A hybrid framework documentation approach for effective use of object-oriented application framework documentation

Abstract

Object-oriented application framework presents one of the most successful approaches to developing reusable software components and libraries of software by providing a general skeleton of classes and behavior patterns for a given application domain. Several challenges and issues of developing, using, and maintaining object-oriented application frameworks have been identified in this research. It was discovered that companies or individuals attempting to build or use any object-oriented application frameworks often fail unless they recognize and resolve challenges such as development effort, steep learning curve, maintainability, validation, defect removal, efficiency, and lack of standards. Framework documentation plays a major role in facing the above challenges, and also in ensuring the success of object-oriented application frameworks in assisting users in developing application software using these frameworks. The researcher has studied various framework documentation approaches, performed a survey, and conducted numerous interviews on framework documentation approaches, and concluded that the current approaches are not very effective in overcoming the above-mentioned challenges and issues, especially in handling the steep learning curve of frameworks. Moreover, currently available documentation approaches are not very effective for new framework users. This research study applies machine learning using the combination of case-based reasoning (CBR) and genetic algorithm (GA) techniques to framework documentation. The main objective is to introduce an effective framework documentation approach that makes learning to use a specific object-oriented application framework less troublesome. In CBR, human reasoning is based on remembering past cases or experiences. A CBR system learns through examples and past experiences. Within this proposed approach, a case refers to a framework usage experience or problem. GA is used in optimizing the search for suitable solutions or cases to a given framework usage problem, and also in the adaptation process of similar cases to suit the requirements of the new case or problem. Nearest neighbor algorithm is applied in finding the most similar case or group of cases to the new case posed by the user. The similarity between an old case and the new case is measured by comparing how close they are in terms of the selected features they have. Knuth-Morris-Pratt pattern matching algorithm is used in finding certain patterns during the adaptation, learning and retaining process of the CBR. This is where machine learning takes place. New cases along with their solutions are saved inside the case base for future retrievals. In order to speed up framework learning process, the minimalist documentation technique is also implemented in the proposed approach. Minimalist technique is applied when framework users are asked to indicate what features of the framework they want to learn. Based on the user’s request, the proposed documentation approach provides the most similar case that gives the simplest solution to the problem at hand. The solution includes only the necessary instructions or statements that perform the requested task. This guarantees only minimal information is provided so that the user will have a better understanding on how to use the features of the framework. A documentation architecture that combines CBR, GA, and minimalist techniques is proposed in order to come up with more effective object-oriented application framework documentation. This is a hybrid documentation approach that applies machine learning in enhancing the system’s knowledge in the form of past framework usage experiences or cases. The evaluation of the approach indicates that the proposed hybrid approach is capable of producing solutions to new cases posed by users efficiently, and effectively. To implement the new hybrid documentation approach,

a prototype that uses this approach is developed in Java. The evaluation of the prototype shows that the proposed hybrid approach is able to reuse existing cases, adapt past cases to suit new cases, and learn new cases correctly. Overall, the prototype is successful in implementing this newly proposed hybrid object-oriented application framework documentation approach.