Dependency Management In Unit Tests

Dependency management

● In practice, cases of well-isolated code are quite rare.

Types of dependencies

● Injecting dependencies (respecting DI)

class A { B b;

public setB (B b) {this.b = b;}

...

}

Compiled dependencies (for instance referencies Filesystem API)

class A {

public void methodToTest () {

...

InputFile inStream = new InputFileStream ("input.txt");

...

inStream.read (buffer);

...

}

}

Dependency Management (“Bad Example”)

class A {
private B bInstance;
public setB (B bInstance) {
this.bInstance = bInstance;
}
public int myMethod (int number) {
...
bInstance.doSomething (1,2);
...
return number;
}
}
class B {
public void doSomething () {
...
}
}

class MyATest extends TestCase {
@Test
public void testWithSImpleInput () {
// initialization
A a = new A ();
B b = new b ();
a.setB (b);
// execution
int actual = a.myMethod (4);
// verification
int expected = 2;
assertEquals (expected, actual);
}
}

Unit Test Problems without Dealing with Addictions
● Floating errors. The difficulty of finding them. Loss of time
● Without eliminating dependencies, unit tests are de facto integration tests.

Dependency management

● For successful unit testing, it is necessary to eliminate all dependencies. Otherwise, potential defects in dependencies can affect the success of tests and complicate the search for defects.

● The solution is that all dependent objects are replaced by “fake” objects that behave exactly as you expect.

● Mocks ... replace a class with a small substitute (“mocks”) that implements the same interface.

● Shims ... modify the compiled code at run time, to inject and run a substitute (“shim”).

● Effective means are needed to create such fake objects on the fly.

Mocking frameworks

Java

C #

Injecting dependencies