Modularizing Control

点击翻译按钮获取中文摘要

Aspect-oriented programming (AOP), an active research area, is interested in modularizing program features, including synchronization, distribution, and instrumentation, all of which crosscut and entangle with more fundamental application concerns.  By encapsulating and abstracting these features, programs can become easier to write, read, and maintain.  One fundamental form of AOP modularizes these features by a novel mechanism comprised of three parts: dynamic join points, pointcuts, and advice.  Existing semantic accounts of this mechanism are disconnected from AOP's goal of modular separation of concerns because the account imposes an ad-hoc extension to a base language.Based on doctoral research showing that dynamic join points arise naturally from the continuation semantics of a programming language, this research proposes a bridge between the modularity goals of AOP and the semantic structures it works with.  That bridge identifies dynamic join point aspects as modularizing control structure of programs.  In the short term, a number of projects are proposed to refine and extend this model of AOP, especially focusing on types and effects descriptions of the modular units made available by AOP.  By identifying and annotating these "abstract control types" in programs, we enable a fuller account of the programmers intentions and provide ways for those intentions to be checked against the actual program code.  This clarity and checking should result in fewer software defects, and possibly greater code reuse.In the long term, this research programme proposes to develop a "standard model of programming languages" which characterizes the elements and interactions at work in programming languages.  This modularity over control is the first part of that model.