Decomposing God Header File via Multi-View Graph Clustering

God Header File refers to a header file with large code size and wide file impact. Such files pose difficulties in code comprehension and slow down compilation, ultimately increasing the maintenance cost during software evolution. Although this concept is similar to God Class, existing refactoring methods for God Classes are inappropriate for God Header Files. The reason lies in the fact that the code elements in header files are mostly short declaration types, and build dependencies of the entire system should be considered with the aim of improving compilation efficiency. Meanwhile, these methods overlook the concern of cyclic dependencies, which holds immense importance in the God Header File decomposition. To address these challenges, this paper proposes a God Header File decomposing approach based on multi-view graph clustering. It first constructs a code element graph with multiple relationships. Then after coarsening the graph, a novel multi-view graph clustering algorithm is applied to identify clusters of closely related code elements, and a heuristic algorithm is introduced to address the cyclic dependencies in the clustering result. We evaluate our approach on a synthetic dataset as well as six real-world God Header Files from different projects. The results show that our approach could achieve 11.5% higher accuracy in comparison to existing God Class refactoring methods. Moreover, our decomposition results attain better modularity on all the real-world God Header Files and reduce recompilation time for historical commits by 15% to 60%.