Hierarchical Small Worlds in Software Architecture
Hierarchical Small Worlds in Software Architecture
复制标题
DOI：
--
复制
发表时间：
2003-07
期刊：
影响因子：
--
通讯作者：
S. Valverde;R. Solé
中科院分区：
文献类型：
--
作者：
S. Valverde;R. Solé
大型软件应用程序的组件不会以随机方式进行交互。取而代之的是，类图与其他天然和人造系统表现出显着的拓扑相似性。大型软件应用程序的组件连接非常良好，因为尽管每个类的连接数量相对较少，但它们之间的平均最短距离非常低。另外，这些图非常异构。这些测量值具有一般性质，在很大程度上与应用程序的特定语义无关。如本文所示，无论分析了每个系统的特定特征，软件演变的最终结果是一个小世界，分层类图，具有明确定义的统计属性。概述了软件演化的后果。
The components of a large software application do not interact in random ways. Instead, class diagrams exhibit remarkable topological similarities to other natural and artificial systems. The components of a large software application are very well connected because the mean shortest distance between them is very low in spite of having a relatively small number of connections per class. In addition, these diagrams are very heterogeneous. These measurements are of a general nature and are largely independent of the particular semantics of the application. As shown in this paper, and irrespective of the specific features of each system analyzed, the final outcome of software evolution is a small world, hierarchical class diagram with well-defined statistical properties. The consequences for software evolution are outlined.