Java bean的setter许可是否会返回？

Question

can I define setter method to return this rather than void?

Like:

ClassA setItem1() {
      return this;
}

ClassA setItem2() {
      return this;
}

then I can use new ClassA().setItem1().setItem2()

Answer 1

There is a lot of misunderstanding about JavaBeans spec.

The main reason for it's existence is the unified Java "component" model. It's a way to interact programatically with a Java Object using Reflection. The API itself is named JavaBeans Introspection . Please, take a look at example usages and You will know a lot more than an average Java programmer does.

Introspection API can be used to manipulate GUI elements in an unified manner. Your component exposes it's properties as a pairs of getters and setters so that they could be discovered and manipulated at run-time on a GUI builder's property sheet.

So, mixing fluent APIs and JavaBeans Spec in my opinion is a no-go. That's two completely unrelated concepts and can disrupt each other. JavaBeans Introspection might not work when method signature differs (return type).

Take a look at this example (taken from linked tutorial):

public class SimpleBean
{
private final String name = "SimpleBean";
private int size;

public String getName()
{
    return this.name;
}

public int getSize()
{
    return this.size;
}

public void setSize( int size )
{
    this.size = size;
}

public static void main( String[] args )
        throws IntrospectionException
{
    BeanInfo info = Introspector.getBeanInfo( SimpleBean.class );
    for ( PropertyDescriptor pd : info.getPropertyDescriptors() )
        System.out.println( pd.getName() );
}
}

This example creates a non-visual bean and displays following properties derived from the BeanInfo object:

• class
• name
• size

You might want to see what happens when You change void return type to anything else. I have done so and the result is the same. So, does that mean it's allowed?

I'm afraid no. The JavaBeans spec is quite strict about those method signatures. It just happened that implementation is forgiving. Nonetheless, I'd disadvise mixing fluent interface with JavaBeans. You can't really rely that, if the discovery works now, it will also in future.

But, from the other side - it looks like You don't use JavaBeans to full extent. Only the getters/setters pair of method. It's up to You how You implement and design Your APIs.

Answer 2

The JavaBeans Specification describes a JavaBean as:

A Java Bean is a reusable software component that can be manipulated visually in a builder tool

They are required to provide introspection, customization, events and persistence among other properties (Section 2.1: What is a bean?)

It is common to call a "Java Bean" to a Plain Old Java Object with accessor methods following the JavaBeans Specification (Section 7.1 and 8.3). The truth is that such object could still be far from being compliant with all the requirements.

If the object your are defining in this class is actually a JavaBean then your method must return void according to JavaBean Specification, section 7.1 where accessor methods are described as follows:

void setFoo(PropertyType value); // simple setter
PropertyType getFoo(); // simple getter

The section 8.3 named designed patterns for properties says:

By default, we use design patterns to locate properties by looking for methods of the form:

public <PropertyType> get<PropertyName>();
public void set<PropertyName>(<PropertyType> a);

In addition, for boolean properties, we allow a getter method to match the pattern:

public boolean is<PropertyName>();

However, if your class is just a POJO then there is nothing wrong with using your method chaining strategy because you are allowed to deviate from the specification since you are not actually building a JavaBean. Not all the classes you define are supposed to be JavaBeans after all, right?

Your might like to take a look at the Oracle JavaBeans Tutorial .

Answer 3

No reason you couldn't do that. Personally, if the setters are being used during object creation, I'd name them withItem1() and withItem2().

ClassA obj = new ClassA().withItem1(item1).withItem2(item2);

Makes it a bit clearer (to me anyway) what the intent of the methods are.

Answer 4

After checking the Oracle JavaBean pages I didn't find anything which explicitly tells you that the setters need to be void. Nevertheless all examples have void set-methods.

The PropertyDescriptor in the Java API support non-void setters, so I think it should be pretty safe to let your setters return this. To be on the safe side you should probably check out if the frameworks you intend to use that uses reflection. For instance didn't Spring support non-void setters in xml config prior to version 3.1 .

Answer 5

I would guess this is not in violation of the JavaBean specification, although I am not sure of it.

Check out the following example:

public class JavaBean {

    private String value;

    public String getValue() {
        return value;
    }

    public JavaBean setValue(String value) {
        this.value = value;
        return this;
    }

    public static void main(String[] args) throws Exception {
        JavaBean bean = new JavaBean();
        JavaBean.class.getMethod("setValue", String.class).invoke(bean, "test");
        System.out.println(bean.getValue());
    }
}

Many frameworks access JavaBeans using the reflection API. As you can see above, accessing a settter which returns 'this' is not influenced by the return type (the return type is not used to locate a method via reflection). It also makes sense, because you cannot have two methods in one scope that are identical except for their return type.

Answer 6

Yes. This is a somewhat common technique called Method Chaining, and can be used to create a "fluent interface".

See: http://en.wikipedia.org/wiki/Method_chaining , http://en.wikipedia.org/wiki/Fluent_interface

Answer 7

Absolutely nothing to stop you doing that but why. If you want to do this create a constructor that takes the args. Bare in mind some software that uses beans would not be expecting return values and may have some unexpected results

If you just want to simplify initialisation, (maybe to set up tests) you could use some groovy code.

Answer 8

Just to add that for people using Spring 3.1+ this is not an issue anymore

see http://static.springsource.org/spring/docs/3.1.0.M2/spring-framework-reference/html/new-in-3.1.html

Answer 9

The Builder pattern is generally used for constructing immutable objects. Even though JavaBeans by nature aren't immutable, I have frequently used the builder pattern on my JavaBeans because it provides a fluent interface that I can use in my tests. The two are easily compatible with each other without breaking the JavaBean specification. You can check out it out on Stack Overflow at Builder Pattern in Effective Java

You just need to make sure you include a default constructor as well as the private Builder constructor, and then put your standard getters and setters in the JavaBean object.

This is much cleaner than constructor chaining, and easier to read as well.

Answer 10

There is nothing preventing you from providing setter methods which return the target object as a convention in your interface...

However , you must also use the canonical signature for Java Bean Simple Property setter methods (eg void setProp(Type t) ) or the bean property will not be recognized as writeable by other software which expects that signature.