When NOT to call super() method when overriding?

sandalone picture sandalone · Apr 20, 2012 · Viewed 45.7k times · Source

When I make my own Android custom class, I extend its native class. Then when I want to override the base method, I always call super() method, just like I always do in onCreate, onStop, etc.

And I thought this is it, as from the very beginning Android team advised us to always call super on every method override.

But, in many books I can see that developers, more experienced than myself, often omit calling super and I really doubt they do it as a lack of knowledge. For example, look at this basic SAX parser class where super is omitted in startElement, characters and endElement:

public class SAXParser extends DefaultHandler{
    public void startElement(String uri, String localName, String qName, Attributes attributes) throws SAXException {
        if(qName.equalsIgnoreCase("XXY")) {
            //do something
        }
    }

    public void characters(char[] ch, int start, int length) throws SAXException {
        //do something
    }

    public void endElement(String uri, String localName, String qName) throws SAXException {
        if(qName.equalsIgnoreCase("XXY")) {
            //do something
        }else () {
            //do something
        }
    }
}

If you try to create any override method via Eclipse or any other IDE, super will always be created as a part of automated process.

This was just a simple example. Books are full of similar code.

How do they know when you must call super and when you can omit it calling?

PS. Do not bind to this specific example. It was just an example randomly picked from many examples.

(This may sound like a beginner question, but I am really confused.)

Answer

Xavi López picture Xavi López · Apr 20, 2012

By calling the super method, you're not overriding the behavior of the method, you're extending it.

A call to super will perform any logic the class you're extending has defined for that method. Take into account that it might be important the moment when you call super's implementation in your method overriding. For instance:

public class A { 
    public void save() { 
         // Perform save logic
    }
}

public class B extends A {
    private Object b;
    @Override
    public void save() { 
        super.save(); // Performs the save logic for A
        save(b); // Perform additional save logic
    }
}

A call to B.save() will perform the save() logic for both A and B, in this particular order. If you weren't calling super.save() inside B.save(), A.save() wouldn't be called. And if you called super.save() after save(b), A.save() would be effectively performed afterwards B.save().

If you want to override super's behavior (that is, fully ignore its implementation and provide it all yourself), you shouldn't be calling super.

In the SAXParser example you provide, the implementations of DefaultHandler for those methods are just empty, so that subclasses can override them and provide a behavior for those methods. In the javadoc for this method this is also pointed out.

public void startElement (String uri, String localName,
    String qName, Attributes attributes) throws SAXException {
    // no op
}

About the super() default call in code generated by IDEs, as @barsju pointed out in his comment, in each constructor there's an implicit call to super() (even if you don't write it in your code), which means, in that context, a call to super's default constructor. The IDE just writes it down for you, but it would also get called if you removed it. Also notice that when implementing constructors, super() or any of its variants with arguments (i.e. super(x,y,z)) can only be called at the very beginning of the method.