of each test, they don't address the problem of cleaning up a fixture at the end of the test. If you just need to perform a side-effect at the beginning or end of a test, and don't need to actually pass any fixture objects into the test, you can override withFixture(NoArgTest), one of ScalaTest's lifecycle methods defined in trait AbstractSuite.

Trait Suite's implementation of runTest passes a no-arg test function to withFixture(NoArgTest). It is withFixture's responsibility to invoke that test function. Suite's implementation of withFixture simply invokes the function, like this:

// Default implementation in trait Suite
protected def withFixture(test: NoArgTest) {
  test()
}

You can, therefore, override withFixture to perform setup before and/or cleanup after invoking the test function. If you have cleanup to perform, you should invoke the test function inside a try block and perform the cleanup in a finally clause, because the exception that causes a test to fail will propagate through withFixture back to runTest. (In other words, if the test fails, the test function invoked by withFixture will throw an exception.)

The withFixture method is designed to be stacked, and to enable this, you should always call the super implementation of withFixture, and let it invoke the test function rather than invoking the test function directly. In other words, instead of writing “test()”, you should write “super.withFixture(test)”, like this:

// Your implementation
override def withFixture(test: NoArgTest) {
  // Perform setup
  try {
    super.withFixture(test) // Invoke the test function
  }
  finally {
    // Perform cleanup
  }
}

Here's an example in which withFixture(NoArgTest) is used to take a snapshot of the working directory if a test fails, and and send that information to the reporter:

package org.scalatest.examples.flatspec.noargtest

import java.io.File
import org.scalatest.FlatSpec

class ExampleSpec extends FlatSpec {

  override def withFixture(test: NoArgTest) {

    try super.withFixture(test)
    catch {
      case e: Exception =>
        val currDir = new File(".")
        val fileNames = currDir.list()
        info("Dir snapshot: " + fileNames.mkString(", "))
        throw e
    }
  }

  "This test" should "succeed" in {
    assert(1 + 1 === 2)
  }

  it should "fail" in {
    assert(1 + 1 === 3)
  }
}

Running this version of ExampleSuite in the interpreter in a directory with two files, hello.txt and world.txt would give the following output:

scala> new ExampleSuite execute
ExampleSuite:
This test
- should succeed
- should fail *** FAILED ***
  2 did not equal 3 (<console>:33)
  + Dir snapshot: hello.txt, world.txt 

Note that the NoArgTest passed to withFixture, in addition to an apply method that executes the test, also includes the test name and the config map passed to runTest. Thus you can also use the test name and configuration objects in your withFixture implementation.

Calling loan-fixture methods

If you need to both pass a fixture object into a test and and perform cleanup at the end of the test, you'll need to use the loan pattern. If different tests need different fixtures that require cleanup, you can implement the loan pattern directly by writing loan-fixture methods. A loan-fixture method takes a function whose body forms part or all of a test's code. It creates a fixture, passes it to the test code by invoking the function, then cleans up the fixture after the function returns.

The following example shows three tests that use two fixtures, a database and a file. Both require cleanup after, so each is provided via a loan-fixture method. (In this example, the database is simulated with a StringBuffer.)

package org.scalatest.examples.flatspec.loanfixture

import java.util.concurrent.ConcurrentHashMap

object DbServer { // Simulating a database server
  type Db = StringBuffer
  private val databases = new ConcurrentHashMap[String, Db]
  def createDb(name: String): Db = {
    val db = new StringBuffer
    databases.put(name, db)
    db
  }
  def removeDb(name: String) {
    databases.remove(name)
  }
}

import org.scalatest.FlatSpec
import DbServer._
import java.util.UUID.randomUUID
import java.io._

class ExampleSpec extends FlatSpec {

  def withDatabase(testCode: Db => Any) {
    val dbName = randomUUID.toString
    val db = createDb(dbName) // create the fixture
    try {
      db.append("ScalaTest is ") // perform setup
      testCode(db) // "loan" the fixture to the test
    }
    finally removeDb(dbName) // clean up the fixture
  }

  def withFile(testCode: (File, FileWriter) => Any) {
    val file = File.createTempFile("hello", "world") // create the fixture
    val writer = new FileWriter(file)
    try {
      writer.write("ScalaTest is ") // set up the fixture
      testCode(file, writer) // "loan" the fixture to the test
    }
    finally writer.close() // clean up the fixture
  }

  // This test needs the file fixture
  "Testing" should "be productive" in withFile { (file, writer) =>
    writer.write("productive!")
    writer.flush()
    assert(file.length === 24)
  }

  // This test needs the database fixture
  "Test code" should "be readable" in withDatabase { db =>
    db.append("readable!")
    assert(db.toString === "ScalaTest is readable!")
  }

  // This test needs both the file and the database
  it should "be clear and concise" in withDatabase { db =>
    withFile { (file, writer) => // loan-fixture methods compose
      db.append("clear!")
      writer.write("concise!")
      writer.flush()
      assert(db.toString === "ScalaTest is clear!")
      assert(file.length === 21)
    }
  }
}

As demonstrated by the last test, loan-fixture methods compose. Not only do loan-fixture methods allow you to give each test the fixture it needs, they allow you to give a test multiple fixtures and clean everything up afterwords.

Also demonstrated in this example is the technique of giving each test its own "fixture sandbox" to play in. When your fixtures involve external side-effects, like creating files or databases, it is a good idea to give each file or database a unique name as is done in this example. This keeps tests completely isolated, allowing you to run them in parallel if desired.

Overriding withFixture(OneArgTest)

If all or most tests need the same fixture, you can avoid some of the boilerplate of the loan-fixture method approach by using a fixture.Suite and overriding withFixture(OneArgTest). Each test in a fixture.Suite takes a fixture as a parameter, allowing you to pass the fixture into the test. You must indicate the type of the fixture parameter by specifying FixtureParam, and implement a withFixture method that takes a OneArgTest. This withFixture method is responsible for invoking the one-arg test function, so you can perform fixture set up before, and clean up after, invoking and passing the fixture into the test function.

To enable the stacking of traits that define withFixture(NoArgTest), it is a good idea to let withFixture(NoArgTest) invoke the test function instead of invoking the test function directly. To do so, you'll need to convert the OneArgTest to a NoArgTest. You can do that by passing the fixture object to the toNoArgTest method of OneArgTest. In other words, instead of writing “test(theFixture)”, you'd delegate responsibility for invoking the test function to the withFixture(NoArgTest) method of the same instance by writing:

withFixture(test.toNoArgTest(theFixture))

Here's a complete example:

package org.scalatest.examples.flatspec.oneargtest

import org.scalatest.fixture
import java.io._

class ExampleSpec extends fixture.FlatSpec {

  case class F(file: File, writer: FileWriter)
  type FixtureParam = F

  def withFixture(test: OneArgTest) {
    val file = File.createTempFile("hello", "world") // create the fixture
    val writer = new FileWriter(file)
    try {
      writer.write("ScalaTest is ") // set up the fixture
      withFixture(test.toNoArgTest(F(file, writer))) // "loan" the fixture to the test
    }
    finally writer.close() // clean up the fixture
  }

  "Testing" should "be easy" in { f =>
    f.writer.write("easy!")
    f.writer.flush()
    assert(f.file.length === 18)
  }

  it should "be fun" in { f =>
    f.writer.write("fun!")
    f.writer.flush()
    assert(f.file.length === 17)
  }
}

In this example, the tests actually required two fixture objects, a File and a FileWriter. In such situations you can simply define the FixtureParam type to be a tuple containing the objects, or as is done in this example, a case class containing the objects. For more information on the withFixture(OneArgTest) technique, see the documentation for fixture.Suite.

Mixing in BeforeAndAfter

In all the shared fixture examples shown so far, the activities of creating, setting up, and cleaning up the fixture objects have been performed during the test. This means that if an exception occurs during any of these activities, it will be reported as a test failure. Sometimes, however, you may want setup to happen before the test starts, and cleanup after the test has completed, so that if an exception occurs during setup or cleanup, the entire suite aborts and no more tests are attempted. The simplest way to accomplish this in ScalaTest is to mix in trait BeforeAndAfter. With this trait you can denote a bit of code to run before each test with before and/or after each test each test with after, like this:

package org.scalatest.examples.flatspec.beforeandafter

import org.scalatest._
import collection.mutable.ListBuffer

class ExampleSpec extends FlatSpec with BeforeAndAfter {

  val builder = new StringBuilder
  val buffer = new ListBuffer[String]

  before {
    builder.append("ScalaTest is ")
  }

  after {
    builder.clear()
    buffer.clear()
  }

  "Testing" should "be easy" in {
    builder.append("easy!")
    assert(builder.toString === "ScalaTest is easy!")
    assert(buffer.isEmpty)
    buffer += "sweet"
  }

  it should "be fun" in {
    builder.append("fun!")
    assert(builder.toString === "ScalaTest is fun!")
    assert(buffer.isEmpty)
  }
}

Note that the only way before and after code can communicate with test code is via some side-effecting mechanism, commonly by reassigning instance vars or by changing the state of mutable objects held from instance vals (as in this example). If using instance vars or mutable objects held from instance vals you wouldn't be able to run tests in parallel in the same instance of the test class unless you synchronized access to the shared, mutable state. This is why ScalaTest's ParallelTestExecution trait extends OneInstancePerTest. By running each test in its own instance of the class, each test has its own copy of the instance variables, so you don't need to synchronize. If you mixed ParallelTestExecution into the ExampleSuite above, the tests would run in parallel just fine without any synchronization needed on the mutable StringBuilder and ListBuffer[String] objects.

Although BeforeAndAfter provides a minimal-boilerplate way to execute code before and after tests, it isn't designed to enable stackable traits, because the order of execution would be non-obvious. If you want to factor out before and after code that is common to multiple test suites, you should use trait BeforeAndAfterEach instead, as shown later in the next section, composing fixtures by stacking traits.

Composing fixtures by stacking traits

In larger projects, teams often end up with several different fixtures that test classes need in different combinations, and possibly initialized (and cleaned up) in different orders. A good way to accomplish this in ScalaTest is to factor the individual fixtures into traits that can be composed using the stackable trait pattern. This can be done, for example, by placing withFixture methods in several traits, each of which call super.withFixture. Here's an example in which the StringBuilder and ListBuffer[String] fixtures used in the previous examples have been factored out into two stackable fixture traits named Builder and Buffer:

package org.scalatest.examples.flatspec.composingwithfixture

import org.scalatest._
import collection.mutable.ListBuffer

trait Builder extends AbstractSuite { this: Suite =>

  val builder = new StringBuilder

  abstract override def withFixture(test: NoArgTest) {
    builder.append("ScalaTest is ")
    try super.withFixture(test) // To be stackable, must call super.withFixture
    finally builder.clear()
  }
}

trait Buffer extends AbstractSuite { this: Suite =>

  val buffer = new ListBuffer[String]

  abstract override def withFixture(test: NoArgTest) {
    try super.withFixture(test) // To be stackable, must call super.withFixture
    finally buffer.clear()
  }
}

class ExampleSpec extends FlatSpec with Builder with Buffer {

  "Testing" should "be easy" in {
    builder.append("easy!")
    assert(builder.toString === "ScalaTest is easy!")
    assert(buffer.isEmpty)
    buffer += "sweet"
  }

  it should "be fun" in {
    builder.append("fun!")
    assert(builder.toString === "ScalaTest is fun!")
    assert(buffer.isEmpty)
    buffer += "clear"
  }
}

By mixing in both the Builder and Buffer traits, ExampleSuite gets both fixtures, which will be initialized before each test and cleaned up after. The order the traits are mixed together determines the order of execution. In this case, Builder is "super" to Buffer. If you wanted Buffer to be "super" to Builder, you need only switch the order you mix them together, like this:

class Example2Suite extends Suite with Buffer with Builder

And if you only need one fixture you mix in only that trait:

class Example3Suite extends Suite with Builder

Another way to create stackable fixture traits is by extending the BeforeAndAfterEach and/or BeforeAndAfterAll traits. BeforeAndAfterEach has a beforeEach method that will be run before each test (like JUnit's setUp), and an afterEach method that will be run after (like JUnit's tearDown). Similarly, BeforeAndAfterAll has a beforeAll method that will be run before all tests, and an afterAll method that will be run after all tests. Here's what the previously shown example would look like if it were rewritten to use the BeforeAndAfterEach methods instead of withFixture:

package org.scalatest.examples.flatspec.composingbeforeandaftereach

import org.scalatest._
import collection.mutable.ListBuffer

trait Builder extends BeforeAndAfterEach { this: Suite =>

  val builder = new StringBuilder

  override def beforeEach() {
    builder.append("ScalaTest is ")
    super.beforeEach() // To be stackable, must call super.beforeEach
  }

  override def afterEach() {
    try super.afterEach() // To be stackable, must call super.afterEach
    finally builder.clear()
  }
}

trait Buffer extends BeforeAndAfterEach { this: Suite =>

  val buffer = new ListBuffer[String]

  override def afterEach() {
    try super.afterEach() // To be stackable, must call super.afterEach
    finally buffer.clear()
  }
}

class ExampleSpec extends FlatSpec with Builder with Buffer {

  "Testing" should "be easy" in {
    builder.append("easy!")
    assert(builder.toString === "ScalaTest is easy!")
    assert(buffer.isEmpty)
    buffer += "sweet"
  }

  it should "be fun" in {
    builder.append("fun!")
    assert(builder.toString === "ScalaTest is fun!")
    assert(buffer.isEmpty)
    buffer += "clear"
  }
}

To get the same ordering as withFixture, place your super.beforeEach call at the end of each beforeEach method, and the super.afterEach call at the beginning of each afterEach method, as shown in the previous example. It is a good idea to invoke super.afterEach in a try block and perform cleanup in a finally clause, as shown in the previous example, because this ensures the cleanup code is performed even if super.afterEach throws an exception.

The difference between stacking traits that extend BeforeAndAfterEach versus traits that implement withFixture is that setup and cleanup code happens before and after the test in BeforeAndAfterEach, but at the beginning and end of the test in withFixture. Thus if a withFixture method completes abruptly with an exception, it is considered a failed test. By contrast, if any of the beforeEach or afterEach methods of BeforeAndAfterEach complete abruptly, it is considered a failed suite, which will result in a SuiteAborted event.

Next, learn about sharing tests.