Learning Api Mappings for Programming Platforms Abstract of the Thesis Learning Api Mappings for Programming Platforms

Software developers often need to port applications written for a source platform to a target platform. One of the key tasks here is to find matching API calls in the target platform for the given API calls in the source platform. This task involves exhaustive reading in target platform API (Application Programming Interface) documentation to identify API methods corresponding to the given API methods of the source platform. We introduce an approach to the problem of inferring mapping between the APIs of a source and target platforms. It is constructed based on independently developed applications on source and target platforms performing similar functionality. We observe that in building these applications, developers exercised knowledge of the corresponding APIs. We develop two dynamic analysis techniques to systematically harvest this knowledge and infer likely mappings between the Graphical APIs of JavaME and Android Graphical platform. Rosetta Mapper: A tool which provides a ranked list of target API methods or method sequences that likely map to each source API method or method sequences. Rosetta Classifier: A supervised learning tool which classifies whether the given mapping between source API and target API is true or false using support vector machines. iii Acknowledgements