The below description is given by one bartender. I think it helps you.
=========================================================================
Case one:
code:
List<Integer> a = new ArrayList(); //warning
a.add(1);
Integer i = a.get(0);
System.out.println(i);
The first line produces a warning.
Because variable a has a generic type List<Integer> and whatever you get out of the list is guaranteed to be an Integer. Therefore the third line will compile and the whole snipped would output a 1.
Why is it allowed to have the ArrayList non-generic here?
Because old classes compiled with
java 1.4 or lower still have to work, and an old 4.1 class can easily have an ArrayList as return type.
Case two:
code:
List a = new ArrayList<Integer>();
a.add(1); // warning
Integer i = a.get(0); //
System.out.println(i);
Here a is of raw type. It stores only objects and you can retrieve only objets from that collection. Even if the object is of type ArrayList<Integer>.
Therefore you get now a compiler error in the third line. Because the cast (Integer) is missing.
You have to insert it manually (in comparison to generics on both side, the compiler would do this for you behind the scenes).
And now the second line produces a warning. Compiler also would warn you if the first line were List a = new ArrayList(); (raw type on both sides).
Because only the compile type matters, and it is non-parameterized.
You can store anything into it. A line a.add("howdy"); in this context would compile also.
Again this "half-generics" is also allowed because it must be guarantied that generic and legacy code can interact.