Question

I am trying to make a Java tool that will scan the structure of a Java application and provide some meaningful information. To do this, I need to be able to scan all of the .class files from the project location (JAR/WAR or just a folder) and use reflection to read about their methods. This is proving to be near impossible.

I can find a lot of solutions based on URLClassloader that allow me to load specific classes from a directory/archive, but none that will allow me to load classes without having any information about the class name or package structure.

EDIT:
I think I phrased this poorly. My issue is not that I can’t get all of the class files, I can do that with recursion etc. and locate them properly. My issue is obtaining a Class object for each class file.

Answer 1

The following code loads all classes from a JAR file. It does not need to know anything about the classes. The names of the classes are extracted from the JarEntry.

JarFile jarFile = new JarFile(pathToJar);
Enumeration<JarEntry> e = jarFile.entries();

URL[] urls = { new URL("jar:file:" + pathToJar+"!/") };
URLClassLoader cl = URLClassLoader.newInstance(urls);

while (e.hasMoreElements()) {
    JarEntry je = e.nextElement();
    if(je.isDirectory() || !je.getName().endsWith(".class")){
        continue;
    }
    // -6 because of .class
    String className = je.getName().substring(0,je.getName().length()-6);
    className = className.replace('/', '.');
    Class c = cl.loadClass(className);

}

edit:

As suggested in the comments above, javassist would also be a possibility.
Initialize a ClassPool somewhere before the while loop form the code above, and instead of loading the class with the class loader, you could create a CtClass object:

ClassPool cp = ClassPool.getDefault();
...
CtClass ctClass = cp.get(className);

From the ctClass, you can get all methods, fields, nested classes, ….
Take a look at the javassist api:
https://jboss-javassist.github.io/javassist/html/index.html