类簇级测试中类测试序的生成技术研究

Determination of class-integration-test order is a key and difficult point of class cluster testing problem. The main topics are as follows: (1) we plan to determine the graph-based class integration test order, satisfying the overall complexity of created test stubs as small as possible: Fully considering the influence of the characteristics of object-oriented programming, such as encapsulation, inheritance and polymorphism, abstract classes cannot be instantiated, etc. and the influence of the three different inheritance relationships between classes such as public inheritance, protect inheritance, and private inheritance on the elimination of loop, we plan to propose the corresponding cycle elimination algorithm and construction rules of test stub, and then determine the class test order. (2) we plan to determine the class integration test order based on particle swarm optimization algorithm, satisfying the overall complexity of created test stubs as small as possible: first, we fully consider the influencing factors of fitness function, and then construct appropriate fitness function according to the specific test requirements that the overall complexity of the test stub should be as small as possible. The particles of entire population solve the optimal value of fitness function in the process of evolving, with which guide the generation of test order, and then the satisfied class integration test order is obtained. (3) in order to verify the feasibility and effectiveness of the proposed theories and methods, we plan to design and implement a prototype system that can automatically generate class integration test order. The project is a novel crossover research issues (study direction) that combining software engineering with mathematics, and has the significant value and wide application foreground. The project shows wide research prospects and has an important theoretical and practical significance.

类测试顺序的确定是类簇级测试中的一个重点和难点问题。本项目研究内容包括（1）基于图论的类集成测试顺序的确定，使测试桩总体复杂度尽可能小：全面考虑面向对象程序的特性，如封装、继承和多态、抽象类不可实例化等，以及类之间不同的继承关系，如公有继承，保护继承，私有继承对消除环路的影响，提出相应的环路消除算法以及测试桩的构造规则，进而确定类测试顺序。（2）基于粒子群算法的类测试顺序的确定，使测试桩总体复杂度尽可能小：首先充分考虑影响适应度函数的因素，再根据满足测试桩的总体复杂度尽可能小的要求构造合适的适应度函数，整个种群的粒子在不断运动过程中求解适应度函数的最优值，再用最优值来引导测试顺序的生成，进而得到满足条件的类集成测试顺序。（3）为了验证所提理论与方法的可行性和有效性，设计并实现一个自动生成类集成测试顺序的原型系统。本项目是软件工程与数学有机结合的新颖交叉研究方向，具有广阔的研究和应用价值。

研究类簇级测试中类测试顺序的确定问题。本项目研究成果主要包括（1）针对基于图论的类集成测试顺序的确定问题，首先提出了类间耦合信息的度量方法，测试桩复杂度的计算方法，分析了抽象类不可实例化的特点以及类之间不同的继承关系对打破环路的影响，提出了环路消除算法，并分析了消除环路之后抽象类不可实例化的特点对类簇的排序阶段的影响；其次提出了基于奖惩机制的类集成测试序列生成方法，实验结果表明，本方法生成类集成测试序列所需时间更少，所需的测试代价较小；最后还提出了基于传递依赖的类集成测试序列生成方法，实验结果表明：传递依赖会对类集成测试序列的测试代价产生影响，对于大部分实验系统而言，生成的类集成测试序列的测试代价更低。（2）针对基于进化算法的类测试顺序的确定问题，提出了基于搜索的类集成测试序列问题中初始种群的生成方法，该方法解决了现有方法初始种群个体适应度较低，导致寻优效果不佳的问题；同时提出了基于粒子群优化算法的类集成测试序列确定方法，实验发现对于规模较大的系统，或者类的个数较多，且包含环路的个数也较多的系统来说，基于搜索的方法更适应于解决这类系统的类集成测试顺序问题。此外还提出了基于梦境粒子群优化的类集成测试序列生成方法，实验发现在收敛速度上，与粒子群优化方法相比，可以减缓收敛速度，避免过早陷入局部最优。将本项目方法应用于开源项目软件中，研究成果进一步深化和丰富了已有的软件测试理论与方法。项目资助共发表期刊论文17篇，其中，SCI论文8篇，国内五大学报论文7篇,中文核心期刊2篇；发表会议论文2篇；出版专著1部；申请7项发明专利，其中，已授权4项。项目培养博士研究生3名，已经毕业2名并取得博士学位，1名在读；培养硕士研究生6名，已经毕业3名并取得硕士学位，3名在读。

内容获取失败，请点击重试