A semiautomated method for classifying program analysis rules into a quality model

Abstract

Most of the software code quality assessment and monitoring methods uses Quality Model (QM) as an aid to capture quality requirements of the software. An important aspect concerning use of QM is classification of Program Analysis (PA) rules into QM according to their relevance to quality attributes such as maintainability, reliability etc. Currently such classification is performed manually by experts and most of the PA tools (such as FxCop for C#, FindBugs for Java, PC-Lint for C/C++) support hundreds of PA rules. Hence performing classification manually can be very effort intensive and time consuming and can lead to concerns like subjectivity and inconsistency. Hence we propose a light weight semiautomated method to expedite classification and make classification activity less effort intensive. Proposed classifier is based on natural language processing (NLP) techniques and uses a keyword matching algorithm. We have computed precision and recall for such a classifier. We have also shown results from applying technique on classifying rules from FxCop, PC-Lint, and FindBugs into the EMISQ QM. We believe that proposed approach will significantly help in reducing the time required to perform classification and hence also to incorporate newer PA tools and rules into QM based methods.