BeanCreationException error when referencing configuration class inside a service class - java

I am trying to #Autowire a #Configuration class inside a #Service class. basically my #Configuration class contains mapping to my custom .properties file. When i try to autowire my configuration class inside my service class, BeanCreationException occurs. I am not sure what happen. Just followed the guide on creating Property classes from spring. There must be something i missed out.
Also, when i try to autowire #Configuration class to another #Configuration class, it runs smoothly
Currently, i know that, prop is always null because when i remove prop.getUploadFileLocation() call, everything will be fine. There must be something wrong during autowiring.
Here is my Service class
#Service
public class ImageService {
public static Logger logger = Logger.getLogger(ImageService.class.getName());
#Autowired
MyProperties prop;
private final String FILE_UPLOAD_LOCATION = prop.getUploadFileLocation() +"uploads/images/";
public void upload(String base64ImageFIle) throws IOException {
logger.info(FILE_UPLOAD_LOCATION);
}
}
Here is my Configuration class
#Data
#Configuration
#ConfigurationProperties (prefix = "my")
public class MyProperties {
private String resourceLocation;
private String resourceUrl;
public String getUploadFileLocation() {
return getResourceLocation().replace("file:///", "");
}
public String getBaseResourceUrl() {
return getResourceUrl().replace("**", "");
}
}
And here is where i can successfully use MyProperties
#Configuration
public class StaticResourceConfiguration implements WebMvcConfigurer {
#Autowired
MyProperties prop;
#Override
public void addResourceHandlers(ResourceHandlerRegistry registry) {
registry.addResourceHandler(prop.getResourceUrl())
.addResourceLocations(prop.getResourceLocation());
}
}

The issue is that you are trying to use an autowired field to set the value in an inline field assignment.
That means
private final String FILE_UPLOAD_LOCATION = prop.getUploadFileLocation() +"uploads/images/";
is executed before the prop is autowired, meaning it will always be null
The way to mitigate this would be to use constructor injection instead.
#Service
public class ImageService {
//Fine since you are using static method
public static Logger logger = Logger.getLogger(ImageService.class.getName());
//Not needed if you are only using it to set FILE_UPLOAD_LOCATION
//Allows field to be final
private final MyProperties prop;
//Still final
private final String FILE_UPLOAD_LOCATION;
//No need for #Autowired since implicit on component constructors
ImageService(MyProperties prop){
//Again not needed if you aren't going to use anywhere else in the class
this.prop = prop;
FILE_UPLOAD_LOCATION = prop.getUploadFileLocation() +"uploads/images/";
}
public void upload(String base64ImageFIle) throws IOException {
logger.info(FILE_UPLOAD_LOCATION);
}
}
See this question for why constructor is preferred over #autowired in general

If you need MyProperties bean to be created before StaticResourceConfiguration bean, you can put #ConditionalOnBean(MyProperties.class) as following. Spring will make sure MyProperties is there before processing StaticResourceConfiguration.
#Configuration
#ConditionalOnBean(MyProperties.class)
public class StaticResourceConfiguration implements WebMvcConfigurer {

Related

How does Autowire work when referring only to a base class

I am trying to extend existing code which auto-wires a number of components which share the same base class and are only referred to by their base class, ei:
#Configuration
public class NavigationHotKeyBindConfigs {
private static final Logger logger = LogManager.getLogger(NavigationHotKeyBindConfigs.class);
#Autowired
private HotKeyConfig hotKeyConfig;
#Autowired
private ANavigation upNavigation;
#Autowired
private ANavigation downNavigation;
#Autowired
private ANavigation centerOnSelection;
#Autowired
private ANavigation handVerify;
public abstract class ANavigation {
#Autowired
protected ScrollSchedulerConfigurer scrollSchedulerConfigurer;
public abstract void execute(DetectionView DetectionView);
}
#Component
public class DownNavigation extends ANavigation{
private static final Logger logger = LogManager.getLogger(DownNavigation.class);
public void execute(DetectionView detectionView){
logger.info(String.format("Received: Navigation key %s", ENavigationKey.DOWN.name()));
scrollSchedulerConfigurer.stop();
detectionView.selectNextDetection();
}
}
#Component
public class UpNavigation extends ANavigation{
private static final Logger logger = LogManager.getLogger(UpNavigation.class);
public void execute(DetectionView detectionView){
logger.info(String.format("Received: Navigation key %s", ENavigationKey.Up.name()));
scrollSchedulerConfigurer.stop();
detectionView.selectPrevDetection();
}
}
I don't understand how spring is ever supposed to match up upNavigation/downNavigation with the UpNavigation/DownNavigation components respectively. Was this just bad design from the original author or is there a subtlety I'm missing?
This lack of understanding is also causing some issues as the centerOnSelection and handVerify objects end up NULL after autowiring though they follow the same pattern in every way I can see. Interestingly, this last behavior is different when run in the intellij IDE vs when run out of a .jar
You can do it like this ...
Let's say we have two beans which have the same base class or interface
#Component("fooFormatter")
public class FooFormatter implements Formatter {
public String format() {
return "foo";
}
}
#Component("barFormatter")
public class BarFormatter implements Formatter {
public String format() {
return "bar";
}
}
Then we inject them like this ...
public class FooService {
#Autowired
#Qualifier("fooFormatter")
private Formatter formatter;
}
From the point of view of clarity and maintainability this seems to be the best solution
#Autowired first looks at the type of the variable and then the name of the variable to match the correct bean. By default, the name of a class marked with #Component is its short name, e.g. upNavigation for the UpNavigation class.
#Autowired
private ANavigation upNavigation;
is the same as
#Autowired
#Qualifier("upNavigation")
private ANavigation upNavigation;
And
#Component
public class UpNavigation
is the same as
#Component("upNavigation")
public class UpNavigation
Since the names match, Spring is able to find the corresponding bean.

Multiple instances of a bean of one class filled with values from application.properties

I intend to write some HealtCheckContributors for a Spring Boot application using spring-boot-actuator. Hence, I implemented two of them. they are intended for checking the health of different apps, of course, but have a nearly identical structure, except the configuration properties, ...
SonarQube complains about that and I wonder if it is possible to have a single health check class but instantiated as many times as defined in application.properties.
An example:
application.properties:
# actuator
app1.management.baseUrl=http://localhost:10000
app1.management.name=app1HealthCheckContributor
app2.management.basUrl=http://localhost:10001
app2.management.name=app2HealthCheckContributor
HealthCheckContributor for app1:
#Slf4j
#Component("xxx")
public class App1HealthCheckContributor extends AbstractHealthIndicator {
private final App1Properties app1Properties;
public App1HealthCheckContributor(final App1Properties app1Properties) {
this.app1Properties = app1Properties;
}
#Override
protected void doHealthCheck(Health.Builder builder) {...}
}
...and this code for each HealthCheckContributor only distinct in its appXProperties.
Isn't it possible to have some kind of base class like:
#Slf4j
#Component()
public class MyHealthCheckContributor extends AbstractHealthIndicator {
private final MyProperties myProperties;
public MyHealthCheckContributor(final MyProperties myProperties) {
this.myProperties = myProperties;
}
#Override
protected void doHealthCheck(Health.Builder builder) {...}
}
and let Spring Boot take care of instantiating two HealthCheckContributors (in our case App1HealthCheckContributor and App2HealthCheckContributor)?
This would eliminate code duplication.
An example of the properties class file:
#Slf4j
#Data
#ConfigurationProperties(prefix = "app1.management")
public class App1Properties {
private String baseUrl;
private String ...;
}
How can I achieve this and how must an application.properties file looks like to achieve what I intend to do?
The final question: How to test multiple instance creation of a bean of one class filled with values from application.properties?
Assuming the code in doHealthCheck is exactly the same for all apps to be checked you could do the following.
You would start by creating a single health check class:
#Slf4j
public class AppHealthCheckContributor extends AbstractHealthIndicator {
private final AppProperties appProperties;
public App1HealthCheckContributor(final AppProperties appProperties) {
this.appProperties = appProperties;
}
#Override
protected void doHealthCheck(Health.Builder builder) {...}
}
And the properties model as follows:
#Slf4j
#Data
public class AppProperties {
private String baseUrl;
private String name;
}
This means that the configuration would be something like the following (in application.yml):
health-check:
apps:
- baseUrl: http://localhost:10000
name: app1
- baseUrl: http://localhost:10001
name: app2
Finally, you would need to create a bean for each app and register them in the application context:
#Slf4j
#Data
#Configuration
#ConfigurationProperties(prefix = "health-check")
public class AllAppPropertiesConfiguration {
private List<AppProperties> apps;
#Autowired
private GenericApplicationContext applicationContext;
#PostConstruct
fun init() {
for (AppProperties app : apps) {
applicationContext.registerBean(app.getName(), AppHealthCheckContributor.class, app);
}
}
}

How to perform an #Entity query inside a ConstraintValidator

The scenario is that before persisting a Log entity class, its property, String description should be checked if it contains at least a word found in the IllegalWord entity class. Here is the mapping of the two entity classes:
// Log.java
#Entity
public class Log {
#GeneratedValue(strategy=GenerationType.SEQUENCE)
#Id
private Long id;
#NotContainingIllegalWords
private String description;
}
// IllegalWord.java
#Entity
public class IllegalWord {
#GeneratedValue(strategy=GenerationType.SEQUENCE)
#Id
private Long id;
private String word;
}
Since I will be performing a select * to the IllegalWord entity class, I created a repository class for it:
// IllegalWordRepository.java
#Repository
public interface IllegalWordRepository extends CrudRepository<IllegalWord, Long> {}
And then created the ConstraintValidator validator class that will be used by NotContainingIllegalWords annotation, that in turn, will be use to annotate the String description field of Log entity class:
// NotContainingIllegalWordsValidator.java
public class NotContainingIllegalWordsValidator implements ConstraintValidator<NotContainingIllegalWords, Object> {
private static final Logger log = LoggerFactory.getLogger(NotContainingIllegalWordsValidator.class);
#Autowired
private IllegalWordRepository illegalWordRepository;
public void initialize(NotContainingIllegalWords constraintAnnotation) {}
public boolean isValid(String value, ConstraintValidatorContext cxt) {
log.debug("illegalWordRepository is null? " + (illegalWordRepository == null));
// Returns "illegalWordRepository is null? true"
// It is not injected even with the #Autowired annotation.
boolean valid = true;
Collection<IllegalWord> illegalWords = illegalWordRepository.findAll();
// Encounters a NullPointerException here.
// valid = ...loop through illegalWords collection and match regex (or whatever optimal approach)
// with #param value to check if it contains the illegal word.
return valid;
}
I thought it will be as straight-forward like that. But the statement illegalWordRepository.findAll() throws an error because the illegalWordRepository variable is null. Notice that I tried to check if it is null in the preceding statement.
I assumed that I have something wrong coded within the repository class so I attempted to used #Autowired private IllegalWordRepository illegalWordRepository inside a #Service annotated class and suprisingly it is injected there properly (e.i. not null):
// IllegalWordService.java
#Service
public class IllegalWordService {
private static final Logger log = LoggerFactory.getLogger(IllegalWordService.class);
#Autowired
private IllegalWordRepository illegalWordRepository;
public IllegalWord generate(String word) {
log.debug("illegalWordRepository is null? " + (illegalWordRepository == null));
// Returns "illegalWordRepository is null? false"
IllegalWord illegalWord = new IllegalWord();
illegalWord.setWord(word);
illegalWordRepository.save(illegalWord);
// Didn't encounter a NullPointerException here.
return illegalWord;
}
}
Therefore, I guess nothing is wrong with the IllegalWordRepository repository class. It's just that it is not injected in NotContainingIllegalWordsValidator validator class as I intended it to be with the #Autowired annotation (if that is how #Autowired annotation was intended to function even, I am sorry I am new in Spring Framework.).
If there is a proper approach on how to perform a #Entity query inside a ConstraintValidator instance, please tell me.
Related unanswered SO question: Inject Repository inside ConstraintValidator with Spring 4 and message interpolation configuration
Failed Attempt:
I tried to annotate the NotContainingIllegalWordsValidator class with #Configurable annotation, like so:
#Configurable(autowire=Autowire.BY_NAME, preConstruction=true)
public class NotContainingIllegalWordsValidator implements ConstraintValidator<NotContainingIllegalWords, Object> {
but the illegalWordRepository property remains null.
Since Your validator is not initialized by Spring, you can't inject anything into it. You'd have to access the ApplicationContext through a static variable.
#SpringBootApplication
public class MyApplication {
private static ApplicationContext applicationContext;
public static void main(final String[] args) {
applicationContext = SpringApplication.run(MyApplication.class, args);
}
public static ApplicationContext getApplicationContext() {
return applicationContext;
}
}
And in your ConstraintValidator:
public class NotContainingIllegalWordsValidator implements ConstraintValidator<NotContainingIllegalWords, Object> {
public boolean isValid(String value, ConstraintValidatorContext cxt) {
ApplicationContext applicationContext = MyApplication.getApplicationContext();
IllegalWordRepository illegalWordRepository = applicationContext.getBean(IllegalWordRepository.class);
...
}
}
From my answer to a similar question:
The minimum setup for #Autowired to work properly in ConstraintValidator implementation is to have this bean in a Spring #Configuration:
#Bean
public Validator defaultValidator() {
return new LocalValidatorFactoryBean();
}
This is the demo project

Spring + TestNG Autowiring failure - NOT due to "new"

Having an issue with Spring Autowiring. I have an Integration Test class declared as follows:
#ContextConfiguration(classes = TestConfig.class, loader = AnnotationConfigContextLoader.class)
public abstract class BaseIntegrationTest
extends AbstractTestNGSpringContextTests {
#Autowired
protected TestProperties properties;
//... more stuff here
}
The Context configuration looks like this:
#Configuration
#EnableAspectJAutoProxy
#ComponentScan(basePackages = {"com.ourcompany.myapp.it"},
excludeFilters = #ComponentScan.Filter(value = com.inin.wfm.it.config.ConfigPackageExcludeFilter.class, type = FilterType.CUSTOM))
public class TestConfig {
private TestProperties testProperties;
private PropertyService propertyService;
//This both creates and registers the bean with Spring
#Bean
public TestProperties getTestProperties() throws IOException {
if (testProperties == null) {
testProperties = new TestProperties(propertyService());
}
return testProperties;
}
#Bean
public PropertyService propertyService() throws IOException {
if (propertyService == null) {
AppAdminConfig config = new AppAdminConfig.Builder(PropertyService.getEnvironment(), TestConfigKey.ApplicationId)
.checkPropertyHasValue(GlobalConfigKey.KafkaBrokerList.key())
.checkPropertyHasValue(GlobalConfigKey.ZookeeperList.key())
.build();
propertyService = new PropertyService(config.getPropertiesConfig());
propertyService.initialize();
}
return propertyService;
}
}
And this is the bean I'm having trouble with:
#Configurable
public class TestProperties {
private PropertyService propertyService;
public TestProperties(PropertyService propertyService) {
this.propertyService = propertyService;
}
public String getCacheUri(){
return propertyService.getPropertyRegistry().getString(TestConfigKey.CacheUri.key(), Default.CACHE_URI);
}
}
I have multiple Test implementation classes that extend BaseIntegrationTest. All of them but one have valid references to their TestProperties field, but exactly one of the test implementation classes is getting a Null Pointer and throwing an NPE when an attempt is made to reference this.
So the question is, why is Spring #Autowire working fine for 3 different classes that extend the same base class, but wiring up null for the fourth? There is no additional configuration logic in any of the implementations.
For MCVE completeness, here's the bare bones of my impl class
public class CacheIT
extends BaseIntegrationTest {
#Test
public void testUserCache() throws InterruptedException, ExecutionException, TimeoutException {
String uri = properties.getCacheUri() //TODO - NPE here
}
}
I know there's not much to go on... I've been working with Spring for a long time and haven't seen it do this kind of thing before. According to everything I can see it should be working.

Overriding beans in Java-based spring configuration hierarchy

Let's assume we have an application that can be customized for some customers. The application is using Java-based spring configuration (a.k.a. Java config) for dependency injection. The application consists of modules and their submodules. Each module and submodule has its own #Configuration class which is imported by parent configuration using #Import. This creates the following hierarchy:
MainConfig
----------+---------------- ....
| |
ModuleAConfig ModuleBConfig
|--------------------|
| |
SubModuleA1Config SubModuleA2Config
For example ModuleAConfig looks like this:
#Configuration
#Import({SubModuleA1Config.class, SubModuleA2Config.class})
public class ModuleAConfig {
// some module level beans
}
Let's say that SubModuleA1Config defines bean someBean of type SomeBean:
#Configuration
public class SubModuleA1Config {
#Bean
public SomeBean someBean() { return new SomeBean(); }
}
Now I want to customize the application for Customer1 (C1) - I want to use C1SomeBean (extending SomeBean) instead of SomeBean as someBean.
How can I achieve this with minimum duplication?
One of my ideas was to prepare alternative hierarchy with C1Config inheriting from MainConfig, C1ModuleAConfig from ModuleAConfig and C1SubModuleA1Config from SubModuleA1Config. C1SubModuleA1Config would override someBean() method returning C1SomeBean. Unfortunately with Spring 4.0.6 I get something like:
Overriding bean definition for bean 'someBean': replacing [someBean defined in class C1SubmoduleA1Config] with [someBean defined in class SubModuleA1Config]
and indeed SomeBean class is returned from context instead of C1SomeBean. This is clearly not what I want.
Note that you cannot override #Import extending configuration classes.
If you want to select which imports to use at runtime, you could use a #ImportSelector instead.
However, #Configuration classes are not more that spring (scoped) managed factories so as you already have a factory method for someBean you don't need to go even further:
#Configuration
public class SubModuleA1Config {
#Autowired
private Environment env;
#Bean
public SomeBean someBean() {
String customerProperty = env.getProperty("customer");
if ("C1".equals(customerProperty))
return new C1SomeBean();
return new SomeBean();
}
}
Update
Using a ImportSelector:
class CustomerImportSelector implements ImportSelector, EnvironmentAware {
private static final String PACKAGE = "org.example.config";
private static final String CONFIG_CLASS = "SubModuleConfig";
private Environment env;
#Override
public String[] selectImports(AnnotationMetadata importingClassMetadata) {
String customer = env.getProperty("customer");
return new String[] { PACKAGE + "." + customer + "." + CONFIG_CLASS };
}
#Override
public void setEnvironment(Environment environment) {
this.env = environment;
}
}
#Configuration
#Import(CustomerImportSelector.class)
public class ModuleAConfig {
// some module level beans
}
However, as every customer has a a separate package, consider also using #ComponentScan. This will pick the configuration class present and don't need a extra configuration property.
#Configuration
#ComponentScan(basePackages="org.example.customer")
public class SubModuleA1Config {
#Autowired
private CustomerFactory customerFactory;
#Bean
public SomeBean someBean() {
return customerFactory.someBean();
}
}
public interface CustomerFactory {
SomeBean someBean();
}
#Component
public class C1CustomerFactory implements CustomerFactory {
#Override
public SomeBean someBean() {
return new C1SomeBean();
}
}

Categories