Cross Platform API Mappings based on API Documentation Graphs

As different versions of the same application might be implemented based on different platforms/programming languages, it is significantly important to build an automated migration tool for the application programming interface (API) mapping relations between different platforms/programming languages. In this paper, we propose an approach to discover API mappings based on the API documentation. We first divide the information in the API documentation into different types of entities, relations, and attributes to construct their respective API Documentation Graphs (ADGs). Then, we encode nodes, edges and triplets of ADGs and input them to a new graph neural network (GNN) for entity alignment to obtain the API mappings between the two different platforms/programming languages. Taking HarmonyOS and Android as representative cases, we evaluate our approach based on their API documentation. The results show that our approach improves top-1, top-5, and top10 accuracies by 50.57%, 56.25%, and 52.66%, respectively, compared with documentation-based baselines.