Thread: Class hierarchy... giving me a headache, so I could use some help

Hi all.

Making a small program that will read a .class file and display information. Using this as a reference point:
VM Spec The class File Format
It describes the class file format, hopefully in an accurate way. Anyway, I am trying to figure out a good way to read from the constant pool. If you read that link, in the section about the constant pool, it talks about several structures that are very similar. So I could just use a class hierarchy. Thing is, I am the master of complicating things for myself, and I am not sure what the best way to represent the different stuff would be. For example, here are two of the classes:

Base class:

public abstract class CPInfo
{
	// Declare the tag holding variable
	protected int mTag;
 
	/**
	 * Returns the tag of this cp info thing
	 * 
	 * @return the tag
	 */
    public int tag()
    {
        return mTag;
    }
 
    /**
     * ============
     * === TAGS ===
     * ============
     */
    public static final int TAG_CLASS = 7;
    public static final int TAG_FIELD_REF = 9;
    public static final int TAG_METHOD_REF = 10;
    public static final int TAG_INTERFACE_METHOD_REF = 11;
    public static final int TAG_STRING = 8;
    public static final int TAG_INTEGER = 3;
    public static final int TAG_FLOAT = 4;
    public static final int TAG_LONG = 5;
    public static final int TAG_DOUBLE = 6;
    public static final int TAG_NAME_AND_TYPE = 12;
    public static final int TAG_UTF8 = 1;
}

Represents an integer in the constant pool.

public class CPInteger extends CPInfo
{
	private int mInt;
 
	public CPInteger(int intval) throws ClassFormatException
	{
		super.mTag = CPInfo.TAG_INTEGER;
		mInt = intval;
	}
 
	public int value()
	{
		return mInt;
	}
}

Represents a float in the constant pool:

public class CPFloatInfo extends CPInfo
{
	private float mFloat;
 
	public CPFloatInfo(float floatval)
	{
		super.mTag = CPInfo.TAG_FLOAT;
		mFloat = floatval;
	}
 
	public float value()
	{
		return mFloat;
	}
}

As you can see, they are basically identical, except that one stores an int and one stores a float. Not happy with it, to be honest, since they are so similar. It feels like I could create a better way to do this. Anyone has a better idea of how to do this? Or is this the best way? I am getting a headache when I think of it too much, lol (figuratively speaking).

And do note that at the moment I am just trying to replicate the structure of a class file until I get a better grip of what I am actually doing. Have not done this before, so to be honest, I have no idea what I am doing.

Take care,
Kerr

EDIT:

One option would be to simply have an object in the CPInfo class and then cast it to whatever data I need (in the case of int and float I would use the Integer or Float classes). Could work, but not very type safe. Not that the way I am currently doing it is any better, for that matter...

Originally Posted by Kerr

...
As you can see, they are basically identical, except that one stores an int and one stores a float. Not happy with it, to be honest, since they are so similar. It feels like I could create a better way to do this. Anyone has a better idea of how to do this? Or is this the best way?
...
And do note that at the moment I am just trying to replicate the structure of a class file until I get a better grip of what I am actually doing. Have not done this before, so to be honest, I have no idea what I am doing.
...
One option would be to simply have an object in the CPInfo class and then cast it to whatever data I need (in the case of int and float I would use the Integer or Float classes). Could work, but not very type safe. Not that the way I am currently doing it is any better, for that matter...

I'd be inclined to use an enum for the tag values, but other than that, it's hard to say - if you don't know what you're doing this for, how can we suggest the best way to go about it? Maybe you need to decide your requirements before trying to any write code...

In any case, I'd avoid using an Object and casting to some type on demand, but you could use generics to make your CPInfo class deal with whatever types are required, something like this:

public class CPInfo<T> {
    // Declare the tag holding variable
    protected TAG mTag;
 
    private T mValue;
 
    public CPInfo(T value) throws ClassFormatException, UnknownTagException {
        mValue = value;
 
        // use the type class name to get the correct tag
        mTag = TAG.getTag(value.getClass().getSimpleName());
    }
 
    public T value() {
        return mValue;
    }
 
    /**
     * Returns the tag of this cp info thing
     *
     * @return the tag
     */
    public TAG tag() {
        return mTag;
    }
 
    /**
     * ============ === TAGS === ============
     * Each tag holds its corresponding class name
     */
 
    public static enum TAG {
        TAG_CLASS("Class"),
        TAG_FIELD_REF("Field"),
        TAG_METHOD_REF("Method"),
        TAG_INTERFACE_METHOD_REF("InterfaceMethod"),
        TAG_STRING("String"),
        TAG_INTEGER("Integer"),
        TAG_FLOAT("Float"),
        TAG_LONG("Long"),
        TAG_DOUBLE("Double"),
        TAG_NAME_AND_TYPE("NameAndType"),
        TAG_UTF8("UTF8");
 
        private String className;
 
        // constructor taking class name
        private TAG(String className) {
            this.className = className;
        }
 
        // return the tag for a given class name
        public static TAG getTag(String className) throws UnknownTagException {
            for (TAG tag : TAG.values()) {
                if (className.equals(tag.className)) {
                    return tag;
                }
            }
            throw new UnknownTagException(className);
        }
    }
 
    public static void main(String[] args) throws ClassFormatException, UnknownTagException {
        // create a float info object
        CPInfo<Float> floatInfo = new CPInfo<Float>(123.45f);
 
        // Print the tag and value
        CPInfo.TAG tag = floatInfo.tag();
        float value = floatInfo.value();
        System.out.println("Tag = "+tag.name());
        System.out.println("value = "+String.valueOf(value));
    }
}

I'm assuming here that you're going to have corresponding classes for each of those tags (e.g. a NameAndType class and a UTF8 class, etc), and that you don't actually need to store primitive types (autoboxing should take care of that). If not, then I don't really know what you want...

The problem is that I would still have to make some casts, since I would have a list of all CPInfo objects (for example, one containing a float and another an int). However, this gives me an idea. Instead of having a class hierarchy I do it the other way around. Instead of posing as a base class, CPInfo could instead simply contain the tag stating what it is and then store the bytes. The other classes would then simply accept a CPInfo as a parameter to their constructor, and then interpret the bytes as they should do. That way I wouldn´t have to cast them.

Enums could be good, yes, chose the way I did because I was striving for simplicity. On the other hand enums can offer more flexibility, since they can contain code. So I guess I will just move the tags into them.

Anyway, will have to play around a little to see if this works.

EDIT2:

Figured out how to do what I wanted with enums. Only issue is... it is not very type safe. This is the enum:

public enum Test
{
	ONE {
		@Override
		public String value()
		{
			return "One111";
		}
	},
	TWO {
		@Override
		public Float value()
		{
			return 2.2f;
		}
	},
	THREE {
		@Override
		public Integer value()
		{
			return 33;
		}
	};
 
	public abstract <T> T value();
}

And this is the code that shows what the issues are:

public class Program
{
	/**
	 * @param args
	 */
	public static void main(String[] args) throws Exception
	{
		String s = Test.ONE.value();
		System.out.println( s );
 
		Float f = Test.TWO.value();
		System.out.println( f );
 
		Float invalid = Test.THREE.value(); // Causes a ClassCastException to be thrown
		System.out.println( invalid );
	}
}

There is not very much safety here, since this code that causes an error will compile. How would I solve this?