| Early detection and correction of errors in the software design phase can reduce total cost and time to market of a software product. Yet, analysis of design models usually consists of subjective walk-throughs and inspections, both of which lack the rigor of systematic evaluation. This dissertation describes an approach to systematically evaluate UML design models. The approach integrates design views described by Class and Sequence Diagrams into a combined model. The combined model, represented by a directed acyclic graph, is used to generate and execute tests. If OCL information is available, the nodes in the graph are decorated with OCL expressions which are used as a partial test oracle.; The approach can also be applied to evaluating coupling and cohesion metrics in UML designs. During Component Based Software Engineering, it is important for component developers to design components that show high cohesion within a component and low coupling between components. Current practice in design metric evaluation relies on extracting structural metrics from individual UML views. This dissertation defines a technique that collects dynamic metrics by observing symbolic execution in a combined model. The metrics are used to define components with high cohesion and low coupling properties.; In the design phase of software development, UML views change often. Thus we need an effective and efficient way to regression test designs. A retest approach that runs all tests can be inefficient resulting in higher software development costs. This dissertation presents an efficient and safe selective regression testing technique for the purpose of regression testing UML designs. The approach addresses how to determine which sections of designs have been modified, how to select tests from an existing test suite for regression testing, and how to generate new tests. |
