How to use verified method of org.mockito.internal.invocation.InvocationBuilder class

On my end if I build your project using the pom.xml you've showed us, with apache-poi declared after jfreechart, then as you've mentioned I get a corrupt JAR. Swapping the order of these two dependencies indeed gives me a correct JAR.

I've some previous experience with the maven-shade-plugin and when I used it I had trouble with the created JAR's META-INF directory, so I've checked that for defects.

I've tried to copy (using Total Commander) the META-INF directory somewhere on my local filesystem and what I've noticed is that there were errors while copying the files in META-INF/licences/. If I've tried to copy those somewhere individually and it worked but I couldn't copy the whole thing. My conclusion was that the JAR/ZIP archive is corrupt.

What I did was I entered that JAR in Total Commander (Ctrl+PgDown over the JAR file) and I renamed thirdpartylicenses.txt to thirdpartylicenses.txt.wtf. Doing this Total Commander offers to save this and it repackages the JAR (I've the Total7zip Total Commander packer plugin installed - if anyone tries this and it doesn't work try it with this installed).

After this. It works.

(I've also tried to repack the whole thing without renaming anything using Cygwin's unzip/zip commands, but that didn't work the new archive was still corrupt. Total Commander or the plugin I've mentioned does some magic.)

I guess the maven-shade-plugin merely creates a corrupt or invalid ZIP/JAR archive. I'm not sure why and maybe what I've described won't work for anyone else but I thought I'd mention it so maybe I could help.

---

I couldn't just leave this alone, so I've dug deeper and I think I've found the answer.

The bad JAR contains 65608 entries. The good JAR contains 65450 entries.

Guess what's the upper limit of the number of entries for a plain ZIP file? Yeah. The Wiki article speaks about a ZIP64 format which overcomes this limitation.

The good JAR has less entries because the actual dependencies change because of the position of the dependency declarations in your pom.xml. (As described by this answer.)

I've counted the entries like this.

Collections.list(new JarFile("...").entries()).size();

I was using Java 7 which seems to support the new ZIP64 format, so maybe if someone tries to count the entries in the bad JAR using Java 5 or 6 will receive an error (I'm not sure though).

I also tried to run the JAR exploded. I've unpacked the whole JAR into a directory and run the whole thing like this.

java -cp <dir/ com.mycompany.test.Start

Worked like a charm.

---

Bottom line. Don't overuse the maven-shade-plugin.

I have a project at work where I build my project like this.

• I copy my project's dependencies using the maven-dependency-plugin. Checkout the copy-dependencies goal. This copies your dependencies into target/dependency if I remember correctly.

• Using the maven-jar-plugin I add these dependencies to my final JAR's MANIFEST.MF as Class-Path entries using these options in the plugins configuration.

  <classpathPrefix>dependency/</classpathPrefix>
  <addClasspath>true</addClasspath>
  

  So I'll have Class-Path entries like dependency/<artifactId>-<version>.jar, etc.

• After this I'm using the maven-assembly-plugin to create a distribution ZIP which contains my final JAR and the whole dependency/ folder.
• When I deploy my application I can run it like java -jar final.jar.

Primarily I've opted to use this solution, because in my project I use some Bouncy Castle JARs which JARs have some extravagant this and that inside their META-INF directory. When I used the maven-shade-plugin to create my final runnable JAR all hell broke loose and I got nasty that method couldn't be found and this signature is not quite right errors.

You should be doing something like this also. This Maven shade business is too shady to be useful (pun intended).

---

Here's a blog post on the whole process that I've tried to describe just above (thanks to baba), maybe it'll help someone in the future.

Not possible through mockito. From their wiki

Why Mockito doesn't mock private methods?

Firstly, we are not dogmatic about mocking private methods. We just don't care about private methods because from the standpoint of testing private methods don't exist. Here are a couple of reasons Mockito doesn't mock private methods:

It requires hacking of classloaders that is never bullet proof and it changes the api (you must use custom test runner, annotate the class, etc.).

It is very easy to work around - just change the visibility of method from private to package-protected (or protected).

It requires me to spend time implementing & maintaining it. And it does not make sense given point #2 and a fact that it is already implemented in different tool (powermock).

Finally... Mocking private methods is a hint that there is something wrong with OO understanding. In OO you want objects (or roles) to collaborate, not methods. Forget about pascal & procedural code. Think in objects.

if you do a employee = new Employee(param1, param2); you may as well skip @InjectMocks.

It is supposed to do the following:

@InjectMocks
ClassUnderTest cut;

@Mock
Dependency1 dep1;
@Mock
Dependency2 dep2;

@Before
public void setup() {
  initMocks(this);
}

omitting @InjectMocks the same behaviour can be achieved with the following code:

ClassUnderTest cut;

@Mock
Dependency1 dep1;
@Mock
Dependency2 dep2;

@Before
public void setup() {
  initMocks(this);
  cut = new ClassUnderTest(dep1, dep2);
}

In your specific case, you should mock param1 and param2. Never call the constructor manually when using @InjectMocks.

You can use a spy on your static class and mock only specific methods:

@RunWith(PowerMockRunner.class)
@PrepareForTest(MyStaticTest.MyStaticClass.class)
public class MyStaticTest {

public static class MyStaticClass {
    public static String getA(String a) {
        return a;
    }
    public static String getB(String b) {
        return b;
    }
}

@Test
public void should_partial_mock_static_class() throws Exception {
    //given
    PowerMockito.spy(MyStaticClass.class);
    given(MyStaticClass.getB(Mockito.anyString())).willReturn("B");
    //then
    assertEquals("A", MyStaticClass.getA("A"));
    assertEquals("B", MyStaticClass.getA("B"));
    assertEquals("C", MyStaticClass.getA("C"));
    assertEquals("B", MyStaticClass.getB("A"));
    assertEquals("B", MyStaticClass.getB("B"));
    assertEquals("B", MyStaticClass.getB("C"));
}

}

You can use PowerMockito

See the example:

Second second = Mockito.mock(Second.class);
whenNew(Second.class).withNoArguments().thenReturn(second);

But re-factoring is better decision.

You need to define the behavior for getPath() for folder as it gets called internally in File class.

You can do it as:

File folder = Mockito.mock(File.class);
when(folder.getPath()).thenReturn("C:\temp\");
File file = new Agent().createNewFile(folder, "fileName");

It will work only till you don't really create a new file but only calling new File.

From the docs:

Sometimes we need to stub with different return value/exception for the same method call. Typical use case could be mocking iterators. Original version of Mockito did not have this feature to promote simple mocking. For example, instead of iterators one could use Iterable or simply collections. Those offer natural ways of stubbing (e.g. using real collections). In rare scenarios stubbing consecutive calls could be useful, though:

when(mock.someMethod("some arg"))
   .thenThrow(new RuntimeException())
  .thenReturn("foo");

//First call: throws runtime exception:
mock.someMethod("some arg");

//Second call: prints "foo"
System.out.println(mock.someMethod("some arg"));

So in your case, you'd want:

when(myMock.doTheCall())
   .thenReturn("You failed")
   .thenReturn("Success");

Mockito, till version 1.8.5, had a bug in the case of polymorphic dispatch. It was fixed and is available in the first release candidate of the version 1.9.0. See issue 200.

So how does it happen in your code base. Note you are mocking these two classes

nodeAutoIndexer = (AutoIndexer<Node>) mock(AutoIndexer.class);
relAutoIndexer = mock(RelationshipAutoIndexer.class);

AutoIndexer happen to be a generic parent interface, in this interface there is this method ReadableIndex<T> getAutoIndex(). RelationshipAutoIndexer is a subtype of the AutoInexer where the generic part is fixed to Relationship, and override the getAutoIndex() method to return the covariant type ReadableRelationshipIndex.

See AutoIndexer and RelationshipIndexer.

Well, in your calling code you have these lines:

AutoIndexer<Node> nodeAutoIndexer = this.graphDb.index().getNodeAutoIndexer();
AutoIndexer<Relationship> relAutoIndexer = this.graphDb.index().getRelationshipAutoIndexer();
this.nodeIndex = nodeAutoIndexer.getAutoIndex();
this.relIndex = relAutoIndexer.getAutoIndex();

Both nodeAutoIndex in your production code and the mock nodeAutoIndexer in your test code have a reference of type AutoIndexer<Node>, so there's no problem regarding polymorphic dispatch. However relAutoIndex in your production code is referenced by the type AutoIndexer<Relationship> and the mock relAutoIndexer in your test code is referenced by the type RelationshipAutoIndexer, so the wrong call is registered on the mock and then fails verification.

Your solution is either to upgrade the mockito version; the 1.9.0 RC1 is very stable and a final release should be coming your way. Or you can migrate your reference type (in your production code) from :

AutoIndexer<Relationship> relAutoIndexer = this.graphDb.index().getRelationshipAutoIndexer();

to :

RelationshipAutoIndexer relAutoIndexer = this.graphDb.index().getRelationshipAutoIndexer();

---

A few other remarks.

• You don't actually need to write an after method here as JUnit creates a new instance on each method run, so your method just adds code that will be done anyway. Note this isn't the case with TestNG.

• Instead of creating your mocks in the before method, you might want to use Mockito annotations. Don't forget the runner.

For example :

@RunWith(MockitoJUnitRunner.class)
public class YourTest {
    @Mock SomeType someTypeMock;
    // ...
}

• The stubbing code is a bit ugly for several reasons.

  • Your should write consistent stubs.

Why not write this in a cleaner way; for example referencing indexManager in both case :

IndexManager indexManager = mock(IndexManager.class);
when(graphDb.index()).thenReturn(indexManager);
when(indexManager.getNodeAutoIndexer()).thenReturn(nodeAutoIndexer);
when(indexManager.getRelationshipAutoIndexer()).thenReturn(relAutoIndexer);

Or don't reference it at all

IndexManager indexManager = mock(IndexManager.class);
when(graphDb.index()).thenReturn(indexManager);
when(graphDb.index().getNodeAutoIndexer()).thenReturn(nodeAutoIndexer);
when(graphDb.index().getRelationshipAutoIndexer()).thenReturn(relAutoIndexer);

Also having a mock that returns a mock is usually a sign of design smell. You are breaking the law of Demeter, and breaking it means you will experience difficult testing, bad maintainability, and difficult evolution. When I say that you could hear me whisper also (without the syllogisms) : it will cost you money. Don't write legacy code! If you are practicing TDD or BDD, you will identify these issues at design time for your own code, which is great to prevent them.

• However if you are dealing with legacy code, you can use this deep stubs syntax :

Using the static methods you could write this

GraphDatabaseService graphdb = mock(GraphDatabaseService.class, RETURNS_DEEP_STUBS);

Or using the annotation you could write this :

@Mock(answer = RETURNS_DEEP_STUBS) GraphDatabaseService graphdb;

And the stubbing :

when(graphDb.index().getNodeAutoIndexer()).thenReturn(nodeAutoIndexer);
when(graphDb.index().getRelationshipAutoIndexer()).thenReturn(relAutoIndexer);

To answer your question directly, you cannot use Mockito to stub a method called from the constructor. Mockito needs an instance of the class before you can begin mocking, and you haven't given yourself a way to create an instance for testing.

More generally, as mentioned in Effective Java item 17, you should not call overridable methods from constructors. If you do so, for instance, you could provide an override in a subclass that refers to a final field but that runs before the final field is set. It probably won't get you in trouble here, but it's a bad habit in Java.

Luckily, you can restructure your code to do this very easily:

public class MyClass {
  public MyClass() {
    this(true);
  }

  /** For testing. */
  MyClass(boolean runSetup) {
    if (runSetup) {
      setup();
    }
  }

  /* ... */
}

To make it even more obvious, you can make the one-parameter MyClass constructor private, and provide a public static factory method:

/* ... */
  public static MyClass createForTesting() {
    return new MyClass(false);
  }

  private MyClass(boolean runSetup) {
/* ... */

Though some developers think it is a bad practice to write any code in methods that is used mostly for tests, remember that you are in charge of the design of your code, and tests are one of few consumers you absolutely know you will need to accommodate. Though it's still a good idea to avoid explicit test setup in "production" code, creating extra methods or overloads for the sake of testing will usually make your code cleaner overall and can drastically improve your test coverage and readability.