Question

I have this parent abstract class which defines an Apache logger static object. Something like this:

import org.apache.log4j.Logger;

public abstract class A {

    private final static Logger logger;

        (...)

}

I know this code is illegal because the logger object is not initialized. The problem is I don’t want to initialize it with logger = Logger.getLogger(A.class); because I want each child class to initialize it with its own class object, that way I will know which class caused which errors.

But at the same time I want to include some of my logging methods on the base class A.

What would be the best practice for this? Should I initialize it with A.class, then reinstantiate it for each child class? Somehow that feels incorrect to me.

Answer 1

Initialize it with the actual class it’s created in:

logger = Logger.getLogger(getClass()); //log4j way of creating loggers

To do this you’ll need to remove static modifier from your logger declaration.
I prefer to keep it private and access it via a getter-method, but you can also make it protected and access directly from A subclasses.

You should not worry that many logger objects will be created, one logger per class instance: under the hood Logger contains a map of loggers, and every time you create a new logger – it’s being cached. When you try to get logger for the same class the second time – it’s just being retrieved from the inner map.

So, if you have 5 classes in your hierarchy – only 5 Logger objects will be created, no matter how many times you call getLogger(getClass()).