Hi say I have some code like:
public class Class1
{
public int MyMethod()
{
return MyOtherMethod();
}
public virtual int MyOtherMethod()
{
return 1;
}
}
Ok, this doesn't do much of relevance but this is just for a simple example.
I then create a new test:
[TestMethod]
public void TestMethod1()
{
var t = new Mock<Class1>();
var w = t.Object.MyMethod();
}
Could someone please tell me why the code runs through the called method MyOtherMethod when it's not designated as virtual but when you make it virtual the test code refuses to go through that method?
You should setup the virtual method before you calling MyMethod:
t.Setup(c => c.MyOtherMethod()).Return(1);
Related
I'm still pretty new to Google Mock so learning as I go. I've just been adding some unit tests and I've run into an issue where I can't get ON_CALL() to correctly stub a method called from within a method.
The following code outlines what I have;
class simpleClass
{
public:
bool simpleFn1() { return simpleFn2(); }
virtual bool simpleFn2() { return FALSE; }
}
In my unit test I have:
class simpleClassMocked : public simpleClass
{
private:
MOCK_METHOD0(simpleFn2, bool());
}
class simpleClassTests : public ::testing::Test
{
}
TEST_F(simpleClassTests, testSimpleFn2)
{
shared_ptr<simpleClassMocked> pSimpleClass = shared_ptr< simpleClassMocked >(new simpleClassMocked());
ON_CALL(*pSimpleClass, simpleF2()).WillByDefault(Return(TRUE));
// This works as expected - simpleFn2() gets stubbed
pSimpleClass->simpleFn2();
// This doesn't work as expected - when simpleFn1 calls simpleFn2 it's not the stubbed expectation???
pSimpleClass->simpleFn1();
}
I figure I must be missing something obvious here, can anyone help? Thanks!
you'll have to Mark the method as virtual and add a corresponding MOCK function in the simpleClassMocked class
class simpleClass
{
public:
virtual bool simpleFn1() { return simpleFn2(); }
virtual bool simpleFn2() { return FALSE; }
}
Also, you need to put the Mock methods in the public area
class simpleClassMocked : public simpleClass
{
public:
MOCK_METHOD0(simpleFn2, bool());
MOCK_METHOD0(simpleFn1, bool());
}
It will work now
Let's say I have a dagger 2 module as follows,
#Module
interface Creator {
MyClass create();
}
and I am using it to create an instance of MyClass
class Instantiator {
void doSomething(){
MyClass clazz = DaggerCreator.create().create();
// do things with clazz
}
}
It seems to me that I cannot effectively test the doSomething method in Instantiator because I cannot provide a mock for MyClass.
Am I wrong? If not are we supposed to use Dagger instantiation sparingly?
You are correct in saying that it is hard to test use of a Component injector, since this is a static method. But harder than what? Here is the same method using instantiation:
class Instantiator {
void doSomething(){
MyClass clazz = new MyClass();
// do things with clazz
}
}
still hard to test, right?
The point is to use as few Component (injectors) as possible and to pass in dependencies in the constructor for your objects. Dagger 2 makes resolving the dependencies in the constructor easy. This thereby makes testing easy since you can pass in mock object in a constructor.
Let's refactor the code you wrote to be testable. Assume that MyClass contains a single method, fireLazers() that you want to test is being invoked inside Instantiator's doSomething() method:
public class DoerOfSomething {
private final MyClass myClass;
#Inject
public DoerOfSomething(MyClass myClazz) {
this.myClass = myClazz;
}
public void doSomething() {
myClass.fireLazers();
}
}
Now you can write a test like this using a mock object:
public void DoerOfSomethingTest {
//mocks
MyClass mockMyClass;
//system under test
DoerOfSomething doerOfSomething;
#Before
public void setUp() {
mockMyClass = Mockito.mock(MyClass.class);
}
#Test
public void whenDoSomething_thenInteractsWithMyClass() {
//arrange
doerOfSomething = new DoerOfSomething(mockMyClass);
//act
doerOfSomething.doSomething();
//assert
verify(mockMyClass).fireLazers();
}
}
Of course, you will now need to inject DoerOfSomething into the top level class where you are injecting, but now you can be certain that the object you are injecting is functioning as expected because it is testable. Your code for using Dagger looks a bit unusual but I'll use your idioms for the sake of parity between the question and the answer.
class Instantiator {
private final DoerOfSomething doerOfSomething;
Instantiator() {
doerOfSomething = DaggerCreator.create().create();
}
void doSomething() {
doerOfSomething.doSomething();
}
}
I am trying to find out whether and how it is possible to nest JUnit test runners, e.g. combine a GuiceJUnitRunner, a Parameterized and a HierarchicalcontextRunner.
To me, it seems that JUnit was not designed to achieve this easily, otherwise BlockJUnit4ClassRunner should have a method which passes the next Runner as an argument.
Someone also implemented a ParallelParameterized runner, which looks to me like combining Parallel and Parameterized was not easily possible.
When googling for "nested" and "JUnit", it comes up with lots of information for nested classes, but I'm looking for nesting Runners, not classes.
There is something called NestedRunner for running plain old Java classes in nested configuration.
your test starts with #RunWith(NestedRunner.class) and here is the example that I found:
#RunWith(NestedRunner.class)
public class ListTest {
// inner class for sharing common context
public class WithArrayList {
// some context for these tests
ArrayList<String> list = new ArrayList<String>();
public class WhenEmpty {
#Test
public void itIsEmpty() {
assertTrue(list.isEmpty());
}
public class AfterAddingAnElement {
// some more context for these tests
String element = "Element";
// you can use instance initializer to initialize your context
// it will be run once per test
{
// the list is still empty in here
assertTrue(list.isEmpty());
list.add(element);
}
#Test
public void itIsNotEmpty() {
assertFalse(list.isEmpty());
}
#Test
public void itContainsTheElement() {
assertTrue(list.contains(element));
}
#Test
public void addingAnotherElementIncreasesSize() {
int sizeBeforeAdding = list.size();
list.add("AnotherElement");
assertThat(list.size(), is(greaterThan(sizeBeforeAdding)));
}
#Test
public void listSizeIsStillOne() {
assertThat(list.size(), is(equalTo(1)));
}
}
#Test
public void isStillEmpty() {
assertTrue(list.isEmpty());
}
}
public class WithTwoElements {
#Before
public void init() {
list.add("Element1");
list.add("Element2");
}
#Test
public void hasSizeOfTwo() {
assertThat(list.size(), is(equalTo(2)));
}
}
}
}
and here is the source for further info
I am unit testing legacy code and I am dealing with a class that instantiates another class. I believe this is testable using Microsoft Fakes, but am wondering if NSubstitute has the capability. I believe the answer is no, but need to be sure.
public class ClassA
{
public int MethodA()
{
int reportId = this.MethodB();
return reportId;
}
public virtual int MethodB()
{
ClassC c = new ClassC();
return c.MethodA();
}
}
public class ClassC
{
public virtual int MethodA()
{
return 2;
}
}
[Test]
public void Test_ClassA()
{
ClassA subclassA = new ClassA();
var subclassC = Substitute.For<ClassC>(); //this is pointless the way I have it here
subclassC.MethodA().Returns(1); //this is pointless the way I have it here
int outValue = subclassA.MethodA();
Assert.AreEqual(outValue, 1); //outvalue is 2 but I would like it to be 1 if possible using Nsubstitute
}
It is possible to override virtual methods in classes by using partial substitution: just make sure that the base code cannot be called by specifying that the base class must not be called:
var A = Substitute.ForPartsOf<ClassA>();
var C = Substitute.ForPartsOf<ClassC>();
C.When(c => c.MethodA()).DoNotCallBase();
C.MethodA().Returns(10);
A.When(a => a.MethodB()).DoNotCallBase();
var cResult = C.MethodA();
A.MethodB().Returns(cResult);
Console.WriteLine(A.MethodB());
Console.WriteLine(C.MethodA());
I am setting up some MSTest based unit tests. To make my life easier I want to use a base class that handles the generic setup and taredown all of my tests require. My base class looks like this:
[TestClass]
public class DBTestBase {
public TestContext TestContext { get; set; }
[ClassInitialize()]
public static void MyClassInitialize(TestContext testContext) {
var config = new XmlConfigurationSource("ARconfig_test.xml");
ActiveRecordStarter.Initialize(Assembly.Load("LocalModels"), config);
}
[TestInitialize()]
public void MyTestInitialize() {
ActiveRecordStarter.CreateSchema();
Before_each_test();
}
protected virtual void Before_each_test() { }
[TestCleanup()]
public void MyTestCleanup() {
After_each_test();
}
protected virtual void After_each_test() { }
}
My actual test class looks like this:
[TestClass]
public class question_tests : DBTestBase {
private void CreateInitialData() {
var question = new Question()
{
Name = "Test Question",
Description = "This is a simple test question"
};
question.Create();
}
protected override void Before_each_test() {
base.Before_each_test();
CreateInitialData();
}
[TestMethod]
public void test_fetching() {
var q = Question.FindAll();
Assert.AreEqual("Test Question", q[0].Name, "Incorrect name.");
}
}
The TestInitialize function works as expected. But the ClassInitialize function never runs. It does run if I add the following to my child class:
[ClassInitialize()]
public static void t(TestContext testContext) {
MyClassInitialize(testContext);
}
Is it possible to get my base class initialize function to run without referencing it in my child class?
ClassInitialize method is executed if and only if the concerned "class" contains at least one TestMethod, and at least one TestMethod from the class is selected for execution.
Confirm this was a problem for me too. I used a constructor on the base and a destructor for the cleanup
[TestClass]
public class question_tests : DBTestBase {
...
[TestCleanup()]
public void TestCleanup()
{
base.MyTestCleanup();
}