A Study of Call Graph Construction for JVM-Hosted Languages

Call graphs have many applications in software engineering, including bug-finding, security analysis, and code navigation IDEs. However, the construction of call requires significant investment program analysis infrastructure. An increasing number programming languages compile to Java Virtual Machine (JVM), frameworks such as WALA SOOT support a broad range algorithms by analyzing JVM bytecode. This approach has been shown work well when applied bytecode produced from code. In this paper, we show that it also works for diverse other JVM-hosted languages: dynamically-typed functional Scheme, statically-typed object-oriented Scala, polymorphic OCaml. Effectively, get graph these free, using existing infrastructure Java, with only minor challenges soundness. This, turn, suggests bytecode-based could serve an implementation vehicle IDE features languages. We present qualitative quantitative analyses soundness precision constructed bytecodes languages, Groovy, Clojure, Python, Ruby. details matter greatly. particular, implementations Ruby produce very unsound graphs, due pervasive use reflection, invokedynamic instructions, run-time generation. Interestingly, dynamic translation schemes employed which result static tend be correlated poor performance at run time.