I have a Spring Boot 2.6.0 application that depends on libraries which use fields initialised with SpEL syntax, e.g.
#Value("${some.value.in.library}") aField;
I attempted to write a Spring MVC test for a controller with #WebMvcTest. However I kept getting failed to load applicationcontext errors due to a SpELEvaluationException for the fields mentioned above.
Even after trying many, many, different test annotations such as #ConfigurationProperties and other fixes. I could see the properties that are defined in application-test.yml being loaded in the startup logs but they were not getting injected into the class.
Finally, I tried using:
MockMvcBuilders.standaloneSetup
and the test was able to start and run successfully.
Unfortunately, due to the complexity of the application and the use of a library where the problem occurs, it is difficult to create a minimal reproducable example.
So my question is, why does MockMvcBuilders.standaloneSetup work, whereas #MockMvcTest doesn't?
This answer suggests that #MockMvcTest should not load the full application context. But that does not seem to be the case based on the failed to load applicationcontext errors.
Related
I was reading through the guide for using Spring Boot Test and there was a paragraph that got me confused.
“As our profiles get richer, it's tempting to swap every now and then in our integration tests. There are convenient tools to do so, like #ActiveProfiles. However, every time we pull a test with a new profile, a new ApplicationContext gets created.”
https://www.baeldung.com/spring-tests
So it assumes that if all tests are run under the same profile, there is only one ApplicationContext created — but how is it possible? I thought that all the objects are recreated for each test suite anyway. Am I missing something?
The official reference says that it's cached.
https://docs.spring.io/spring-framework/docs/current/spring-framework-reference/testing.html#testing-ctx-management
But how does it get loaded into the JUnit runner or Spock one across multiple test suites?
What was missing in my understanding is the fact that all the test suites are run as a part of a single program, so it's easy to cache any objects that are required by all of them, including Spring context.
I just tried to run an OptaPlanner project in Spring Boot, but there's only very simple text in OptaPlanner User Guide for Spring.
Actually, I think it is very easy to copy all domain objects, configuration files and drools files from an OptaPlanner project to Spring Boot project without any changes, but the only question is how to call Solver's solve method.
I made it run after Spring Boot startup with a class (named CommandLineAppStartupRunner) which implements CommandLineRunner interface, and I called solve method in its run method. Finally, I got an exception like follows:
Caused by: java.lang.IllegalArgumentException: Can not set org.optaplanner.core.api.score.buildin.hardmediumsoft.HardMediumSoftScore field springbootcloudbalance.domain.CloudBalance.score to springbootcloudbalance.domain.CloudBalance
at sun.reflect.UnsafeFieldAccessorImpl.throwSetIllegalArgumentException(UnsafeFieldAccessorImpl.java:167)
at sun.reflect.UnsafeFieldAccessorImpl.throwSetIllegalArgumentException(UnsafeFieldAccessorImpl.java:171)
at sun.reflect.UnsafeFieldAccessorImpl.ensureObj(UnsafeFieldAccessorImpl.java:58)
at sun.reflect.UnsafeObjectFieldAccessorImpl.get(UnsafeObjectFieldAccessorImpl.java:36)
at java.lang.reflect.Field.get(Field.java:393)
at org.optaplanner.core.impl.domain.common.accessor.ReflectionFieldMemberAccessor.executeGetter(ReflectionFieldMemberAccessor.java:54)
at org.optaplanner.core.impl.domain.solution.descriptor.SolutionDescriptor.getScore(SolutionDescriptor.java:1071)
at org.optaplanner.core.impl.score.director.AbstractScoreDirector.cloneSolution(AbstractScoreDirector.java:212)
at org.optaplanner.core.impl.solver.scope.DefaultSolverScope.setWorkingSolutionFromBestSolution(DefaultSolverScope.java:230)
at org.optaplanner.core.impl.solver.AbstractSolver.solvingStarted(AbstractSolver.java:75)
at org.optaplanner.core.impl.solver.DefaultSolver.solvingStarted(DefaultSolver.java:210)
at org.optaplanner.core.impl.solver.DefaultSolver.solve(DefaultSolver.java:190)
at springbootcloudbalance.CommandLineAppStartupRunner.run(CommandLineAppStartupRunner.java:55)
at org.springframework.boot.SpringApplication.callRunner(SpringApplication.java:818)
... 10 common frames omitted
I checked the code, and found the exception throws because the object from field.getDeclaringClass() is a different instance from the one from var1.getClass(). I'm afraid it due to the implementation of java reflection conflicts between OptaPlanner and Spring Boot.
The version I used is as follows:
OptaPlanner 7.11.0.Final
Spring Boot 2.0.5.RELEASE
JVM 1.8.0_181
Removing the spring-boot-devtools dependency fixes this error. Another SO question similar to this one explains it has something to do with different classloaders: Optaplanner's Drools working memory is empty. The accepted answer also mentions a possible fix:
To fix it, configure spring dev tools to load Drools libraries in the RestartClassLoader, together with the project's classes: using-boot-devtools-customizing-classload
Nick's answer is correct. This is just to figure out what's going on.
This line means that optaplanner is extracting CloudBalance.getScore():
at org.optaplanner.core.impl.domain.solution.descriptor.SolutionDescriptor.getScore(SolutionDescriptor.java:1071)
This line means that it's using a ReflectionFieldMemberAccessor for that, which is just way to read a field through reflection (including private fields):
at org.optaplanner.core.impl.domain.common.accessor.ReflectionFieldMemberAccessor.executeGetter(ReflectionFieldMemberAccessor.java:54)
Now the error message is where it gets interesting:
Can not set ...HardMediumSoftScore field ...CloudBalance.score to ...CloudBalance
It looks like basically doing:
CloudBalance cloudBalance2 = cloudBalance.getScore();
Huh?
Musa provided this answer, but someone deleted it, despite that the JIRA link is extremely relevant, as it points out which version of OptaPlanner will deal better with this problem:
"An issue was submitted to OptaPlanner to provide better error messages for such cases: https://issues.jboss.org/browse/PLANNER-1586. Feel free to add any comments or suggestions."
I have a large spring project, using xml configuration. I'm looking for a quick way to verify changes to the xml configuration.
I can load the whole project locally - the problem is this takes more than 5 minutes, loads a huge amount of data.
My XML editor catches XML formatting errors.
I'm looking for something intermediate - to catch obvious problems like references to beans that aren't defined, or calling constructors with the wrong arguments. Is there a quick way to do this, without having to actually invoke all the constructors and bring up the whole environment?
I'm building with Maven and editing with Eclipse, although my question isn't specific to either.
Since you already use Eclipse, you could try Spring Tool Suite (comes either standalone or as an add-on). It's essentially Eclipse with extra Spring-specific features, like Beans Validator. I'm not sure how thorough the validation is, but it should catch most configuration problems.
It's maintained by SpringSource so its integration with Spring "just works" and it's guaranteed not be more or less in sync with Spring Framework's release cycle.
Beanoh :
http://beanoh.org/overview.html#Verify
this project does exactly what I'm looking for. Verify obvious problems with spring config, but without the overhead of initializing everything.
You can use a Spring testing support to integration test your Spring configuration. However if the loading of the context is taking 5 mins, then the tests will also take the same amount of time. Spring does cache the context so if you have multiple tests using the same set of Spring contexts, then once cached the tests should be very quick.
I can suggest a few ways to more efficiently test your configuration:
Organize your project in modules, with each module being responsible for its own Spring configuration - this way, each module can be independently developed and tested.
If you have a modular structure, the testing can be more localized by mocking out the dependent modules, again this is for speed.
We are using cucumber-jvm to write an integration test layer in our application. One of the challenges we are finding is managing the database between the tests and the web application.
A typical scenario is that we want to persist some entities in a Given step of a scenario, then perform some actions on the user interface that may, in turn, persoist more entities. At the end, we want to clean the database. Because the cucumber-jvm tests are in one jvm and the web application is running in another jvm we cannot share a transaction (at least in a way of which I am aware) so the database must be cleaned manually.
My initial thought was to use an Embedded Tomcat server running off of an embedded in-memory database (HSQLDB) in the same JVM as the cucumber-jvm test. This way we might be able to share a single spring container, and by extension a single transaction, from which all objects could be retrieved.
During my initial tests it looks like Spring gets loaded and configured twice: once when the test starts and the cucumber.xml is read, and a second time when the embedded tomcat starts and the web application reads its applicationContext.xml. These appear to be in two completely separate containers because if I try to resolve an object in one container that is specified in the other container then it doesn't resolve. If I duplicate my configuration then I get errors about duplicate beans with the same id.
Is there a way that I can tell Spring to use the same container for both my test application and the embedded tomcat?
I'm using Spring 3.2.2.GA and Embedded Tomcat 7.0.39 (latest versions of both libraries).
Am I crazy? Do I need to provide more technical details? Apologies if I use some incorrect terminology.
Thanks
p.s. If my problem seems familiar to you and you can suggest an alternative solution to the one I am trying, please let me know!
Jeff,
It is normal that spring is loaded twice. There are two places where two spring contexts are created:
In the servlet container listener org.springframework.web.context.ContextLoaderListener that is configured in web.xml. This one reads its configuration from the file set by the context-param contextConfigLocation.
In the implementation of ObjectFactory provided by cucumber-spring plugin cucumber.runtime.java.spring.SpringFactory. This one reads its configuration from cucumber.xml.
The two spring contexts are totally different and their instances are kept in two different places. As a servlet context attribute for the former and kept by the JavaBackend for the latter.
When starting the embedded tomcat, it is possible to get access to the servlet context and thus set ourself the spring context used bt tomcat with the one from cucumber. But, spring has a special class called WebApplicationContext for context used in a servlet container. The cucumber SpringFactory on other hand creates its context through ClassPathXmlApplicationContext. So unless there is a way to specify the type of application context from the xml config, we will have to provide an ObjectFactory that shoots a WebApplicationContext.
What we can do is to have two web.xml. One for the normal and one for the test. For the test, we use our version of the ContexLoader listener.
So this is a rather "big" question, but what I'm trying to accomplish is the following:
I have a Spring application, MVC, JDBC (MySQL) and JSP running on tomcat.
My objective is to test the entire "stack" using a proper method.
What I have so far is Junit using Selenium to simulate an actual user interacting with the application (requires a dummy account for that), and performing different validations such as, see if element is present in the page, see if the database has a specific value or if a value matches the database.
1st concern is that this is actually using the database so it's hard to test certain scenarios. I would really like to be able to mock the database. Have it emulate specific account configs, data states etc
2nd concern is that given the fact that I use what is in the database, and data is continuously changing, it is hard to predict behavior, and therefore properly asserting
I looked at Spring Test but it allows for testing outside a servlet container, so no JSP and no Javascript testing possible.
I saw DBUtils documentation but not sure if it will help me in this case
So, to my fellow developers, I would like to ask for tips to:
Run selenium tests on top of a mocked database
Allow different configs per test
Keep compatibility with Maven/Gradle
I have started with an ordered autowire feature to support this kind of stubbing.
It's basically an idea that i took over from the Seam framework i was working with in the past but i couldnt find yet a similar thing in spring.
The idea is to have a precedence annotation (fw, app,mock,...) that will be used to resolve the current implementation of an autowired bean. This is easy already in xml but not with java config.
So we have our normal repository beans in with app precedence and a test package stubbing these classes with mock precedence.
If both are in the classpath spring would normally fail with a duplicate bean found exception. In our case the extended beanfactory simply takes the bean with the highest precedence.
Im not sure if the order annotation of spring could be used directly but i prefered to have "well defined" precedence scopes anyway, so it will be clear for our developers what this is about.
! While this is a nice approach to stub so beans for testing i would not use it to replace a database definition but rather go with an inmemory database like hsql, like some previous answers mentionned already. !