Statistical properties of software networks have been extensively studied. However, in the previous works, software networks are usually considered as a single-layer network, which cannot capture the authentic characteristics of software since software in its nature should be multilayer. In this paper, we explore the structural properties of the multilayer software network at the class level by progressively merging layers together, where each coupling type such as inheritance, implements, and method call defines a specific layer. A case study in software Tomcat is conducted using a set of 10 measures widely used in complex network literatures. The results show that some structural properties that are widely observed in software network researches can only emerge when several layers are merged together, such as high clustering coefficient, small value of average shortest path length, and high global efficiency. Our study highlights the importance of taking into consideration the multilayer nature of software systems. The results we found can provide valuable insights to our understanding and modeling of the dynamical processes taking place in the design and development of software systems.

中文翻译：

---

多层软件网络的结构特性：以 TOMCAT 为例

软件网络的统计特性已被广泛研究。然而，在以前的工作中，软件网络通常被认为是单层网络，由于软件本质上应该是多层的，因此无法捕捉软件的真实特征。在本文中，我们通过逐步将层合并在一起来探索类级别的多层软件网络的结构特性，其中继承、实现和方法调用等每种耦合类型都定义了一个特定的层。使用在复杂网络文献中广泛使用的一组 10 项措施对软件 Tomcat 进行了案例研究。结果表明，在软件网络研究中广泛观察到的一些结构特性只有在几个层合并在一起时才会出现，例如高聚类系数，平均最短路径长度值小，全局效率高。我们的研究强调了考虑软件系统的多层性质的重要性。我们发现的结果可以为我们理解和建模软件系统设计和开发中发生的动态过程提供有价值的见解。