Conference Papers: Understanding Class Hierarchies with KABA

KABA is a prototype implementation of the Snelting/Tip analysis for JAVA. KABA combines dataflow analysis, type inference and concept lattices in order to perform a fine-grained analysis of member-access patterns in a class hierarchy together with a given set of applications. KABA computes a transformed hierarchy which is guaranteed to be 1. operationally equivalent, 2. maximally factorized, 3. minimal. The new hierarchy in particular makes obvious which classes can be splitted and which cannot; which inheritance relations must be retained and which can be discarded. The paper presents several case studies on medium-sized JAVA programs.