基于变型空间代数的自动程序修复方法

发布时间：2018-11-24 14:48

【摘要】：基于代码枚举的自动程序修复方法借助变异算子对程序中错误语句进行变更操作,从而得到程序修复解.由于缺乏文法制导及变异算子数量的有限性,该方法的有效性有待进一步提高.本文提出一种基于变型空间代数的自动程序修复方法,即将回归测试用例集视为训练实例,通过归纳学习得到程序中出错语句的修复解.具体而言,该方法包括以下特征:(1)从文法到变型空间的自动构造生成方法;(2)根据变型空间树中变型空间的不同类别,分别给出一致性定义;(3)结合静态及类型检查的变型空间代数运算.实验结果表明:与基于代码枚举及基于搜索的修复方法相比,本文提出的方法在修复成功率方面更具优势;与此同时,方法中的静态及类型检查机制可以有效地削减假设空间的规模.
[Abstract]:The automatic program repair method based on code enumeration changes the error statement in the program by using mutation operator and obtains the program repair solution. Due to the lack of the finite number of grammatical guidance and mutation operators, the effectiveness of this method needs to be further improved. In this paper, an automatic program repair method based on variant space algebra is proposed. The regression test case set is regarded as a training example, and the repair solution of the error sentence in the program is obtained by inductive learning. In particular, the method includes the following features: (1) automatic generation method from grammar to variant space; (2) definition of consistency according to different categories of variant space in variant space tree; (3) the algebraic operation of variant space combined with static and type checking. The experimental results show that the proposed method is superior to the code enumeration and search-based repair methods in the restoration success rate. At the same time, static and type checking mechanisms in the method can effectively reduce the size of the hypothetical space.
【作者单位】：广东肇庆学院数学与统计学院;武汉大学计算机学院;上海工程技术大学电子电气工程学院计算机系;
【基金】：国家自然科学基金(No.61640221,No.61603242) 上海高校青年教师培养资助计算专项基金(No.ZZGCD15088) 肇庆学院科研基金(No.201734) 肇庆市科技创新指导类项目(No.201704030409)
【分类号】：TP311.5

【相似文献】