Collaborative Research: Formal Methods for Behavioral Subclassing and Callbacks

提案0429894/Naumann 0429567/Leavens合作研究：行为子类化和回调的形式化方法。Naumann和加里T. Leavens为了提高演化性、可伸缩性和生产力，软件系统必须由可扩展的组件组成。面向对象编程语言的特性，如继承和动态分派，以及回调和向下调用等技术，是至关重要的，但它们颠倒了通常的抽象层次。 对象之间的别名对于效率至关重要，但可能会突破封装边界。 然而，抽象和封装是必要的，分别验证单个组件。这个项目将推进面向对象软件的规范，开发和验证的理论，重点是行为子类化，别名限制和回调。 将研究Java建模语言（JML）的行为接口规范的核心特征，使用限制规则来控制别名和模型程序来指定回调。 这些想法将被呈现在一个令人信服的，简单的，和强大的理论，以促进工具的应用程序，规范符号和证明规则的替代方案之间的比较，和教学。 该理论将被编码在一个定理证明器中，并对关键结果进行机器检查。该项目将为程序设计和规范语言的设计者提供理论指导。 研究结果还将有助于澄清和改进实践中使用的技术。 预计将直接应用于基于JML的项目和我们巴西合作者的工作。

Proposals 0429894/Naumann 0429567/LeavensCollaborative Research:Formal Methods for Behavioral Subclassing and CallbacksDavid A. Naumann and Gary T. LeavensFor evolvability, scalability, and productivity, software systems must be composed of extensible components. Features of object-oriented programming languages such as inheritance and dynamic dispatch, andtechniques like callbacks and downcalls, are crucial but they invert the usual layering of abstractions. Aliasing among objects is crucial for efficiency but can breach encapsulation boundaries. Yet abstractionand encapsulation are necessary to separately validate individual components.This project will advance the theory of specification, development, and verification for object-oriented software, focusing on behavioral subclassing, alias confinement, and callbacks. Core features of theJava Modeling Language (JML) for behavioral interface specification will be studied, using a confinement discipline to control aliasing and model programs to specify callbacks. The ideas will be presentedin a cogent, simple, and robust theory, to facilitate applications by tools, comparison between alternative proposals for specification notations and proof rules, and teaching. The theory will be encodedin a theorem prover and key results machine-checked.This project will provide theoretical guidance for the designers of programming and specification languages. The results will also help clarify and improve techniques used in practice. Direct application is expected in projects based on JML and in work by our Brazilian collaborators.