How to do setup/teardown in specs2 when using "in new WithApplication" - scala

I am using Specs2 with play 2.2.1 built with Scala 2.10.2 (running Java 1.7.0_51). I have been reading about how to do setup/teardown with Specs2. I have seen examples using the "After" trait as follows:
class Specs2Play extends org.specs2.mutable.Specification {
"this is the first example" in new SetupAndTeardownPasswordAccount {
println("testing")
}
}
trait SetupAndTeardownPasswordAccount extends org.specs2.mutable.After {
println("setup")
def after = println("teardown ")
}
This works fine, except that all of my tests are using "in new WithApplication". It seems what I need is to have an object which is both a "WithApplication" and an "After". Below does not compile, but is essentially what I want:
trait SetupAndTeardownPasswordAccount extends org.specs2.mutable.After with WithApplication
So, my question is, how do I add setup/teardown to my tests which are already using "in WithApplication"? My primary concern is that all of our tests make use of fake routing like this (so they need the With Application).
val aFakeRequest = FakeRequest(method, url).withHeaders(headers).withBody(jsonBody)
val Some(result) = play.api.test.Helpers.route(aFakeRequest)
result

This is the code for WithApplication:
abstract class WithApplication(val app: FakeApplication = FakeApplication()) extends Around with Scope {
implicit def implicitApp = app
override def around[T: AsResult](t: => T): Result = {
Helpers.running(app)(AsResult.effectively(t))
}
}
It's actually quite easy to modify this to suit your needs without creating a bunch of other traits. The missing piece here is the anonymous function t, which you provide the implementation for in your tests (using WithApplication). It would be nice to make WithApplication a little more robust to be able to execute arbitrary blocks of code before and after the tests, if necessary.
One approach could be to create a similar class to WithApplication that accepts two anonymous functions setup and teardown that both return Unit. All I really need to do is modify what's happening inside AsResult.effectively(t). To keep this simple, I'm going to remove the app parameter from the parameter list, and use FakeApplication always. You don't seem to be providing a different configuration, and it can always be added back.
abstract class WithEnv(setup: => Unit, teardown: => Unit) extends Around with Scope {
implicit def implicitApp = app
override def around[T: AsResult](t: => T): Result = {
Helpers.running(app)(AsResult.effectively{
setup
try {
t
} finally {
teardown
}
})
}
}
Instead of simply calling the anonymous function t, I first call setup, then t, then teardown. The try/finally block is important because failed tests in specs2 throw exceptions, and we want to be sure that teardown will be executed no matter what the outcome.
Now you can easily setup test environments using functions.
import java.nio.files.{Files, Paths}
def createFolder: Unit = Files.createDirectories(Paths.get("temp/test"))
def deleteFolder: Unit = Files.delete("temp/test")
"check if a file exists" in new WithEnv(createFolder, deleteFolder) {
Files.exists(Paths.get("temp/test")) must beTrue
}
(This might not compile, but you get the idea.)

If your after method doesn't need anything from the WithApplication trait you can mix in your specification the AfterExample trait and define the after behaviour for the whole spec:
import org.specs2.specification._
class Specs2Play extends org.specs2.mutable.Specification with AfterExample {
"this is the first example" in new SetupAndTeardownPasswordAccount {
pending("testing")
}
trait SetupAndTeardownPasswordAccount extends WithApplication
def after = println("cleanup")
}

Related

How to run ConfigMapWrapperSuite?

I need to write an integration test and it requires starting a server executable. I want to make location of the server configurable, so that I could set it on my box and on integration server.
ConfigMapWrapperSuite seems to be doing exactly what I want:
#WrapWith(classOf[ConfigMapWrapperSuite])
class ConsulTest(configMap: ConfigMap) extends FlatSpec with ShouldMatchers {
val consulPath = configMap("consul.path")
"Consul" should "list keys under root" in {
...
}
But when I set my IDE (IntelliJ) to execute all tests in the project, I get an exception saying that constructor with Map parameter not found. Looking into source code of scalatest revealed:
final class ConfigMapWrapperSuite(clazz: Class[_ <: Suite]) extends Suite {
private lazy val wrappedSuite = {
val constructor = clazz.getConstructor(classOf[Map[_, _]])
constructor.newInstance(Map.empty)
}
So in contrary to what documentation says, suite must have constructor with a Map and not ConfigMap.
Ok, I changed constructor to take a Map[String,String] but now I get NoSuchElementException at val consulPath = configMap("consul.path"). Lookung up the stack down to ConfigMapWrapperSuite and I see that constructor.newInstance(Map.empty) WTF? So wrapped suite class is instantiated with empty map, and than another time, during the suite run with actual map of parameters? How do I suppose to get parameters if I'm given an empty map?
I looked up scalatest's unit tests. They are so rudimentary that actually retrieving a value from configMap is not performed.
I do not want to use ConfigMapFixture because it will make me initializing every single test with the same code.
So, how do I not only pass but also get global setting in test suite?
Scalatest version: 3.0.0-M15
Ok, answering my own question. ConfigMapWrapperSuite seems to be not used too much and essentially is broken.
Instead I've used BeforeAndAfterAllConfigMap as in here:
class ConsulTest extends FlatSpec with ShouldMatchers with OneInstancePerTest with BeforeAndAfterAllConfigMap {
var consulProcess: Process = null
override def beforeAll(conf: ConfigMap): Unit = {
consulProcess = Seq("bin/"+exe, "agent", "-advertise", "127.0.0.1", "-config-file", "bin/config.json").run()
}
override def afterAll(conf: ConfigMap): Unit = {
consulProcess.destroy()
}
You need override instance
#WrapWith(classOf[ConfigMapWrapperSuite])
class AcceptanceTest(configMap: Map[String, Any])
extends FlatSpec
with OneInstancePerTest {
override def newInstance = new AcceptanceTest(configMap)
}

Testing object which calls another object in Scala using Specs2

I'm working with a project which already has some legacy code written in Scala. I was given a task to write some unit tests for one of its classes when I discovered it's not so easy. Here's the problem I've encountered:
We have an object, say, Worker and another object to access the database, say, DatabaseService which also extends other class (I don't think it matters, but still). Worker, in its turn, is called by higher classes and objects.
So, right now we have something like this:
object Worker {
def performComplexAlgorithm(id: String) = {
val entity = DatabaseService.getById(id)
//Rest of the algorithm
}
}
My first though was 'Well, I can probably make a trait for DatabaseService with the getById method'. I don't really like the idea to create an interface/trait/whatever just for the sake of testing because I believe it doesn't necessarily lead to a nice design, but let's forget about it for now.
Now, if Worker was a class, I could easily use DI. Say, via constructor like this:
trait DatabaseAbstractService {
def getById(id: String): SomeEntity
}
object DatabaseService extends SomeOtherClass with DatabaseAbstractService {
override def getById(id: String): SomeEntity = {/*complex db query*/}
}
//Probably just create the fake using the mock framework right in unit test
object FakeDbService extends DatabaseAbstractService {
override def getById(id: String): SomeEntity = {/*just return something*/}
}
class Worker(val service: DatabaseService) {
def performComplexAlgorithm(id: String) = {
val entity = service.getById(id)
//Rest of the algorithm
}
}
The problem is, Worker is not a class so I can't make an instance of it with another service. I could do something like
object Worker {
var service: DatabaseAbstractService = /*default*/
def setService(s: DatabaseAbstractService) = service = s
}
However, it scarcely makes any sense to me since it looks awful and leads to an object with mutable state which doesn't seem very nice.
The question is, how can I make the existing code easily testable without breaking anything and without making any terrible workarounds? Is it possible or should I change the existing code instead so that I could test it easier?
I was thinking about using extending like this:
class AbstractWorker(val service: DatabaseAbstractService)
object Worker extends AbstractWorker(DatabaseService)
and then I somehow could create a mock of Worker but with different service. However, I didn't figure out how to do it.
I'd appreciate any advice as to how either change the current code to make it more testable or test the existing.
If you can alter the code for Worker, you can change it to still allow it to be an object and also allow for swapping of the db service via an implicit with a default definition. This is one solution and I don't even know if this is possible for you, but here it is:
case class MyObj(id:Long)
trait DatabaseService{
def getById(id:Long):Option[MyObj] = {
//some impl here...
}
}
object DatabaseService extends DatabaseService
object Worker{
def doSomething(id:Long)(implicit dbService:DatabaseService = DatabaseService):Option[MyObj] = {
dbService.getById(id)
}
}
So we set up a trait with concrete impl of the getById method. Then we add an object impl of that trait as a singleton instance to use in the code. This is a good pattern to allow for mocking of what was previously only defined as an object. Then, we make Worker accept an implicit DatabaseService (the trait) on it's method and give it a default value of the object DatabaseService so that regular use does not have to worry about satisfying that requirement. Then we can test it like so:
class WorkerUnitSpec extends Specification with Mockito{
trait scoping extends Scope{
implicit val mockDb = mock[DatabaseService]
}
"Calling doSomething on Worker" should{
"pass the call along to the implicit dbService and return rhe result" in new scoping{
mockDb.getById(123L) returns Some(MyObj(123))
Worker.doSomething(123) must beSome(MyObj(123))
}
}
Here, in my scope, I make an implicit mocked DatabaseService available that will supplant the default DatabaseService on the doSomething method for my testing purposes. Once you do that, you can start mocking out and testing.
Update
If you don't want to take the implicit approach, you could redefine Worker like so:
abstract class Worker(dbService:DatabaseService){
def doSomething(id:Long):Option[MyObj] = {
dbService.getById(id)
}
}
object Worker extends Worker(DatabaseService)
And then test it like so:
class WorkerUnitSpec extends Specification with Mockito{
trait scoping extends Scope{
val mockDb = mock[DatabaseService]
val testWorker = new Worker(mockDb){}
}
"Calling doSomething on Worker" should{
"pass the call along to the implicit dbService and return rhe result" in new scoping{
mockDb.getById(123L) returns Some(MyObj(123))
testWorker.doSomething(123) must beSome(MyObj(123))
}
}
}
In this way, you define all the logic of importance in the abstract Worker class and that's what you till focus your testing on. You provide a singleton Worker via an object that is used in the code for convenience. Having an abstract class let's you use a constructor param to specify the database service impl to use. This is semantically the same as the previous solution but it's cleaner in that you don't need the implicit on every method.

Wrapping a test in a transaction

Assuming I have several tests that do something like this
"should do something with the database" in new WithApplication {
DB.withTransaction { implicit con =>
// Query the database.
// Perform my tests.
// Rollback the transaction to return database to initial state
con.rollback()
}
}
I don't want to remember to wrap all my database tests in a transaction and call con.rollback() at the end of each test manually.
Instead I tried writing a trait that wraps the entire test in a transaction so I could just write my tests like this
"do something to the database" in new WithApplication with DatabaseTest {
SQL("...").execute()
}
This is what I have so far
trait DatabaseTest extends Around with Scope {
abstract override def around[T: AsResult](test: => T): Result = {
super.around = {
import play.api.Play.current
DB.withTransaction { implicit con =>
val result = test
con.rollback()
result
}
}
}
However when I try running the test above I get this error
could not find implicit value for parameter connection: java.sql.Connection
So my question is whether what I'm trying to do is possible, and if so, what's wrong with my code and what would I have to change to get it to work.
Implicits in Scala is not quite as magic as transactions added to some kind of context in Java application servers for example, instead they work by telling the compiler "if this parameter of type T is missing, look in the scope around where I'm used for a value of type T marked as implicit".
This is why your usual code that looks like this works:
def doStuff()(implicit aConnection: Connection) = ???
DB.withTransaction { implicit theConnection =>
doStuff() // secretly the compiler adds theConnection here
}
But this also means that you cannot get it for free because some class you inherited have an implicit connection, it needs to be in scope to be available, and Around does not let you change the signature of the test, so you cant really get it into scope using that.
One way to do it might be to avoid withTransaction and instead use DB.getConnection in a scope, something like this (from the top of my head, so might not be exact/compile):
trait Tx extends Scope with After {
implicit lazy val conn = DB.getConnection(autocommit = false)
override def after() = conn.rollback()
}
Then use it in your tests, which will in the implicit connection into scope:
"something something" in new Tx {
SQL("...").execute()
}
Based on johanandren's answer I've come up with the following solution
trait DatabaseIsolation extends Scope with After {
implicit val app: FakeApplication = FakeApplication()
implicit lazy val con = DB.getConnection(autocommit = false)
override def after = con.rollback()
}
And then use it in the test like this
"do something" in new DatabaseIsolation {
SQL("...").execute()
}

Overriding a trait and selftype

I want to override the ScalaTest trait BeforeAndAfterEach to have that stuff implemented once for all my tests. Finally I got it to compile, but I don't understand why.
trait MySetup extends BeforeAndAfterEach {
this : org.scalatest.BeforeAndAfterEach with org.scalatest.Suite =>
var service: String = _
abstract override def beforeEach(): Unit = {
service = "apa"
super.beforeEach()
}
abstract override def afterEach(): Unit = {
service = ""
}
}
The thing that got it working was the line:
this : org.scalatest.BeforeAndAfterEach with org.scalatest.Suite =>
I found it in the beginning of the BeforeAndAfterEach implementation and copied it.
What does it do, and why do I need it?
Update:
This is a simpler version.
trait MySetup extends FlatSpec with BeforeAndAfterEach {
var service: String = _
override def beforeEach {
service = "apa"
super.beforeEach
}
override def afterEach {
service = ""
super.afterEach
}
}
BeforeAndAfterEach has a self-type of Suite, meaning that BeforeAndAfterEach can only be mixed in to a type that extends Suite. ScalaTest wants to you pick a primary suite type first and then mix-in behavior afterwards.
The self-type declaration is not inherited in sub-traits so you have to redeclare the self-type.
The following question has some tradeoffs between self-types and sub-traits: What is the difference between self-types and trait subclasses?
For some background on ScalaTest design, see: http://www.artima.com/scalazine/articles/selfless_trait_pattern.html
The way I'd probably write MySetup is this:
import org.scalatest.Suite
import org.scalatest.BeforeAndAfterEach
trait MySetup extends BeforeAndAfterEach { this: Suite =>
var service: String = _
abstract override def beforeEach(): Unit = {
service = "apa"
super.beforeEach()
}
abstract override def afterEach(): Unit = {
try {
super.afterEach() // To be stackable, must call super.afterEach
}
finally {
service = ""
}
}
}
That way the self type is less intrusive. Mike's answer is correct. The idea is to allow traits to stack, so you can mix in multiple traits like this, in different orders if you want. Another relevant article is "The Stackable Trait Pattern":
http://www.artima.com/scalazine/articles/stackable_trait_pattern.html
You may also find the examples in the relevant Scaladoc section helpful:
http://www.scalatest.org/scaladoc-1.6.1/org/scalatest/FlatSpec.html#composingFixtures
Note I call super.afterEach too. To be stackable, you need to call super on both beforeEach and afterEach. I do it in a try so that if the super.afterEach blows up with an exception, you still get the after behavior of this trait. (Though likely your Suite will abort anyway at that point, so in this case it probably doesn't matter. But in general it's a good idea.)
This is Scalas syntax for dependency injection.
this: <dependency> =>
It means literally this trait depends upon <dependency> trait. Read this article for more.

How to mock a method with functional arguments in Scala?

I'm trying to mock a method call that takes a call-by-name argument:
import org.scalatest.WordSpec
import org.scalatest.mock.MockitoSugar
import org.mockito.Mockito._
import org.junit.runner.RunWith
import org.scalatest.junit.JUnitRunner
trait Collaborator {
def doSomething(t: => Thing)
}
trait Thing
#RunWith(classOf[JUnitRunner])
class Test extends WordSpec with MockitoSugar {
"The subject under test" should {
"call the collaborator" in {
// setup
val m = mock[Collaborator]
val t = mock[Thing]
// test code: this would actually be invoked by the SUT
m.doSomething(t)
// verify the call
verify(m).doSomething(t)
}
}
}
I'm primarily interested in Mockito since that's what I'm using, but I'd be interested to see whether any of the major mock frameworks is capable of this kind of testing. The Test fails at runtime on the verify line, with an error like
Argument(s) are different! Wanted:
collaborator.doSomething(
($anonfun$apply$3) <function>
);
-> at Test$$anonfun$1$$anonfun$apply$1.apply(Test.scala:27)
Actual invocation has different arguments:
collaborator.doSomething(
($anonfun$apply$2) <function>
);
-> at Test$$anonfun$1$$anonfun$apply$1.apply(Test.scala:24)
If I'm understanding the situation correctly, the compiler is implicitly wrapping t in a nullary function that returns t. The mock framework is then comparing that function to the one produced in the test code, which is equivalent but not equals().
My case is a relatively simple version of the problem, but I think this would be an issue with any higher-order function.
This looks ugly, but hopefully it can help you to find good solution:
import org.scalatest.mock.MockitoSugar
import org.mockito.Mockito._
trait Collaborator {
def doSomething(t: => Thing)
}
trait Thing
new MockitoSugar {
// setup
val m = mock[Collaborator]
val t = mock[Thing]
m.doSomething(t)
classOf[Collaborator].getMethod("doSomething", classOf[Function0[_]]).invoke(
verify(m),
new Function0[Thing] {
def apply() = null
override def equals(o: Any): Boolean = t == o.asInstanceOf[Function0[Thing]].apply()
})
}
You can try specs2. In specs2, we "hijack" the Mockito Invocation class to account for byname parameters:
trait ByName { def call(i: =>Int) = i }
val byname = mock[ByName]
byname.call(10)
there was one(byname).call(10)
This problem seems to be specific to by-name invocations because in regular higher order functions you can match against the explicit FunctionX object:
verify(collaborator).somethingElse(any(Function2[String, Thing]))
in the by-name case the wrapping of the argument into a Function0 is done implicitly, and Alexey's answer shows how to invoke the mock with an explicit parameter.
You could write something akin to your own verify which would apply arguments captured by mockito.
Mockito internally records invocation and their arguments with e.g.:
http://code.google.com/p/mockito/source/browse/trunk/src/org/mockito/internal/matchers/CapturingMatcher.java