Abstract:
Today a massive amount of source code is available on the Internet and open to serve
as a means for code reuse. Developers can reduce the time cost and resource cost by
reusing these external open source code in their own projects. Even though a number
of Code Search Engines (CSE) are available, finding the most relevant source code is
often challenging. In this research, we proposed a framework that can be used to
overcome the problem faced by developers in code searching and reusing. The
framework starts with the software architecture design in XML format (Class
Diagram), extracts information from the XML file, and then based on the extracted
information, fetches relevant projects using three types of crawler from GitHub,
SourceForge, and GoogleCode. We will have a huge amount of projects by
downloading process using the crawlers and need to find most relevant projects among
them.
In this research, we particularly focus on projects developed using Java language.
Each project will have a number of .java files, and all files will be represented as
Abstract Syntax Trees (AST) to extract identifiers (class names, method names, and
attributes name) and comments from the .java files. Then, on one hand, we will have
the identifiers which are extracted from the XML file (Class diagram), and the other
hand the identifiers and the action words (verbs) extracted from downloaded projects.
Action words are extracted from comments using Part of Speech technique (POS).
These two group of identifiers need to be analyzed for matching, if the identifiers are
matched, an amount of marks will be given to these identifiers, likewise marks will
be added together and then if the total marks is greater than 50%, the .java file belongs
to these identifier will be suggested as relevant code. Otherwise, synonyms of the
identifiers will be discovered using WordNet, and the matching process will be
repeated for the synonyms. For the composite identifiers, camel case splitter is used to
separate these words. If the programmers do not follow camel case naming convention,
N-gram technique is used to separate these word. The Stanford Spellchecker is used to
identify abbreviated words. Evaluation of our developed framework resulted in
95.25% of average accuracy of four subsystem [project downloader (100%), identifier
analyzer (94%), word finder (87%), and comments analyzer (100%)] accuracy.
