Multi-granularity Characteristics Analysis In Large Scale Software Network

As the software scales are gradually increasing, the complexities of software are continuously improving too, in software architecture; there are various forms of integration in different levels, different sizes, and different organizing types. If only research the single level of software network cannot reflect the complete characteristics of the overall structure of the software, so we need a new methods to analysis the different granularity of software structure.In view of such problems, this paper abstracts the object-oriented software into three granularities of the software network; through analyze the different granularity of software network characteristics to find the rules which are contained in software architecture. This paper designed and implemented an object-oriented multi-granularity network analysis tool, it can be used to resolve the object-oriented software source code, and extract the package granularity, class granularity and method granularity of three different levels of software network and calculate the value of their network features.Analysis the network characteristics in three different granularity of software network, the results show that all of three granularity of software network of object-oriented software have the complex network features which are small-world and free-scale. The power-law exponent of out-degree is bigger than power-law exponent of in-degree distribution in package-granularity and class-granularity software network, while in-degree distribution is greater than the out-degree distribution, it reflect the design ideas of encourage the reuse and reduce the complexity in object-oriented software. The cluster-degree correlation of nodes in multi-granularity software network shows a linear approach to k-1, it shows that all of the three granularity of software network have structural feature of hierarchy, which reflects the modular design idea. The betweenness and degree correlations of nodes appear in three different areas of the division. In three different granularity of software network, the distribution of correlations between structural holes and degree fit the power-law distribution, there are two significant division in closeness-degree correlations, all the three granularity software network have the nodes which have smaller closeness, but the degree of these nodes are not necessarily small, these nodes are on the edge of network, the nodes which are in the center of network have the bigger closeness and the value of closeness is gradually stabilized with the increase of degree, these have the great significance in researching the collaborative between the modules and analyzing the structural feature of software.Through the object-oriented software multi-granularity software network analysis, it will be more complete and in-depth analysis of the structural characteristics of different levels of software, it has great practical significance in controlling the complexity of software and researching how software design ideas impact of software architecture.