Visualizing interaction patterns in program executions.

摘要

Implementing, validating, modifying, or reengineering a software system requires an understanding of the program's behavior. Many factors make this understanding difficult, including the gulf in terms of abstraction between design models and source code, and the design drift that occurs over the lifetime of a system. During this research several generations of visualizations have been designed and built to support particular behavioral understanding tasks. As a result, an innovative visualization technique for creating global overviews of large information spaces was developed. To help bridge the gap between low-level events in program executions and higher-level models of program behavior, the notion of interaction patterns as recurring interaction scenarios in program executions was defined. The result is a framework to identify, visualize, and analyze interactions in program executions as a means for examining and understanding behavior. The thesis that visualizing interaction patterns in program executions can facilitate program behavioral understanding during design recovery, design/implementation validation, and reengineering tasks has been validated by several case studies.