...
 ClassPath path = new ClassPath("/usr/lib/java/lib/classes.zip");
 ClassInfo clazz = path.getClassInfo("java.util.Hashtable");

 try {
   clazz.load(ClassInfo.DECLARATIONS);
 } catch (ClassFormatException ex) {
   System.err.println("Something is wrong with HashTable, giving up!");
   return;
 } catch (IOException ex) {
   System.err.println("Can't load HashTable, giving up!");
   return;
 }

 MethodInfo[] methods = clazz.getMethods();
 for (int i = 0; i < methods.length; i++) {
     String type = methods[i].getType();
     if (TypeSignature.getReturnType(type) == TypeSignature.INT_TYPE)
         System.out.println("Found integer method: "+method.getName());
 }
 ...

 ...
 ClassPath path = new ClassPath("/usr/lib/java/lib/classes.zip");
 ClassInfo clazz = path.getClassInfo("my.new.Class");
 clazz.setModifiers(Modifier.PUBLIC);
 clazz.setSourceFile("Class.pl");
 clazz.set...
 clazz.write(zipOutputStream);
 ...

Advantages of this bytecode package

• You don't need to think of the constant pool, except when you want to write your custom attributes.
• The set of opcodes is drastically reduced: For example you don't have to handle 20 different opcodes that all push a constant value on the stack. When reading it will automatically convert them to ldc or ldc2 and on writing it will convert them back.
• Wide instructions are automatically generated when needed, large methods are supported.
• The code is organized in {@link net.sf.jode.bytecode.BasicBlocks} which makes flow analysis much easier.
• The memory consumption is quite moderate.

Disadvantages

• You can't change every byte. For example Jode decides itself if a lookup switch or table switch is generated.
• Jode does a lot of checks when reading the bytecode and it is impossible to recover from errors. This makes it sometime hard to find out why the bytecode of a particular class files is invalid.