An algebraic axiom environment for software testing (axenvironment).

Abstract

This dissertation describes the design and implementation of an algebraic axiom support environment for software testing. Since absolute software correctness is undecidable, "approximate" correctness is as good as software engineering can hope to do. The approximately correct behavior of a software system with respect to a specification can only be demonstrated incrementally, beginning with the modules of a system and finishing with the external interface. Software module specification in the form of algebraic axioms provides a base from which we can be complete and concise in developing and testing the behavior of modules. Algebraic axioms can also be useful for a variety of software issues such as reusability, completeness and consistency of a requirements specification, and the definition of abstract and hierarchical data types. The primary focus of this dissertation is that algebraic axioms can provide a complete and consistent means to record a specification with which to test a software system's behavior at the module level. A major aim of this research has been to specify and develop sufficient support software to demonstrate the viability of this approach in actual software development, making design for testability a development parameter. This research focuses on the following issues: (1) The relationship between algebraic axioms and other formal methods for specifying software behavior. (2) Extensions needed to make the algebraic axiom method encompass testing. (3) What software support is necessary to make algebraic specifications, with our extensions, useful for real-world software development. Results indicate that using the formal method of algebraic specifications can have a positive impact on software development when adequate and realistic support software is introduced into the process. The approach results in additional initial labor for a software system, but is shown to be economical in terms of testing completeness, maintenance, and potential reuse.