Decompiling Java using staged encapsulation

• Published in: Proceedings Eighth Working Conference on Reverse Engineering pp. 368 - 374
• Main Authors: Miecznikowski, J., Hendren, L.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 2001
• Subjects: Computer science; Constraint optimization; Data mining; Design optimization; Encapsulation; High level languages; Java; Pattern matching; Tree graphs; Visualization
• ISBN: 0769513034; 9780769513034
• ISSN: 1095-1350
• DOI: 10.1109/WCRE.2001.957845

The paper presents an approach to program structuring for use in decompiling Java bytecode to Java source. The structuring approach uses three intermediate representations: (1) a list of typed, aggregated statements with an associated exception table, (2) a control flow graph, and (3) a structure encapsulation tree. The approach works in six distinct stages, with each stage focusing on a specific family of Java constructs, and each stage contributing more detail to the structure encapsulation tree. After completion of all stages the structure encapsulation tree contains enough information to allow a simple extraction of a structured Java program. The approach targets general Java bytecode including bytecode that may be the result of front-ends for languages other than Java, and also bytecode that has been produced by a bytecode optimizer. Thus, the techniques have been designed to work for bytecode that may not exhibit the typical structured patterns of bytecode produced by a standard Java compiler. The structuring techniques have been implemented as part of the Dava decompiler which has been built using the Soot framework.