Object-Oriented Program Specialization: Aspects Into Aspects — Or Maybe Not?

Automatic program specialization is a software engineering technique that configures a program fragment by generating an implementation dedicated to a specific usage context. Given a generic component that solves a whole family of problems and that is implemented in a standard programming language, program specialization can automatically configure this component by generating a specialized implementation. We consider automatic program specialization implemented using partial evaluation, which performs aggressive interprocedural constant propagation of all data types, and performs constant folding and control-flow simplifications based on the usage context [1]. Automatic program specialization has recently been defined in the context of object-oriented languages, and is here referred to as object-oriented program specialization [4, 5, 6, 7, 8]. As part of this work, we have developed an automatic program specializer for Java, named JSpec, which has been shown to give significant execution-time speedups on large programs. Specialization of an object-oriented program generates new, specialized methods, which must be reintroduced back into the class structure of the program. The dependencies between the specialized methods cross-cut the class structure of the program, which brings aspect-oriented programming to mind [2]. Using aspect-oriented programming, the result of specializing an object-oriented program can be elegantly expressed by encapsulating the specialized methods into an aspect. Concretely, the JSpec specializer generates specialized programs as AspectJ aspects [13], which are woven into the generic program as part of the compilation process. Specialization of an object-oriented program can be controlled using the specialization class framework [12]. A specialization class provides specialization information about a class in the program, by indicating what information is