Spring Data: Service layer unit testing

Question

In my project I'm having trouble doing unit testing. One issue is that just doing an integration test is much faster to write and also tests that the components actually work together. Unit testing novel "algorithms" or so seems much easier. Unit Testing service classes it just feels wrong and useless.

I'm using mockito to mock spring data repository (and hence DB access). The thing is if i tell the mocked repository to return entity A on method call getById it will obviously return that and the service will return it too. Yes, the service does some extra stuff, but very minor things, like load lazy collections (from hibernate). Obviously I don't have any lazy collections (proxies) in a unit test.

Example:

@Test
public void testGetById() {
    System.out.println("getById");
    TestCompound expResult = new TestCompound(id, "Test Compound", "9999-99-9", null, null, null);

    TestCompoundRepository mockedRepository = mock(TestCompoundRepository.class);
    when(mockedRepository.findOne(id)).thenReturn(expResult);

    ReflectionTestUtils.setField(testCompoundService, "testCompoundRepository",
            mockedRepository, TestCompoundRepository.class);

    TestCompound result = testCompoundService.getById(id);
    assertEquals(expResult, result);
}

hooray, the rest succeeds. What a surprise! Not really no.

Can some one explain to me what I'm doing wrong? Or else what the point of such a test is? I mean I tell to return expResult and then it is returned. Wow. What a surprise! Feels like I'm testing if mockito works and not my Service.

EDIT:

The only benefit I see if some were stupid error happens like leaving an unwanted line there that sets return value to null or something similar stupid. Such cases would be caught by the unit test. Still the "reward-effort" ratio seems bad?

Answer 1

Question might be a bit old but I will put an answer in case someone stumbles across.

• I'm using Mockito and JUnit.
• AccountRepository is a plain spring data repository extending JPARepository.
• Account is a plain JPA entity.

To test your services and mock Spring Data repositories, you need something like below.

package foo.bar.service.impl;

import foo.bar.data.entity.Account;
import foo.bar.data.repository.AccountRepository;
import foo.bar.service.AccountService;

import org.junit.Assert;
import org.junit.Test;
import org.junit.runner.RunWith;
import org.mockito.InjectMocks;
import org.mockito.Mock;
import org.mockito.Mockito;
import org.mockito.runners.MockitoJUnitRunner;

@RunWith(MockitoJUnitRunner.class)
public class AccountServiceImplTest {

    @Mock
    private static AccountRepository accountRepository;

    @InjectMocks
    private static AccountService accountService = new AccountServiceImpl();

    private Account account;

    @Test
    public void testFindAccount() {

        Integer accountId = new Integer(1);

        account = new Account();
        account.setId(accountId);
        account.setName("Account name");
        account.setCode("Accont code");
        account.setDescription("Account description");

        Mockito.when(accountRepository.findOne(accountId)).thenReturn(account);

        Account retrivedAccount = accountService.findAccount(accountId);

        Assert.assertEquals(account, retrivedAccount);

    }

}

Answer 2

One of the reasons I like testing my Spring Data repositories is to test that I have defined my JPA mappings correctly. I do not use a mocking framework for these tests, I use the Spring Test framework which actually bootstraps the container allowing me to autowire the actual repository into the Junit test so that I may execute tests against it.

I agree with your thoughts that mocking the repository is pretty useless. Since your using Spring I would suggest leveraging the Spring Test framework to perform real tests against your repositories, which can be executed against an embedded database such as H2 in a more unit test based fashion or your actual database implementation such as Oracle or MySql, to conduct more of an integration test. (Execute these against a copy of a development database) These tests will reveal fallacies in your JPA mappings and other items such as improper cascades setup in the database.

Here is an example of one of my tests on GitHub. Notice how the framework actually autowires the repository into the test. The repository also contains an example of how to configure the Spring Test framework, which I have also demonstrated in this blog post.

In conclusion, I do not believe you will receive any of the benefits of testing a repository that I have discussed from using a mock of the repository.

One additional note I wanted to add, is that mocks are not really intended for use in the actual class under test. Their use is for providing required dependencies to a class under test.

Answer 3

You can use this library: https://github.com/agileapes/spring-data-mock

This will mock your repository for you, while allowing you to implement custom functionality for any method as well as your native query methods.

Answer 4

You exactly right. It is clear unit test. And it will never fail (so, it is useless) I think you need at integration test to test real JPA repository with real database (H2 in memory for example) (as I always do).

And it is better to test your services (theirs interfaces). If after some time you will change your storage (to Mongo for example) - you will be able to use your service tests to ensure all works as before.

After some time you will be suprised how many DB\JPA-related problems (constraints, optimistic locks, lazy-loading, duplicate id, some hibernate issues and so on) you find.

Also, try to develop via tests - not just write test after implementation. Instead before creation of new method in service - create test for it, implement service method and only after just recheck it in real application. At least it is much faster to start test than a server.

So, do not create tests to have a lot of them. Find how they may help you.

Usage of mocks for repositories is not good idea. Test how your services work together with Hibernate\JPA\Database. Most part of problems is located beetwen layers.

Answer 5

You can mock the repository and inject it to the service, this is the way; but, if you just instantiate the service with @Mock of repositories, it would be better, if you define the repositories as private final fields in the service and use a constructor of all repositories. That way, if you add another repository to the service, the test will fail and you have to change it, which is the purpose.

Imagine this service:

class OrderService {
    private final UserRepository userRepos;

    public OrderService(UserRepository userRepos) {
        this.userRepos = userRepos;
    }

    ...

}

And this test:

class OrderServiceTests {
    @Mock
    private UserRepository userRepos;

    private OrderService service;

    private OrderServiceTests() {
        this.service = new OrderService(this.userRepos);
    }
}

Now, if we add another dependency to the service:

class OrderService {
    private final UserRepository userRepos;
    private final AddressRepository addRepos;

    public OrderService(UserRepository userRepos, AddressRepository addRepos) {
        this.userRepos = userRepos;
        this.addRepos = addRepos;

    ...

}

The previous test will fail because the constructor has changed. If you use @InjectMocks this will not happen; the injection happens behind the curtain and we are not clear what happens; this may not be desirable.

Another thing is, I don't agree that integration test will cover all the cases that unit tests will cover; it may but not always the case. Even the controller can be unit-tested with mocks; after all the tests are meant to cover all the code we have written, so they must be fine-grained; imagine when we follow TTD and we only complete the controller and services level: how we proceed without controller unit testing?

Answer 6

Assuming that we have the below Service class

@Service 
public class EmployeeServiceImpl implements EmployeeService {

    @Autowired
    private EmployeeRepository employeeRepository;

    @Override
    public Employee getEmployeeByName(String name) {
        return employeeRepository.findByName(name);
    }
}

Test class:

@RunWith(SpringRunner.class)
public class EmployeeServiceImplIntegrationTest {

    @TestConfiguration
    static class EmployeeServiceImplTestContextConfiguration {

       @Bean
       public EmployeeService employeeService() {
           return new EmployeeServiceImpl();
        }
    }

    @Autowired
    private EmployeeService employeeService;

    @MockBean
    private EmployeeRepository employeeRepository;

// write test cases here
}

To check the Service class, we need to have an instance of Service class created and available as a @Bean so that we can @Autowire it in our test class. This configuration is achieved by using the @TestConfiguration annotation.

During component scanning, we might find components or configurations created only for specific tests accidentally get picked up everywhere. To help prevent that, Spring Boot provides @TestConfiguration annotation that can be used on classes in src/test/java to indicate that they should not be picked up by scanning.

Another interesting thing here is the use of @MockBean. It creates a Mock for the EmployeeRepository which can be used to bypass the call to the actual EmployeeRepository:

@Before
public void setUp() {
    Employee alex = new Employee("alex");

    Mockito.when(employeeRepository.findByName(alex.getName()))
      .thenReturn(alex);
}

After the setup, we can easily test our service like:

@Test
public void whenValidName_thenEmployeeShouldBeFound() {
    String name = "alex";
    Employee found = employeeService.getEmployeeByName(name);

    assertThat(found.getName())isEqualTo(name);
}

For more in depth knowledge check: https://www.baeldung.com/spring-boot-testing