I just announced the new Spring 5 modules in REST With Spring:


1. Overview

CDI (Contexts and Dependency Injection) is a standard dependency injection framework included in Java EE 6 and higher.

It allows us to manage the lifecycle of stateful components via domain-specific lifecycle contexts and inject components (services) into client objects in a type-safe way.

In this tutorial, we’ll take an in-depth look at CDI’s most relevant features and implement different approaches for injecting dependencies in client classes.

2. DYDI (Do-it-Yourself Dependency Injection)

In a nutshell, it’s possible to implement DI without resorting to any framework at all.

This approach is popularly known as DYDI (Do-it-Yourself Dependency Injection).

With DYDI, we keep application code isolated from object creation by passing the required dependencies into the client classes through plain old factories/builders.

Here’s how a basic DYDI implementation might look like:

public interface TextService {
    String doSomethingWithText(String text);
    String doSomethingElseWithText(String text);    
public class SpecializedTextService implements TextService { ... }
public class TextClass {
    private TextService textService;
    // constructor
public class TextClassFactory {
    public TextClass getTextClass() {
        return new TextClass(new SpecializedTextService(); 

Of course, DYDI is suitable for some relatively simple use cases.

If our sample application grew in size and complexity, implementing a larger network of interconnected objects, we would end up polluting it with tons of object graph factories.

This would require a lot of boilerplate code just for creating object graphs. This is not a fully-scalable solution.

Can we do DI any better? Of course, we can. Here’s exactly where CDI comes into the picture.

3. A Simple Example

CDI turns DI into a no-brainer process, boiled down to just decorating the service classes with a few simple annotations, and defining the corresponding injection points in the client classes.

To showcase how CDI implements DI at the most basic level, let’s suppose that we want to develop a simple image file editing application. Capable of opening, editing, writing, saving an image file and so on.

3.1. The “beans.xml” File

First, we must place a “beans.xml” file in the “src/main/resources/META-INF/” folder. Even if this file doesn’t contain any specific DI directives at all, it’s required for getting CDI up and running:

<beans xmlns="http://java.sun.com/xml/ns/javaee" 

3.2. The Service Classes

Next, let’s create the service classes that perform the file mentioned above operations on GIF, JPG and PNG files:

public interface ImageFileEditor {
    String openFile(String fileName);
    String editFile(String fileName);
    String writeFile(String fileName);
    String saveFile(String fileName);
public class GifFileEditor implements ImageFileEditor {
    public String openFile(String fileName) {
        return "Opening GIF file " + fileName;
    public String editFile(String fileName) {
      return "Editing GIF file " + fileName;
    public String writeFile(String fileName) {
        return "Writing GIF file " + fileName;

    public String saveFile(String fileName) {
        return "Saving GIF file " + fileName;
public class JpgFileEditor implements ImageFileEditor {
    // JPG-specific implementations for openFile() / editFile() / writeFile() / saveFile()
public class PngFileEditor implements ImageFileEditor {
    // PNG-specific implementations for openFile() / editFile() / writeFile() / saveFile()

3.3. The Client Class

Finally, let’s implement a client class that takes an ImageFileEditor implementation in the constructor, and let’s define an injection point with the @Inject annotation:

public class ImageFileProcessor {
    private ImageFileEditor imageFileEditor;
    public ImageFileProcessor(ImageFileEditor imageFileEditor) {
        this.imageFileEditor = imageFileEditor;

Simply put, the @Inject annotation is CDI’s actual workhorse. It allows us to define injection points in the client classes.

In this case, @Inject instructs CDI to inject an ImageFileEditor implementation in the constructor.

Furthermore, it’s also possible to inject a service by using the @Inject annotation in fields (field injection) and setters (setter injection). We’ll look at these options later.

3.4. Building the ImageFileProcessor Object Graph with Weld

Of course, we need to make sure that CDI will inject the right ImageFileEditor implementation into the ImageFileProcessor class constructor.

To do so, first, we should get an instance of the class.

As we won’t rely on any Java EE application server for using CDI, we’ll do this with Weld, the CDI reference implementation in Java SE:

public static void main(String[] args) {
    Weld weld = new Weld();
    WeldContainer container = weld.initialize();
    ImageFileProcessor imageFileProcessor = container.select(ImageFileProcessor.class).get();

Here, we’re creating a WeldContainer object, then getting an ImageFileProcessor object, and finally calling its openFile() method.

As expected, if we run the application, CDI will complain loudly by throwing a DeploymentException:

Unsatisfied dependencies for type ImageFileEditor with qualifiers @Default at injection point...

We’re getting this exception because CDI doesn’t know what ImageFileEditor implementation to inject into the ImageFileProcessor constructor.

In CDI’s terminology, this is known as an ambiguous injection exception.

3.5. The @Default and @Alternative Annotations

Solving this ambiguity is easy. CDI, by default, annotates all the implementations of an interface with the @Default annotation.

So, we should explicitly tell it which implementation should be injected into the client class:

public class GifFileEditor implements ImageFileEditor { ... }
public class JpgFileEditor implements ImageFileEditor { ... }
public class PngFileEditor implements ImageFileEditor { ... }

In this case, we’ve annotated GifFileEditor and JpgFileEditor with the @Alternative annotation, so CDI now knows that PngFileEditor (annotated by default with the @Default annotation) is the implementation that we want to inject.

If we rerun the application, this time it’ll be executed as expected:

Opening PNG file file1.png

Furthermore, annotating PngFileEditor with the @Default annotation and keeping the other implementations as alternatives will produce the same above result.

This shows, in a nutshell, how we can very easily swap the run-time injection of implementations by simply switching the @Alternative annotations in the service classes.

4. Field Injection

CDI supports both field and setter injection out of the box.

Here’s how to perform field injection (the rules for qualifying services with the @Default and @Alternative annotations remain the same):

private final ImageFileEditor imageFileEditor;

5. Setter Injection

Similarly, here’s how to do setter injection:

public void setImageFileEditor(ImageFileEditor imageFileEditor) { ... }

6. The @Named Annotation

So far, we’ve learned how to define injection points in client classes and inject services with the @Inject, @Default , and @Alternative annotations, which cover most of the use cases.

Nevertheless, CDI also allows us to perform service injection with the @Named annotation.

This method provides a more semantic way of injecting services, by binding a meaningful name to an implementation:

public class GifFileEditor implements ImageFileEditor { ... }

public class JpgFileEditor implements ImageFileEditor { ... }

public class PngFileEditor implements ImageFileEditor { ... }

Now, we should refactor the injection point in the ImageFileProcessor class to match a named implementation:

public ImageFileProcessor(@Named("PngFileEditor") ImageFileEditor imageFileEditor) { ... }

It’s also possible to perform field and setter injection with named implementations, which looks very similar to using the @Default and @Alternative annotations:

private final @Named("PngFileEditor") ImageFileEditor imageFileEditor;

public void setImageFileEditor(@Named("PngFileEditor") ImageFileEditor imageFileEditor) { ... }

7. The @Produces Annotation

Sometimes, a service requires some configuration to be fully-initialized before it gets injected to handle additional dependencies.

CDI provides support for these situations, through the @Produces annotation.

@Produces allows us to implement factory classes, whose responsibility is the creation of fully-initialized services.

To understand how the @Produces annotation works, let’s refactor the ImageFileProcessor class, so it can take an additional TimeLogger service in the constructor.

The service will be used for logging the time at which a certain image file operation is performed:

public ImageFileProcessor(ImageFileEditor imageFileEditor, TimeLogger timeLogger) { ... } 
public String openFile(String fileName) {
    return imageFileEditor.openFile(fileName) + " at: " + timeLogger.getTime();
// additional image file methods

In this case, the TimeLogger class takes two additional services, SimpleDateFormat and Calendar:

public class TimeLogger {
    private SimpleDateFormat dateFormat;
    private Calendar calendar;
    // constructors
    public String getTime() {
        return dateFormat.format(calendar.getTime());

How do we tell CDI where to look at for getting a fully-initialized TimeLogger object?

We just create a TimeLogger factory class and annotate its factory method with the @Produces annotation:

public class TimeLoggerFactory {
    public TimeLogger getTimeLogger() {
        return new TimeLogger(new SimpleDateFormat("HH:mm"), Calendar.getInstance());

Whenever we get an ImageFileProcessor instance, CDI will scan the TimeLoggerFactory class, then call the getTimeLogger() method (as it’s annotated with the @Produces annotation), and finally inject the Time Logger service.

If we run the refactored sample application with Weld, it’ll output the following:

Opening PNG file file1.png at: 17:46

8. Custom Qualifiers

CDI supports the use of custom qualifiers for qualifying dependencies and solving ambiguous injection points.

Custom qualifiers are a very powerful feature. They not only bind a semantic name to a service, but they bind injection metadata too. Metadata such as the RetentionPolicy and the legal annotation targets (ElementType).

Let’s see how to use custom qualifiers in our application:

@Target({ElementType.FIELD, ElementType.METHOD, ElementType.TYPE, ElementType.PARAMETER})
public @interface GifFileEditorQualifier {}
@Target({ElementType.FIELD, ElementType.METHOD, ElementType.TYPE, ElementType.PARAMETER})
public @interface JpgFileEditorQualifier {}
@Target({ElementType.FIELD, ElementType.METHOD, ElementType.TYPE, ElementType.PARAMETER})
public @interface PngFileEditorQualifier {}

Now, let’s bind the custom qualifiers to the ImageFileEditor implementations:

public class GifFileEditor implements ImageFileEditor { ... }
public class JpgFileEditor implements ImageFileEditor { ... }
public class PngFileEditor implements ImageFileEditor { ... }

Lastly, let’s refactor the injection point in the ImageFileProcessor class:

public ImageFileProcessor(@PngFileEditorQualifier ImageFileEditor imageFileEditor, TimeLogger timeLogger) { ... }

If we run our application once again, it should generate the same output shown above.

Custom qualifiers provide a neat semantic approach for binding names and annotation metadata to implementations.

In addition, custom qualifiers allow us to define more restrictive type-safe injection points (outperforming the functionality of the @Default and @Alternative annotations).

If only a subtype is qualified in a type hierarchy, then CDI will only inject the subtype, not the base type.

9. Conclusion

Unquestionably, CDI makes dependency injection a no-brainer, the cost of the extra annotations is very little effort for the gain of organized dependency injection.

There are times when DYDI does still have its place over CDI. Like when developing fairly simple applications that only contain simple object graphs.

As always, all the code samples shown in this article are available over on GitHub.

I just announced the new Spring 5 modules in REST With Spring:


Leave a Reply

Notify of