Question

In the source for JDK 1.6, the Collections class’ swap method is written like this:

public static void swap(List<?> list, int i, int j) {
    final List l = list;
    l.set(i, l.set(j, l.get(i)));
}

What reason is there for creating a final copy of the passed list? Why don’t they simply modify the passed list directly? In this case, you also get the raw type warning.

Answer 1

There is no copy of the list, there is only a copy of the reference to the list. The final keyword is not important. However, it is important that a raw type is used. If the parameter would be used instead, the compiler would report an error:

public static void swap(List<?> list, int i, int j) {
    // ERROR: The method set(int, capture#3-of ?) in the type List<capture#3-of ?>
    // is not applicable for the arguments (int, capture#4-of ?)
    list.set(i, list.set(j, list.get(i)));
}

This means, that they are using the intermediate variable to circumvent the shortcomings of generics, and to get rid of the error message.

The interesting question is: Why don’t they use a generic method? The following code works:

public static <T> void swap(List<T> list, int i, int j) {
    list.set(i, list.set(j, list.get(i)));
}

The answer is, that this method produces warnings in old code invoking the method with raw types:

List list = ...;
// WARNING: Type safety: Unchecked invocation swap2(List, int, int)
// of the generic method swap2(List<T>, int, int) of type Swap
Collections.swap(list, 0, 1);