Bad-Smells in software are patterns of poor design and bad programming. These patterns can be removed from the software system by using refactoring techniques which improve the readability, maintainability and comprehension of the software system. This thesis describes an automated bad smell detection process in the Java source-code by static analysis of code-structure.

A prototype tool, JSmell, has been developed to detect the bad smell. Once the bad smell has been detected by the tool, various refactoring techniques are suggested to eliminate the smell. The developers can then focus on the elimination of poor design constructs. The analysis and detection process involves software-metrics evaluation and parsing technique, respectively. The success rate for bad smell detection process is 85%-90% when tested against seven different test cases for a single version of system. JSmell also represents the structural analysis of source code in terms of classes, methods and data fields which facilitates the developers in understanding the high level system architecture.

Keywords. Bad-smell, Code Smell, Refactoring, Object-Oriented Design, Software Metrics, Taxonomy, Design defects, Source-code, Software Maintenance, Reverse Engineering, Lexer, Parser