A Language-Independent Framework for Software Miniaturization

One of the undesired effects of software evolution is the proliferation of unused components, or components unlikely to be used by a given subset of the applications. As a consequence, the size of binaries and libraries tends to grow. One of the major trends of today’s software market is the porting of applications on hand-held devices or, in general, on devices having a limited amount of resources available. Several forms of refactoring and, in particular, the miniaturization of libraries and applications, are therefore necessary. We propose a framework and a toolkit covering several aspects of software miniaturization, such as removing unused objects and code clones, as well as creating small, cohese libraries refactoring the existing ones. The last step has been implemented using a hybrid approach based on hierarchical clustering, genetic algorithms and hill climbing, also incorporating the developer’s knowledge. Most of the framework activities are language independent (relying on object module analysis) thus they do not require any kind of source code parsing and are applicable to software systems developed with 1 DR AF T different programming languages. The proposed framework has been applied to GRASS, an over 1 million LOCs open source Geographical Information System, reducing of 50% the average number of objects linked by each application, and thus application’s memory requirements.