Chapter 5. Evolution, Not Revolution

One motto underpinning the design of generics for Java is evolution, not revolution. It must be possible to migrate a large, existing body of code to use generics gradually (evolution) without requiring a radical, all-at-once change (revolution). The generics design ensures that old code compiles against the new Java libraries, avoiding the unfortunate situation in which half of your code needs old libraries and half of your code needs new libraries.

The requirements for evolution are much stronger than the usual backward compatibility. With simple backward compatibility, one would supply both legacy and generic versions for each application; this is exactly what happens in C#, for example. If you are building on top of code supplied by multiple suppliers, some of whom use legacy collections and some of whom use generic collections, this might rapidly lead to a versioning nightmare.

What we require is that the same client code works with both the legacy and generic versions of a library. This means that the supplier and clients of a library can make completely independent choices about when to move from legacy to generic code. This is a much stronger requirement than backward compatibility; it is called migration compatibility or platform compatibility.

Java implements generics via erasure, which ensures that legacy and generic versions usually generate identical class files, save for some auxiliary information about types. It is possible to replace a legacy class file by a generic class file without changing, or even recompiling, any client code; this is called binary compatibility.

We summarize this with the motto binary compatibility ensures migration compatibilityor, more concisely, erasure eases evolution.

This section shows how to add generics to existing code; it considers a small example, a library for stacks that extends the Collections Framework, together with an associated client. We begin with the legacy stack library and client (written for Java before generics), and then present the corresponding generic library and client (written for Java with generics). Our example code is small, so it is easy to update to generics all in one go, but in practice the library and client will be much larger, and we may want to evolve them separately. This is aided by raw types, which are the legacy counterpart of parameterized types.

The parts of the program may evolve in either order. You may have a generic library with a legacy client; this is the common case for anyone that uses the Collections Framework in Java 5 with legacy code. Or you may have a legacy library with a generic client; this is the case where you want to provide generic signatures for the library without the need to rewrite the entire library. We consider three ways to do this: minimal changes to the source, stub files, and wrappers. The first is useful when you have access to the source and the second when you do not; we recommend against the third.

In practice, the library and client may involve many interfaces and classes, and there may not even be a clear distinction between library and client. But the same principles discussed here still apply, and may be used to evolve any part of a program independently of any other part.