PathAnyFunSpec

A sister class to org.scalatest.funspec.AnyFunSpec that isolates tests by running each test in its own instance of the test class, and for each test, only executing the path leading to that test.

Class PathAnyFunSpec behaves similarly to class org.scalatest.funspec.AnyFunSpec, except that tests are isolated based on their path. The purpose of PathAnyFunSpec is to facilitate writing specification-style tests for mutable objects in a clear, boilerpate-free way. To test mutable objects, you need to mutate them. Using a path class, you can make a statement in text, then implement that statement in code (including mutating state), and nest and combine these test/code pairs in any way you wish. Each test will only see the side effects of code that is in blocks that enclose the test. Here's an example:

import org.scalatest.funspec
import org.scalatest.matchers.should.Matchers
import scala.collection.mutable.ListBuffer

class ExampleSpec extends funspec.PathAnyFunSpec with Matchers {

  describe("A ListBuffer") {

    val buf = ListBuffer.empty[Int] // This implements "A ListBuffer"

    it("should be empty when created") {

      // This test sees:
      //   val buf = ListBuffer.empty[Int]
      // So buf is: ListBuffer()

      buf should be ('empty)
    }

    describe("when 1 is appended") {

      buf += 1 // This implements "when 1 is appended", etc...

      it("should contain 1") {

        // This test sees:
        //   val buf = ListBuffer.empty[Int]
        //   buf += 1
        // So buf is: ListBuffer(1)

        buf.remove(0) should equal (1)
        buf should be ('empty)
      }

      describe("when 2 is appended") {

        buf += 2

        it("should contain 1 and 2") {

          // This test sees:
          //   val buf = ListBuffer.empty[Int]
          //   buf += 1
          //   buf += 2
          // So buf is: ListBuffer(1, 2)

          buf.remove(0) should equal (1)
          buf.remove(0) should equal (2)
          buf should be ('empty)
        }

        describe("when 2 is removed") {

          buf -= 2

          it("should contain only 1 again") {

            // This test sees:
            //   val buf = ListBuffer.empty[Int]
            //   buf += 1
            //   buf += 2
            //   buf -= 2
            // So buf is: ListBuffer(1)

            buf.remove(0) should equal (1)
            buf should be ('empty)
          }
        }

        describe("when 3 is appended") {

          buf += 3

          it("should contain 1, 2, and 3") {

            // This test sees:
            //   val buf = ListBuffer.empty[Int]
            //   buf += 1
            //   buf += 2
            //   buf += 3
            // So buf is: ListBuffer(1, 2, 3)

            buf.remove(0) should equal (1)
            buf.remove(0) should equal (2)
            buf.remove(0) should equal (3)
            buf should be ('empty)
          }
        }
      }

      describe("when 88 is appended") {

        buf += 88

        it("should contain 1 and 88") {

          // This test sees:
          //   val buf = ListBuffer.empty[Int]
          //   buf += 1
          //   buf += 88
          // So buf is: ListBuffer(1, 88)

          buf.remove(0) should equal (1)
          buf.remove(0) should equal (88)
          buf should be ('empty)
        }
      }
    }

    it("should have size 0 when created") {

      // This test sees:
      //   val buf = ListBuffer.empty[Int]
      // So buf is: ListBuffer()

      buf should have size 0
    }
  }
}

Note that the above class is organized by writing a bit of specification text that opens a new block followed by, at the top of the new block, some code that "implements" or "performs" what is described in the text. This is repeated as the mutable object (here, a ListBuffer), is prepared for the enclosed tests. For example:

describe("A ListBuffer") {
  val buf = ListBuffer.empty[Int]

Or:

describe("when 2 is appended") {
  buf += 2

Note also that although each test mutates the ListBuffer, none of the other tests observe those side effects:

it("should contain 1") {

  buf.remove(0) should equal (1)
  // ...
}

describe("when 2 is appended") {

  buf += 2

  it("should contain 1 and 2") {

    // This test does not see the buf.remove(0) from the previous test,
    // so the first element in the ListBuffer is again 1
    buf.remove(0) should equal (1)
    buf.remove(0) should equal (2)

This kind of isolation of tests from each other is a consequence of running each test in its own instance of the test class, and can also be achieved by simply mixing OneInstancePerTest into a regular org.scalatest.funspec.AnyFunSpec. However, PathAnyFunSpec takes isolation one step further: a test in a PathAnyFunSpec does not observe side effects performed outside tests in earlier blocks that do not enclose it. Here's an example:

describe("when 2 is removed") {

  buf -= 2

  // ...
}

describe("when 3 is appended") {

  buf += 3

  it("should contain 1, 2, and 3") {

    // This test does not see the buf -= 2 from the earlier "when 2 is removed" block,
    // because that block does not enclose this test, so the second element in the
    // ListBuffer is still 2
    buf.remove(0) should equal (1)
    buf.remove(0) should equal (2)
    buf.remove(0) should equal (3)

Running the full ExampleSpec, shown above, in the Scala interpeter would give you:

scala> import org.scalatest._
import org.scalatest._

scala> run(new ExampleSpec)
ExampleSpec:
A ListBuffer
- should be empty when created
  when 1 is appended
  - should contain 1
    when 2 is appended
    - should contain 1 and 2
      when 2 is removed
      - should contain only 1 again
      when 3 is appended
      - should contain 1, 2, and 3
    when 88 is appended
    - should contain 1 and 88
- should have size 0 when created

Note: class PathAnyFunSpec's approach to isolation was inspired in part by the specsy framework, written by Esko Luontola.

Shared fixtures

A test fixture is objects or other artifacts (such as files, sockets, database connections, etc.) used by tests to do their work. If a fixture is used by only one test, then the definitions of the fixture objects can be local to the method. If multiple tests need to share an immutable fixture, you can simply assign them to instance variables. If multiple tests need to share mutable fixture objects or vars, there's one and only one way to do it in a PathAnyFunSpec: place the mutable objects lexically before the test. Any mutations needed by the test must be placed lexically before and/or after the test. As used here, "Lexically before" means that the code needs to be executed during construction of that test's instance of the test class to reach the test (or put another way, the code is along the "path to the test.") "Lexically after" means that the code needs to be executed to exit the constructor after the test has been executed.

The reason lexical placement is the one and only one way to share fixtures in a PathAnyFunSpec is because all of its lifecycle methods are overridden and declared final. Thus you can't mix in BeforeAndAfter or BeforeAndAfterEach, because both override runTest, which is final in a PathAnyFunSpec. You also can't override withFixture, because PathAnyFunSpec extends Suite not TestSuite, where withFixture is defined. In short:

The reason the life cycle methods are final, by the way, is to prevent users from attempting to combine a PathAnyFunSpec's approach to isolation with other ways ScalaTest provides to share fixtures or execute tests, because doing so could make the resulting test code hard to reason about. A PathAnyFunSpec's execution model is a bit magical, but because it executes in one and only one way, users should be able to reason about the code. To help you visualize how a PathAnyFunSpec is executed, consider the following variant of ExampleSpec that includes print statements:

import org.scalatest.funspec
import org.scalatest.matchers.Matchers
import scala.collection.mutable.ListBuffer

class ExampleSpec extends funspec.PathAnyFunSpec with Matchers {

  println("Start of: ExampleSpec")
  describe("A ListBuffer") {

    println("Start of: A ListBuffer")
    val buf = ListBuffer.empty[Int]

    it("should be empty when created") {

      println("In test: should be empty when created; buf is: " + buf)
      buf should be ('empty)
    }

    describe("when 1 is appended") {

      println("Start of: when 1 is appended")
      buf += 1

      it("should contain 1") {

        println("In test: should contain 1; buf is: " + buf)
        buf.remove(0) should equal (1)
        buf should be ('empty)
      }

      describe("when 2 is appended") {

        println("Start of: when 2 is appended")
        buf += 2

        it("should contain 1 and 2") {

          println("In test: should contain 1 and 2; buf is: " + buf)
          buf.remove(0) should equal (1)
          buf.remove(0) should equal (2)
          buf should be ('empty)
        }

        describe("when 2 is removed") {

          println("Start of: when 2 is removed")
          buf -= 2

          it("should contain only 1 again") {

            println("In test: should contain only 1 again; buf is: " + buf)
            buf.remove(0) should equal (1)
            buf should be ('empty)
          }

          println("End of: when 2 is removed")
        }

        describe("when 3 is appended") {

          println("Start of: when 3 is appended")
          buf += 3

          it("should contain 1, 2, and 3") {

            println("In test: should contain 1, 2, and 3; buf is: " + buf)
            buf.remove(0) should equal (1)
            buf.remove(0) should equal (2)
            buf.remove(0) should equal (3)
            buf should be ('empty)
          }
          println("End of: when 3 is appended")
        }

        println("End of: when 2 is appended")
      }

      describe("when 88 is appended") {

        println("Start of: when 88 is appended")
        buf += 88

        it("should contain 1 and 88") {

          println("In test: should contain 1 and 88; buf is: " + buf)
          buf.remove(0) should equal (1)
          buf.remove(0) should equal (88)
          buf should be ('empty)
        }

        println("End of: when 88 is appended")
      }

      println("End of: when 1 is appended")
    }

    it("should have size 0 when created") {

      println("In test: should have size 0 when created; buf is: " + buf)
      buf should have size 0
    }

    println("End of: A ListBuffer")
  }
  println("End of: ExampleSpec")
  println()
}

Running the above version of ExampleSpec in the Scala interpreter will give you output similar to:

scala> import org.scalatest._
import org.scalatest._

scala> run(new ExampleSpec)
ExampleSpec:
Start of: ExampleSpec
Start of: A ListBuffer
In test: should be empty when created; buf is: ListBuffer()
End of: A ListBuffer
End of: ExampleSpec

Start of: ExampleSpec
Start of: A ListBuffer
Start of: when 1 is appended
In test: should contain 1; buf is: ListBuffer(1)
ExampleSpec:
End of: when 1 is appended
End of: A ListBuffer
End of: ExampleSpec

Start of: ExampleSpec
Start of: A ListBuffer
Start of: when 1 is appended
Start of: when 2 is appended
In test: should contain 1 and 2; buf is: ListBuffer(1, 2)
End of: when 2 is appended
End of: when 1 is appended
End of: A ListBuffer
End of: ExampleSpec

Start of: ExampleSpec
Start of: A ListBuffer
Start of: when 1 is appended
Start of: when 2 is appended
Start of: when 2 is removed
In test: should contain only 1 again; buf is: ListBuffer(1)
End of: when 2 is removed
End of: when 2 is appended
End of: when 1 is appended
End of: A ListBuffer
End of: ExampleSpec

Start of: ExampleSpec
Start of: A ListBuffer
Start of: when 1 is appended
Start of: when 2 is appended
Start of: when 3 is appended
In test: should contain 1, 2, and 3; buf is: ListBuffer(1, 2, 3)
End of: when 3 is appended
End of: when 2 is appended
End of: when 1 is appended
End of: A ListBuffer
End of: ExampleSpec

Start of: ExampleSpec
Start of: A ListBuffer
Start of: when 1 is appended
Start of: when 88 is appended
In test: should contain 1 and 88; buf is: ListBuffer(1, 88)
End of: when 88 is appended
End of: when 1 is appended
End of: A ListBuffer
End of: ExampleSpec

Start of: ExampleSpec
Start of: A ListBuffer
In test: should have size 0 when created; buf is: ListBuffer()
End of: A ListBuffer
End of: ExampleSpec

A ListBuffer
- should be empty when created
  when 1 is appended
  - should contain 1
    when 2 is appended
    - should contain 1 and 2
      when 2 is removed
      - should contain only 1 again
      when 3 is appended
      - should contain 1, 2, and 3
    when 88 is appended
    - should contain 1 and 88
- should have size 0 when created

Note that each test is executed in order of appearance in the PathAnyFunSpec, and that only those println statements residing in blocks that enclose the test being run are executed. Any println statements in blocks that do not form the "path" to a test are not executed in the instance of the class that executes that test.

How it executes

To provide its special brand of test isolation, PathAnyFunSpec executes quite differently from its sister class in org.scalatest.funspec. An org.scalatest.funspec.AnyFunSpec registers tests during construction and executes them when run is invoked. An org.scalatest.funspec.PathAnyFunSpec, by contrast, runs each test in its own instance while that instance is being constructed. During construction, it registers not the tests to run, but the results of running those tests. When run is invoked on a PathAnyFunSpec, it reports the registered results and does not run the tests again. If run is invoked a second or third time, in fact, a PathAnyFunSpec will each time report the same results registered during construction. If you want to run the tests of a PathAnyFunSpec anew, you'll need to create a new instance and invoke run on that.

A PathAnyFunSpec will create one instance for each "leaf" node it contains. The main kind of leaf node is a test, such as:

// One instance will be created for each test
it("should be empty when created") {
  buf should be ('empty)
}

However, an empty scope (a scope that contains no tests or nested scopes) is also a leaf node:

 // One instance will be created for each empty scope
describe("when 99 is added") {
  // A scope is "empty" and therefore a leaf node if it has no
  // tests or nested scopes, though it may have other code (which
  // will be executed in the instance created for that leaf node)
  buf += 99
}

The tests will be executed sequentially, in the order of appearance. The first test (or empty scope, if that is first) will be executed when a class that mixes in PathAnyFunSpec is instantiated. Only the first test will be executed during this initial instance, and of course, only the path to that test. Then, the first time the client uses the initial instance (by invoking one of run, expectedTestsCount, tags, or testNames on the instance), the initial instance will, before doing anything else, ensure that any remaining tests are executed, each in its own instance.

To ensure that the correct path is taken in each instance, and to register its test results, the initial PathAnyFunSpec instance must communicate with the other instances it creates for running any subsequent leaf nodes. It does so by setting a thread-local variable prior to creating each instance (a technique suggested by Esko Luontola). Each instance of PathAnyFunSpec checks the thread-local variable. If the thread-local is not set, it knows it is an initial instance and therefore executes every block it encounters until it discovers, and executes the first test (or empty scope, if that's the first leaf node). It then discovers, but does not execute the next leaf node, or discovers there are no other leaf nodes remaining to execute. It communicates the path to the next leaf node, if any, and the result of running the test it did execute, if any, back to the initial instance. The initial instance repeats this process until all leaf nodes have been executed and all test results registered.

Ignored tests

You mark a test as ignored in an org.scalatest.funspec.PathAnyFunSpec in the same manner as in an org.scalatest.funspec.AnyFunSpec. Please see the Ignored tests section in its documentation for more information.

Note that a separate instance will be created for an ignored test, and the path to the ignored test will be executed in that instance, but the test function itself will not be executed. Instead, a TestIgnored event will be fired.

Informers

You output information using Informers in an org.scalatest.funspec.PathAnyFunSpec in the same manner as in an org.scalatest.funspec.AnyFunSpec. Please see the Informers section in its documentation for more information.

Pending tests

You mark a test as pending in an org.scalatest.funspec.PathAnyFunSpec in the same manner as in an org.scalatest.funspec.AnyFunSpec. Please see the Pending tests section in its documentation for more information.

Note that a separate instance will be created for a pending test, and the path to the ignored test will be executed in that instance, as well as the test function (up until it completes abruptly with a TestPendingException).

Tagging tests

You can place tests into groups by tagging them in an org.scalatest.funspec.PathAnyFunSpec in the same manner as in an org.scalatest.funspec.AnyFunSpec. Please see the Tagging tests section in its documentation for more information.

Note that one difference between this class and its sister class org.scalatest.funspec.AnyFunSpec is that because tests are executed at construction time, rather than each time run is invoked, an org.scalatest.funspec.PathAnyFunSpec will always execute all non-ignored tests. When run is invoked on a PathAnyFunSpec, if some tests are excluded based on tags, the registered results of running those tests will not be reported. (But those tests will have already run and the results registered.) By contrast, because an org.scalatest.funspec.AnyFunSpec only executes tests after run has been called, and at that time the tags to include and exclude are known, only tests selected by the tags will be executed.

In short, in an org.scalatest.funspec.AnyFunSpec, tests not selected by the tags to include and exclude specified for the run (via the Filter passed to run) will not be executed. In an org.scalatest.funspec.PathAnyFunSpec, by contrast, all non-ignored tests will be executed, each during the construction of its own instance, and tests not selected by the tags to include and exclude specified for a run will not be reported. (One upshot of this is that if you have tests that you want to tag as being slow so you can sometimes exclude them during a run, you probably don't want to put them in a PathAnyFunSpec. Because in a path.Freespec the slow tests will be run regardless, with only their registered results not being reported if you exclude slow tests during a run.)

Shared tests

You can factor out shared tests in an org.scalatest.funspec.PathAnyFunSpec in the same manner as in an org.scalatest.funspec.AnyFunSpec. Please see the Shared tests section in its documentation for more information.

Nested suites

Nested suites are not allowed in a PathAnyFunSpec. Because a PathAnyFunSpec executes tests eagerly at construction time, registering the results of those test runs and reporting them later when run is invoked, the order of nested suites versus test runs would be different in a org.scalatest.funspec.PathAnyFunSpec than in an org.scalatest.funspec.AnyFunSpec. In org.scalatest.funspec.AnyFunSpec's implementation of run, nested suites are executed then tests are executed. A org.scalatest.funspec.PathAnyFunSpec with nested suites would execute these in the opposite order: first tests then nested suites. To help make PathAnyFunSpec code easier to reason about by giving readers of one less difference to think about, nested suites are not allowed. If you want to add nested suites to a PathAnyFunSpec, you can instead wrap them all in a Suites object. They will be executed in the order of appearance (unless a Distributor is passed, in which case they will execute in parallel).

Durations

Many ScalaTest events include a duration that indicates how long the event being reported took to execute. For example, a TestSucceeded event provides a duration indicating how long it took for that test to execute. A SuiteCompleted event provides a duration indicating how long it took for that entire suite of tests to execute.

In the test completion events fired by a PathAnyFunSpec (TestSucceeded, TestFailed, or TestPending), the durations reported refer to the time it took for the tests to run. This time is registered with the test results and reported along with the test results each time run is invoked. By contrast, the suite completion events fired for a PathAnyFunSpec represent the amount of time it took to report the registered results. (These events are not fired by PathAnyFunSpec, but instead by the entity that invokes run on the PathAnyFunSpec.) As a result, the total time for running the tests of a PathAnyFunSpec, calculated by summing the durations of all the individual test completion events, may be greater than the duration reported for executing the entire suite.