MITRAS : modelo inteligente para transformação de aplicações de software

Abstract

Software maintenance is costly for developers and customers as well. In this regard, new ideas and technologies are necessary to minimize the effort in this type of activity. This dissertation presents MITRAS project, an intelligent model for software transformation that enables users without programming knowledge to do modifications in systems via interaction in natural language. This is made possible combining state of the art tools and techniques in natural language processing and graph transformations applied to software models. MITRAS is designed to work with adaptive and perfective type of maintenances (as classified by IEEE), this maintenance do not imply that the system has an error, but that there is need for improvements identified by the user in already deployed systems User requests and the respective modifications are mapped in graphs that can be used in a graph transformation process that models system modifications. Synchronism between graph model and source code is maintained through automatic model extraction and differential code injection. This work presents both formal theoretical foundations and practical aspects of MITRAS computational model and a prototype developed and tested against a real open source Project. The model results are assessed using two methods, the first one aims to check execution time and complexity of the system, the second one compares the results of a user of MITRAS and software developers. The experiments are presenting good results so far, indicating that users can do simple maintenance tasks with MITRAS in less time then developers but with similar correctness.