Semantic code search and analysis

Title from PDF of title page, viewed on July 28, 2014Thesis advisor: Yugyung LeeVitaIncludes bibliographical references (pages 33-35)Thesis (M. S.)--School of Computing and Engineering. University of Missouri--Kansas City, 2014As open source software repositories have been enormously growing, the high quality source codes have been widely available. A greater access to open source software also leads to an increase of software quality and reduces the overhead of software development. However, most of the available search engines are limited to lexical or code based searches and do not take semantics that underlie the source codes. Thus, object oriented (OO) principles, such as inheritance and composition, cannot be efficiently utilized for code search or analysis. This thesis proposes a novel approach for searching source code using semantics and structure. This approach will allow users to analyze software systems in terms of code similarity. For this purpose, a semantic measurement, called CoSim, was designed based on OO programing models including Package, Class, Method and Interface. We accessed and extracted the source code from open source repositories like Github and converted them into Resource Description Framework (RDF) model. Using the measurement, we queried the source code with SPARQL Query Language and analyzed the systems. We carried out a pilot study for preliminary evaluation of seven different versions of Apache Hadoop systems in terms of their similarities. In addition, we compared the search outputs from our system with those by the Github Code Search. It was shown that our search engine provided more comprehensive and relevant information than the Github does. In addition, the proposed CoSim measurement precisely reflected the significant and evolutionary properties of the systems in the similarity comparison of Hadoop software systemsAbstract -- Illustrations -- Tables - Introduction -- Background and related work -- Semantic code search and analysis model -- Semantic code search and analysis implementation -- Results and evaluation -- Conclusion and future work -- Reference