Abstract: All of the automatic plan-based instruments in program comprehension acquire relevant semantic information by comparing the source codes with the standard plans in pre-constructed plan database. The rough-match based on applying characteristic symbol as the index can greatly accelerate the matching progress, but it also brings the disadvantage of information leak. This article proposes a refined-match method based on fuzzy reasoning to analyze the similarity between the potential plans and the source codes, and gets the ranked candidate plan set.

Key words: Program comprehension, Characteristic symbol, Fuzzy reasoning, Refined-match

摘要： 所有基于Plan的自动程序理解工具都采用把源码段跟标准Plan相比较的方法来得到源码语义信息。采用程序段的特征信息作为匹配索引项进行粗匹配可以大大加快匹配效率，但完全依赖特征信息作为匹配代理造成了信息流失。该文提出了一种基于模糊推理的细匹配方法来进行相似度分析，目的是对粗匹配得到的潜在Plan集进行相似度分析以得到最终候选Plan集。

关键词: 程序理解, 特征符号, 模糊推理, 细匹配

CLC Number: 

• TP18