Quarkus - CDI Integration Guide

ArC, the CDI container, is bootstrapped at build time. The downside of this approach is that CDI Portable Extensions cannot be supported. Nevertheless, the functionality can be achieved using the Quarkus-specific extensions API.

The container is bootstrapped in multiple phases. From a high level perspective these phases go as follows:

  1. Initialization

  2. Bean discovery

  3. Registration of synthetic components

  4. Validation

In the initialization phase the preparatory work is being carried out and custom contexts are registered. Bean discovery is then the process where the container analyzes all application classes, identifies beans and wires them all together based on the provided metadata. Subsequently, the extensions can register synthetic components. Attributes of these components are fully controlled by the extensions, i.e. are not derived from an existing class. Finally, the deployment is validated. For example, the container validates every injection point in the application and fails the build if there is no bean that satisfies the given required type and qualifiers.

You can see more information about the bootstrap by enabling additional logging. Simply run the Maven build with -X or --debug and grep the lines that contain io.quarkus.arc. In the development mode, you can use quarkus.log.category."io.quarkus.arc.processor".level=DEBUG and two special endpoints are also registered automatically to provide some basic debug info in the JSON format.

Quarkus build steps can produce and consume various build items and hook into each phase. In the following sections we will describe all the relevant build items and common scenarios.

1. Metadata Sources

Classes and annotations are the primary source of bean-level metadata. The initial metadata are read from the bean archive index, an immutable Jandex index which is built from various sources during bean discovery. However, extensions can add, remove or transform the metadata at certain stages of the bootstrap. Moreover, extensions can also register synthetic components. This is an important aspect to realize when integrating CDI components in Quarkus.

This way, extensions can turn classes, that would be otherwise ignored, into beans and vice versa. For example, a class that declares a @Scheduled method is always registered as a bean even if it is not annotated with a bean defining annotation and would be normally ignored.

2. Use Case - My Class Is Not Recognized as a Bean

An UnsatisfiedResolutionException indicates a problem during typesafe resolution. Sometimes an injection point cannot be satisfied even if there is a class on the classpath that appears to be eligible for injection. There are several reasons why a class is not recognized and also several ways to fix it. In the first step we should identify the reason.

2.1. Reason 1: Class Is Not discovered

Quarkus has a simplified discovery. It might happen that the class is not part of the application index. For example, classes from the runtime module of a Quarkus extension are not indexed automatically.

Solution: Use the AdditionalBeanBuildItem. This build item can be used to specify one or more additional classes to be analyzed during the discovery. Additional bean classes are transparently added to the application index processed by the container.

AdditionalBeanBuildItem Example
@BuildStep
AdditionalBeanBuildItem additionalBeans() {
     return new AdditionalBeanBuildItem(SmallRyeHealthReporter.class, HealthServlet.class)); (1)
}
1 AdditionalBeanBuildItem.Builder can be used for more complex use cases.

Bean classes added via AdditionalBeanBuildItem are removable by default. If the container considers them unused, they are just ignored. However, you can use AdditionalBeanBuildItem.Builder.setUnremovable() method to instruct the container to never remove bean classes registered via this build item. See also Removing Unused Beans and Reason 3: Class Was Discovered and Has a Bean Defining Annotation but Was Removed for more details.

It is aso possible to set the default scope via AdditionalBeanBuildItem.Builder#setDefaultScope(). The default scope is only used if there is no scope declared on the bean class.

If no default scope is specified the @Dependent pseudo-scope is used.

2.2. Reason 2: Class Is Discovered but Has No Bean Defining Annotation

In Quarkus, the application is represented by a single bean archive with the bean discovery mode annotated. Therefore, bean classes that don’t have a bean defining annotation are ignored. Bean defining annotations are declared on the class-level and include scopes, stereotypes and @Interceptor.

Solution 1: Use the AutoAddScopeBuildItem. This build item can be used to add a scope to a class that meets certain conditions.

AutoAddScopeBuildItem Example
@BuildStep
AutoAddScopeBuildItem autoAddScope() {
   return AutoAddScopeBuildItem.builder().containsAnnotations(SCHEDULED_NAME, SCHEDULES_NAME) (1)
      .defaultScope(BuiltinScope.SINGLETON) (2)
      .build();
}
1 Find all classes annotated with @Scheduled.
2 Add @Singleton as default scope. Classes already annotated with a scope are skipped automatically.

Solution 2: If you need to process classes annotated with a specific annotation then it’s possible to extend the set of bean defining annotations via the BeanDefiningAnnotationBuildItem.

BeanDefiningAnnotationBuildItem Example
@BuildStep
BeanDefiningAnnotationBuildItem additionalBeanDefiningAnnotation() {
   return new BeanDefiningAnnotationBuildItem(Annotations.GRAPHQL_API); (1)
}
1 Add org.eclipse.microprofile.graphql.GraphQLApi to the set of bean defining annotations.

Bean classes added via BeanDefiningAnnotationBuildItem are not removable by default, i.e. the resulting beans must not be removed even if they are considered unused. However, you can change the default behavior. See also Removing Unused Beans and Reason 3: Class Was Discovered and Has a Bean Defining Annotation but Was Removed for more details.

It is also possible to specify the default scope. The default scope is only used if there is no scope declared on the bean class.

If no default scope is specified the @Dependent pseudo-scope is used.

2.3. Reason 3: Class Was Discovered and Has a Bean Defining Annotation but Was Removed

The container attempts to remove all unused beans during the build by default. This optimization allows for framework-level dead code elimination. In few special cases, it’s not possible to correctly identify an unused bean. In particular, Quarkus is not able to detect the usage of the CDI.current() static method yet. Extensions can eliminate possible false positives by producing an UnremovableBeanBuildItem.

UnremovableBeanBuildItem Example
@BuildStep
UnremovableBeanBuildItem unremovableBeans() {
   return UnremovableBeanBuildItem.targetWithAnnotation(STARTUP_NAME); (1)
}
1 Make all classes annotated with @Startup unremovable.

3. Use Case - My Annotation Is Not Recognized as a Qualifier or an Interceptor Binding

It is likely that the annotation class is not part of the application index. For example, classes from the runtime module of a Quarkus extension are not indexed automatically.

Solution: Use the AdditionalBeanBuildItem as described in Reason 1: Class Is Not discovered.

4. Use Case - I Need To Transform Metadata

In some cases, it’s useful to be able to modify the metadata. Quarkus provides a powerful alternative to javax.enterprise.inject.spi.ProcessAnnotatedType. With an AnnotationsTransformerBuildItem it’s possible to override the annotations that exist on bean classes.

For example, you might want to add an interceptor binding to a specific bean class. Here is how to do it:

AnnotationsTransformerBuildItem Example
@BuildStep
AnnotationsTransformerBuildItem transform() {
   return new AnnotationsTransformerBuildItem(new AnnotationsTransformer() {

      public boolean appliesTo(org.jboss.jandex.AnnotationTarget.Kind kind) {
         return kind == org.jboss.jandex.AnnotationTarget.Kind.CLASS; (1)
      }

      public void transform(TransformationContext context) {
         if (context.getTarget().asClass().name().toString().equals("org.acme.Bar")) {
            context.transform().add(MyInterceptorBinding.class).done(); (2)
         }
      }
    });
}
1 The transformer is only applied to classes.
2 If the class name equals to org.acme.Bar then add @MyInterceptorBinding. Don’t forget to invoke Transformation#done().
Keep in mind that annotation transformers must be produced before the bean discovery starts.

Build steps can query the transformed annotations for a given annotation target via the TransformedAnnotationsBuildItem.

TransformedAnnotationsBuildItem Example
@BuildStep
void queryAnnotations(TransformedAnnotationsBuildItem transformedAnnotations, BuildProducer<MyBuildItem> myBuildItem) {
   ClassInfo myClazz = ...;
   if (transformedAnnotations.getAnnotations(myClazz).isEmpty()) { (1)
     myBuildItem.produce(new MyBuildItem());
   }
}
1 TransformedAnnotationsBuildItem.getAnnotations() will return a possibly transformed set of annotations.
There are other build items specialized in transformation: Use Case - Additional Interceptor Bindings and Use Case - Injection Point Transformation.

5. Use Case - Inspect Beans, Observers and Injection Points

5.1. Solution 1: BeanDiscoveryFinishedBuildItem

Consumers of BeanDiscoveryFinishedBuildItem can easily inspect all class-based beans, observers and injection points registered in the application. However, synthetic beans and observers are not included because this build item is produced before the synthetic components are registered.

Additionaly, the bean resolver returned from BeanDiscoveryFinishedBuildItem#getBeanResolver() can be used to apply the type-safe resolution rules, e.g. to find out whether there is a bean that would satisfy certain combination of required type and qualifiers.

BeanDiscoveryFinishedBuildItem Example
@BuildStep
void doSomethingWithNamedBeans(BeanDiscoveryFinishedBuildItem beanDiscovery, BuildProducer<NamedBeansBuildItem> namedBeans) {
   List<BeanInfo> namedBeans = beanDiscovery.beanStream().withName().collect(toList())); (1)
   namedBeans.produce(new NamedBeansBuildItem(namedBeans));
}
1 The resulting list will not contain @Named synthetic beans.

5.2. Solution 2: SynthesisFinishedBuildItem

Consumers of SynthesisFinishedBuildItem can easily inspect all beans, observers and injection points registered in the application. Synthetic beans and observers are included because this build item is produced after the synthetic components are registered.

Additionaly, the bean resolver returned from SynthesisFinishedBuildItem#getBeanResolver() can be used to apply the type-safe resolution rules, e.g. to find out whether there is a bean that would satisfy certain combination of required type and qualifiers.

SynthesisFinishedBuildItem Example
@BuildStep
void doSomethingWithNamedBeans(SynthesisFinishedBuildItem synthesisFinished, BuildProducer<NamedBeansBuildItem> namedBeans) {
   List<BeanInfo> namedBeans = synthesisFinished.beanStream().withName().collect(toList())); (1)
   namedBeans.produce(new NamedBeansBuildItem(namedBeans));
}
1 The resulting list will contain @Named synthetic beans.

6. Use Case - The Need for Synthetic Beans

Sometimes it is practical to be able to register a synthetic bean. Bean attributes of a synthetic bean are not derived from a Java class, method or field. Instead, all the attributes are defined by an extension. In regular CDI, this could be achieved using the AfterBeanDiscovery.addBean() methods.

Solution: If you need to register a synthetic bean then use the SyntheticBeanBuildItem.

SyntheticBeanBuildItem Example 1
@BuildStep
SyntheticBeanBuildItem syntheticBean() {
   return SyntheticBeanBuildItem.configure(String.class)
             .qualifiers(new MyQualifierLiteral())
             .creator(mc -> mc.returnValue(mc.load("foo"))) (1)
             .done();
}
1 Generate the bytecode of the javax.enterprise.context.spi.Contextual#create(CreationalContext<T>) implementation.

The output of a bean configurator is recorded as bytecode. Therefore, there are some limitations in how a synthetic bean instance is created at runtime. You can:

  1. Generate the bytecode of the Contextual#create(CreationalContext<T>) method directly via ExtendedBeanConfigurator.creator(Consumer<MethodCreator>).

  2. Pass a io.quarkus.arc.BeanCreator implementation class via ExtendedBeanConfigurator#creator(Class<? extends BeanCreator<U>>), and possibly specify some parameters via ExtendedBeanConfigurator#param().

  3. Produce the runtime instance through a proxy returned from a @Recorder method and set it via ExtendedBeanConfigurator#runtimeValue(RuntimeValue<?>) or ExtendedBeanConfigurator#supplier(Supplier<?>).

SyntheticBeanBuildItem Example 2
@BuildStep
@Record(STATIC_INIT) (1)
SyntheticBeanBuildItem syntheticBean(TestRecorder recorder) {
   return SyntheticBeanBuildItem.configure(Foo.class).scope(Singleton.class)
                .runtimeValue(recorder.createFoo()) (2)
                .done();
}
1 By default, a synthetic bean is initialized during STATIC_INIT.
2 The bean instance is supplied by a value returned from a recorder method.

It is possible to mark a synthetic bean to be initialized during RUNTIME_INIT. See the Three Phases of Bootstrap and Quarkus Philosophy for more information about the difference between STATIC_INIT and RUNTIME_INIT.

RUNTIME_INIT SyntheticBeanBuildItem Example
@BuildStep
@Record(RUNTIME_INIT) (1)
SyntheticBeanBuildItem syntheticBean(TestRecorder recorder) {
   return SyntheticBeanBuildItem.configure(Foo.class).scope(Singleton.class)
                .setRuntimeInit() (2)
                .runtimeValue(recorder.createFoo())
                .done();
}
1 The recorder must be executed in the ExecutionTime.RUNTIME_INIT phase.
2 The bean instance is initialized during RUNTIME_INIT.

Synthetic beans initialized during RUNTIME_INIT must not be accessed during STATIC_INIT. RUNTIME_INIT build steps that access a runtime-init synthetic bean should consume the SyntheticBeansRuntimeInitBuildItem:

@BuildStep
@Record(RUNTIME_INIT)
@Consume(SyntheticBeansRuntimeInitBuildItem.class) (1)
void accessFoo(TestRecorder recorder) {
   recorder.foo(); (2)
}
1 This build step must be executed after syntheticBean() completes.
2 This recorder method results in an invocation upon the Foo bean instance and thus we need to make sure that the build step is executed after all synthetic beans are initialized.
It is also possible to use the BeanRegistrationPhaseBuildItem to register a synthetic bean. However, we recommend extension authors to stick with SyntheticBeanBuildItem which is more idiomatic for Quarkus.

7. Use Case - Synthetic Observers

Similar to synthetic beans, the attributes of a synthetic observer method are not derived from a Java method. Instead, all the attributes are defined by an extension.

Solution: If you need to register a synthetic observer, use the ObserverRegistrationPhaseBuildItem.

A build step that consumes the ObserverRegistrationPhaseBuildItem should always produce an ObserverConfiguratorBuildItem or at least inject a BuildProducer for this build item, otherwise it could be ignored or processed at the wrong time (e.g. after the correct CDI bootstrap phase).
ObserverRegistrationPhaseBuildItem Example
@BuildStep
void syntheticObserver(ObserverRegistrationPhaseBuildItem observerRegistrationPhase,
            BuildProducer<MyBuildItem> myBuildItem,
            BuildProducer<ObserverConfiguratorBuildItem> observerConfigurators) {
   observerConfigurators.produce(new ObserverConfiguratorBuildItem(observerRegistrationPhase.getContext()
       .configure()
       .beanClass(DotName.createSimple(MyBuildStep.class.getName()))
       .observedType(String.class)
       .notify(mc -> {
           // do some gizmo bytecode generation...
       })));
   myBuildItem.produce(new MyBuildItem());
}

The output of a ObserverConfigurator is recorded as bytecode. Therefore, there are some limitations in how a synthetic observer is invoked at runtime. Currently, you must generate the bytecode of the method body directly.

8. Use Case - I Have a Generated Bean Class

No problem. You can generate the bytecode of a bean class manually and then all you need to do is to produce a GeneratedBeanBuildItem instead of GeneratedClassBuildItem.

GeneratedBeanBuildItem Example
@BuildStep
void generatedBean(BuildProducer<GeneratedBeanBuildItem> generatedBeans) {
    ClassOutput beansClassOutput = new GeneratedBeanGizmoAdaptor(generatedBeans); (1)
    ClassCreator beanClassCreator = ClassCreator.builder().classOutput(beansClassOutput)
                .className("org.acme.MyBean")
                .build();
    beanClassCreator.addAnnotation(Singleton.class);
    beanClassCreator.close(); (2)
}
1 io.quarkus.arc.deployment.GeneratedBeanGizmoAdaptor makes it easy to produce GeneratedBeanBuildItems from Gizmo constructs.
2 The resulting bean class is something like public class @Singleton MyBean { }.

9. Use Case - I Need to Validate the Deployment

Sometimes extensions need to inspect the beans, observers and injection points, then perform additional validations and fail the build if something is wrong.

Solution: If an extension needs to validate the deployment it should use the ValidationPhaseBuildItem.

A build step that consumes the ValidationPhaseBuildItem should always produce a ValidationErrorBuildItem or at least inject a BuildProducer for this build item, otherwise it could be ignored or processed at the wrong time (e.g. after the correct CDI bootstrap phase).
@BuildStep
void validate(ValidationPhaseBuildItem validationPhase,
            BuildProducer<MyBuildItem> myBuildItem,
            BuildProducer<ValidationErrorBuildItem> errors) {
   if (someCondition) {
     errors.produce(new ValidationErrorBuildItem(new IllegalStateException()));
     myBuildItem.produce(new MyBuildItem());
   }
}
You can easily filter all registered beans via the convenient BeanStream returned from the ValidationPhaseBuildItem.getContext().beans() method.

10. Use Case - Register a Custom CDI Context

Sometimes extensions need to extend the set of built-in CDI contexts.

Solution: If you need to register a custom context, use the ContextRegistrationPhaseBuildItem.

A build step that consumes the ContextRegistrationPhaseBuildItem should always produce a ContextConfiguratorBuildItem or at least inject a BuildProducer for this build item, otherwise it could be ignored or processed at the wrong time (e.g. after the correct CDI bootstrap phase).

ContextRegistrationPhaseBuildItem Example

@BuildStep
ContextConfiguratorBuildItem registerContext(ContextRegistrationPhaseBuildItem phase) {
      return new ContextConfiguratorBuildItem(phase.getContext().configure(TransactionScoped.class).normal().contextClass(TransactionContext.class));
}

Additionally, each extension that registers a custom CDI context via ContextRegistrationPhaseBuildItem should also produce the CustomScopeBuildItem in order to contribute the custom scope annotation name to the set of bean defining annotations.

CustomScopeBuildItem Example

@BuildStep
CustomScopeBuildItem customScope() {
   return new CustomScopeBuildItem(DotName.createSimple(TransactionScoped.class.getName()));
}

10.1. What if I Need to Know All the Scopes Used in the Application?

Solution: You can inject the CustomScopeAnnotationsBuildItem in a build step and use the convenient methods such as CustomScopeAnnotationsBuildItem.isScopeDeclaredOn().

11. Use Case - Additional Interceptor Bindings

In rare cases it might be handy to programmatically register an existing annotation that is not annotated with @javax.interceptor.InterceptorBinding as an interceptor binding. This is similar to what CDI achieves through BeforeBeanDiscovery#addInterceptorBinding(). We are going to use InterceptorBindingRegistrarBuildItem to get it done.

InterceptorBindingRegistrarBuildItem Example
@BuildStep
InterceptorBindingRegistrarBuildItem addInterceptorBindings() {
    return new InterceptorBindingRegistrarBuildItem(new InterceptorBindingRegistrar() {
        @Override
        public Map<DotName, Set<String>> registerAdditionalBindings() {
            return Collections.singletonMap(DotName.createSimple(NotAnInterceptorBinding.class.getName()),
                                        Collections.emptySet());
        }
    });
}

12. Use Case - Additional Qualifiers

Sometimes it might be useful to register an existing annotation that is not annotated with @javax.inject.Qualifier as a CDI qualifier. This is similar to what CDI achieves through BeforeBeanDiscovery#addQualifier(). We are going to use QualifierRegistrarBuildItem to get it done.

QualifierRegistrarBuildItem Example
@BuildStep
QualifierRegistrarBuildItem addQualifiers() {
    return new QualifierRegistrarBuildItem(new QualifierRegistrar() {
        @Override
        public Map<DotName, Set<String>> getAdditionalQualifiers() {
            return Collections.singletonMap(DotName.createSimple(NotAQualifier.class.getName()),
                                        Collections.emptySet());
        }
    });
}

13. Use Case - Injection Point Transformation

Every now and then it is handy to be able to change the qualifiers of an injection point programmatically. You can do just that with InjectionPointTransformerBuildItem. The following sample shows how to apply transformation to injection points with type Foo that contain qualifier MyQualifier:

InjectionPointTransformerBuildItem Example
@BuildStep
InjectionPointTransformerBuildItem transformer() {
    return new InjectionPointTransformerBuildItem(new InjectionPointsTransformer() {

        public boolean appliesTo(Type requiredType) {
            return requiredType.name().equals(DotName.createSimple(Foo.class.getName()));
        }

        public void transform(TransformationContext context) {
            if (context.getQualifiers().stream()
                    .anyMatch(a -> a.name().equals(DotName.createSimple(MyQualifier.class.getName())))) {
                context.transform()
                        .removeAll()
                        .add(DotName.createSimple(MyOtherQualifier.class.getName()))
                        .done();
            }
        }
    });
}
In theory, you can use an AnnotationsTransformer to achieve the same goal. However, there are few differences that make InjectionPointsTransformer more suitable for this particular task: (1) annotation transformers are applied to all classes during bean discovery, whereas InjectionPointsTransformer is only applied to discovered injection points after bean discovery; (2) with InjectionPointsTransformer you don’t need to handle various types of injection points (field, parameters of initializer methods, etc.).

14. Use Case - Resource Annotations and Injection

The ResourceAnnotationBuildItem can be used to specify resource annotations that make it possible to resolve non-CDI injection points, such as Jakarta EE resources. An integrator must also provide a corresponding io.quarkus.arc.ResourceReferenceProvider service provider implementation.

ResourceAnnotationBuildItem Example
@BuildStep
void setupResourceInjection(BuildProducer<ResourceAnnotationBuildItem> resourceAnnotations, BuildProducer<GeneratedResourceBuildItem> resources) {
    resources.produce(new GeneratedResourceBuildItem("META-INF/services/io.quarkus.arc.ResourceReferenceProvider",
        MyResourceReferenceProvider.class.getName().getBytes()));
    resourceAnnotations.produce(new ResourceAnnotationBuildItem(DotName.createSimple(MyAnnotation.class.getName())));
}

15. Available Build Time Metadata

Any of the above extensions that operates with BuildExtension.BuildContext can leverage certain build time metadata that are generated during build. The built-in keys located in io.quarkus.arc.processor.BuildExtension.Key are:

ANNOTATION_STORE

Contains an AnnotationStore that keeps information about all AnnotationTarget annotations after application of annotation transformers

INJECTION_POINTS

Collection<InjectionPointInfo> containing all injection points

BEANS

Collection<BeanInfo> containing all beans

REMOVED_BEANS

Collection<BeanInfo> containing all the removed beans; see Removing unused beans for more information

OBSERVERS

Collection<ObserverInfo> containing all observers

SCOPES

Collection<ScopeInfo> containing all scopes, including custom ones

QUALIFIERS

Map<DotName, ClassInfo> containing all qualifiers

INTERCEPTOR_BINDINGS

Map<DotName, ClassInfo> containing all interceptor bindings

STEREOTYPES

Map<DotName, StereotypeInfo> containing all stereotypes

To get hold of these, simply query the extension context object for given key. Note that these metadata are made available as build proceeds which means that extensions can only leverage metadata that were built before the extensions are invoked. If your extension attempts to retrieve metadata that wasn’t yet produced, null will be returned. Here is a summary of which extensions can access which metadata:

AnnotationsTransformer

Shouldn’t rely on any metadata as it could be used at any time in any phase of the bootstrap

ContextRegistrar

Has access to ANNOTATION_STORE, QUALIFIERS, INTERCEPTOR_BINDINGS, STEREOTYPES

InjectionPointsTransformer

Has access to ANNOTATION_STORE, QUALIFIERS, INTERCEPTOR_BINDINGS, STEREOTYPES

ObserverTransformer

Has access to ANNOTATION_STORE, QUALIFIERS, INTERCEPTOR_BINDINGS, STEREOTYPES

BeanRegistrar

Has access to ANNOTATION_STORE, QUALIFIERS, INTERCEPTOR_BINDINGS, STEREOTYPES, BEANS (class-based beans only), OBSERVERS (class-based observers only), INJECTION_POINTS

ObserverRegistrar

Has access to ANNOTATION_STORE, QUALIFIERS, INTERCEPTOR_BINDINGS, STEREOTYPES, BEANS, OBSERVERS (class-based observers only), INJECTION_POINTS

BeanDeploymentValidator

Has access to all build metadata