Question

I’m using a third party library that returns a raw Iterator, e.g.

Iterator<?> children = element.getChildElements();

I know the actual type, but I don’t necessarily trust the third party lib to stick with it in the future. There are two (that I can think of) somewhat risky ways to traverse this:

@SuppressWarnings("unchecked")
Iterator<ActualObject> currentChildren = (Iterator<ActualObject>)currentElement.getChildElements();

or

Iterator<?> children = element.getChildElements();
while (null != children && children.hasNext()) {
  ActualObject child = (ActualObject)children.next(); //Possible ClassCastException @ runtime
  ...
}

The only “safe” way I could come up with to traverse this sort of iterator is the following:

Iterator<?> children = element.getChildElements();
while (null != children && children.hasNext()) {
  Object obj = children.next();
  ActualObject child = null;
    if (obj instanceof ActualObject)
      child = (ActualObject)obj;
    ...
}

This seems overly verbose. Is there a better, yet equally “safe” way to traverse a raw iterator?

EDIT: I realize I can catch/log the exception in an else block, I was looking (hoping) for a Java language equivalent of what ColinD has mentioned below.

Answer 1

Guava makes this easy with its Iterators.filter(Iterator<?>, Class<T>) method. It returns an unmodifiable Iterator<T> that basically just skips every element of the given iterator that isn’t an instance of type T:

Iterator<ActualObject> children = Iterators.filter(element.getChildElements(),
    ActualObject.class);

You can obviously then traverse the resulting iterator without needing to cast each element to ActualObject AND without having to worry about ClassCastException.