CPS Persistence layer (semi-integration) testing
Table Of Contents
Related Jiras:
CPS-95: Persistence layer unit testing - setup and exampleClosed
CPS-124: Update Jenkins configuration to support Test Containers Delivered
CPS-128: Share container based persistince testing knowledge with teamDelivered
Overview
The CPS-RI module which is responsible for data persistence and retrieval from the database is based on Spring Data framework.
Following components are used:
Persistence service implementations (logic layer) → these are defined as Java classes
JPA repositories (data access layer) → these are defined as Java interfaces, corresponding objects
are created by Spring framework at runtime
In order to reach the desirable level of reliability it's expected the functionality to be covered with tests.
Below is comparison between unit test vs integration test approaches.
While using unit tests is a common approach for Java application testing, it seems insufficient in a context of testing the
actual data persistence/retrieval functionality:
Test covers a small slice of functionality while the major piece served by external services (Spring data framework,
JDBC driver and database itself) remain not used (not tested)JPA repositories can be only used as mocks to test the service implementation but JPA repositories are not subject
for tests themselves because of definition as interfacesThe effort to setup the expected behaviors for JPA repository mocks is comparable with (or exceeds) the effort
required to setup the actual data for same cases;
From other hand the benefits of using integration tests with real database instance are following:
All levels of abstraction are involved in tests, it means all of them are impacting on test results same way as within
a deployed applicationJPA repositories are subjects of test
Complete validation of expected behavior, including database schema initialization
Can be used on build, keeping desired level of reliability and consistency
Tests can be used on development stage making the database deployment (on dev environment) optional
Components
Components used to serve the integration tests are shown on diagram below.
Docker provides a container with running database instance which is used during the test
Test container component communicates with Docker via CLI (or Engine API depending on Docker deployment),
manages required container to be created/started before test and stopped/removed after the test is completedTest container component is used by JUnit test as a Class Rule
The connection to test database is served using standard Spring framework components
More details on implementation are below.
Test Containers
Running instance of database for integration test is provided by TestContainers Java library component.
Docker availability is a prerequisite.
Test container life circle management
Test Containers library is integrated with JUnit via Rule interface implementation:
On start() method invocation the requested test container is created (if absent) within a Docker and started (if not running yet)
On stop() method invocation the test container is stopped and removed
If annotated as @Rule
the test container creation/start then stop/removal operations will be performed for each test method.
If annotated as @ClassRule
these operation will be performed for each test class.
In order to save the resources (and all tests execution time as result) the default TestContainer implementation was wrapped to custom
(singleton type of) artifact. This way the database test container is initialized only once for all the tests. The test container removal is
linked to JVM shutdown event to avoid this container remain running in docker.
NB. The default remaining test containers removal (on JVM termination) was served by dedicated RYUK container (part of TestContainers library).
However this container requires running as privileged one, so it was disabled in order make TestContainers available on Jenkins. It made it
necessary to explicitly invoke test container removal logic on JVM termination event.
@SpringBootTest
Using @SpringBootTest
annotation for tests allows @Autowired
Spring components (including JPA repository instances) initialization for testing
same way as it's done on regular Spring Boot application execution.
To be used the @SpringBootApplication
annotated class is required. However both application class and associated configuration are allocated
in CPS-REST maven module which cannot be accessed from CPS-RI module on build (when the tests classes are compiled and executed).
So in order to fulfill the @SpringBootTest requirements and to properly initialize DataSource object the substitutions for both Spring Boot application
class and configuration (DataSource properties) are set directly within CPS-RI module (test scope).
NB. There is no need to initialize the Spring Boot application within CPS-RI in same scale as it's done for CPS-REST module, so the
application.yaml configuration within CPS-RI module contains mainly database related parameters. In case these parameters are updated
within primary configuration file same changes require to be provided to test configuration file as well.
Database initialization
The datasource is configured in application.yml test configuration file like below
spring:
datasource:
url: ${DB_URL}
username: ${DB_USERNAME}
password: ${DB_PASSWORD}
driverClassName: org.postgresql.Driver
initialization-mode: always
The database connection parameters are expected to be defined using environment variables.
Due to ports exposed for a test containers (part of DB_URL) are variable (to allow concurrent tests executions and to avoid conflicts with
existing containers) the expected variables are set after the test container being initialized directly from TestContainer wrapper artifact
DatabaseTestContainer.java
@Override
public void start() {
super.start();
System.setProperty("DB_URL", databaseTestContainer.getJdbcUrl());
System.setProperty("DB_USERNAME", databaseTestContainer.getUsername());
System.setProperty("DB_PASSWORD", databaseTestContainer.getPassword());
}
The database schema initialization is performed by Spring framework (same way for both runtime and testing) using schema.sql file.
See https://docs.spring.io/spring-boot/docs/current/reference/html/howto.html#howto-database-initialization
Test data
Spring framework allows data setting for testing using @Sql
annotation.
Single or multiple SQL scripts can be defined to be executed before the test method (resources folder is a root).
@Test
@Sql({"/data/clear.sql", "/data/test-data.sql"})
public void test() {
// test
}
Note. During 2023 CPS moved away from using SQL scripts to insert Test Data. Instead we insert dat programmatically during the test for more details see CPS (semi-) Integration Tests
Resources
Frameworks and libraries:
Tutorials: