How to load DBUnit test data once per case with Spring Test - junit

Question

Spring Test helpfully rolls back any changes made to the database within a test method. This means that it is not necessary to take the time to delete/reload the test data before each test method.
But if you use the #BeforeClass Junit annotation, then that forces the data loader to be static. A question that is explored here: Why must jUnit's fixtureSetup be static?
If the data initialization method is static, so must the data connection methods and the data source..and on and on...forcing everything to be static...which won't work. At which point, I ask - what good is Spring Test's ability to rollback changes when you have to delete/reload the test data anyway for every test??!?!

Answer 1

One approach that works is to create a "data initialiser" class, add it to a test Spring application context that also has your data source, and wire this application context into your tests. This relies on the fact that Spring caches the application context between test invocations.
For example, a test superclass:
#RunWith(SpringJUnit4ClassRunner.class)
#ContextConfiguration(locations={"classpath:test-application-context.xml"})
#Transactional
public abstract class DataLoadingTest {
#Autowired
protected DatabaseInitialiser databaseInitialiser;
}
With test-application-context.xml:
<beans xmlns="http://www.springframework.org/schema/beans"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans.xsd">
<bean id="dataSource" .../>
<bean class="DatabaseInitialiser">
<property name="dataSource" ref="dataSource"/>
</bean>
</beans>
And
public class DatabaseInitialiser extends JdbcDaoSupport {
#PostConstruct
public void load() {
// Initialise your database here: create schema, use DBUnit to load data, etc.
}
}
In this example:
all tests that rely on the database extend DataLoadingTest;
Spring initialises the application context upon first test invocation;
this calls DatabaseInitialiser.load(), via the #PostConstruct annotation;
Spring keeps the application context in a cache;
further test invocations wire in the DatabaseInitialiser from the application context, which is already cached;
tests are transactional, and roll back at the end to the initial data set.
Likewise, DatabaseInitialiser can have a method annotated #PostDestroy to perform any rollback necessary at the end of the whole test run.

Answer 2

We use DBUnit in conjunction with Spring Test extensively. But we do not use the DBUnit functionality to delete data at the end of the test.
We put a bunch of DBUnit inserts for our test data in the #Before method to initialise the test. Then when the test is complete we let the spring rollback functionality bring the database back to its original state.
The biggest problem we have with this is that the DBUnit data has to be loaded before each test, which can be a major performance hit. Most of our tests using DBUnit are read only, testing the behaviour of the application based on certain predefined behaviour. So we have a habit of creating master tests that then run all the fine grain tests in a batch within the same transaction.

Answer 3

Spring Test and DbUnit is two excellent frameworks. But it doesn't make sense to combine them. Since Spring Test execute a rollback on the connection, it cleans up afterwards, while DbUnit cleans up and insert test data in the #Before method.
Use Spring if you're not dependent on any dynamic data and dbUnit otherwise.

Answer 4

Methods annotated with #BeforeTransaction run, like its name suggests, before the transaction of each test is started. If in such method you can detect if the test data is loaded, then one can load the data when needed.
Beware though that the data is left in your (in-memory) database for all subsequent tests.
We use this to load "static" data that would, in a production environment, also be bootstrapped into our database when starting it. This way we actually use exactly the same code and data for our tests, rather than relying on (DbUnit) exports that might become outdated.

Answer 5

You can create a data initializer "bean", since the config is only run once. It follows the same principle as the main answer, but with less code and classes
#Configuration
class DBUnitTest_Config {
protected String PATH = "";
#Bean
public DataSetConfig setupData(DataSource dataSource) throws SQLException {
DataSetExecutorImpl executor = DataSetExecutorImpl.instance(new ConnectionHolderImpl(dataSource.getConnection()));
DataSetConfig dataSetConfig = new DataSetConfig(PATH);
executor.createDataSet(dataSetConfig);
return dataSetConfig;
}
}