A Systematic Approach for Structuring Exception Handling in Robust Component-Based Software
5. Case Study
In this section, we describe a case study that has been conducted to assess the feasibility and benefits obtained from applying our approach to part of a real system. The main goal of this case study was to analyse the impact of the proposed approach when applied to an existing system, in terms of both separation of normal and exceptional activities and reuse of the implementation of existing components. The target system, called Telestrada, is a large traveller information system being developed for a Brazilian national highway administrator. It comprises five subsystems: Central Database Subsystem, GIS (Geographic Information System) Subsystem, Call-Centre Operations Subsystem, Roadside Operations Subsystem, and Complaint Management Subsystem.
The case study consisted in applying our exception handling strategy presented in Section 4 to the Complaint Management Subsystem (CMS), in order to model its exceptional behaviour. This subsystem is a web-based application implemented in Java using the COSMOS model (Section 3.2). The implementation of the CMS comprises 12175 lines of source code (1598 automatically generated), as measured by the Unix wc (word count) command, and more than 300 classes. It is based on popular technologies, such as Enterprise Java Beans, Java Server Pages and Servlets, and the Struts framework.
The case study covered two iterations of the implementation of the CMS. During the first iteration, it was produced an initial implementation of the CMS in which exception handling was introduced in an ad hoc manner. The development of the proposed approach occurred after the conclusion of this initial implementation. Hence, the first iteration was not influenced by the proposed approach. During the second iteration, our approach was applied to obtain a robust implementation with a structured exceptional behaviour of the CMS. Another developer that was familiar with the proposed approach but had no previous contact with the CMS conducted this second iteration. Hence, the conditions under which the second iteration was conducted were similar to those of a real software development effort.