Programming Logic Modeling And Application Of Object-Oriented Source Code

In 2014,the US Department of Defense Advanced Research Projects Agency proposed the Mining and Understanding Software Enclaves program to explore the method of deeply mining specifications and protocols on large-scale open source code bases,and to guide developers to complete complex work in the software field.As one of the key technologies,the mining of programming logic has gradually become a research hotspot for domestic and foreign scholars.This thesis takes open source codes as the research object of study,focusing on the general rules and individual habits of the human programming process.By exploring the object-oriented open source code programming logic modeling and application,this thesis mines the potential logic and rules of human programming,intending to provide basic support to the technology of vulnerability-mining,the programming behavior prediction,and the repairing and building of complex software.The summery of this thesis is as follows:1.On the bases of the concept of programming logic and the analysis of relating issues regarding the research of programming logic,it proposes a research framework of object-oriented open source code programming logic,and clarified the connection between various research contents,laying the foundation for the study of programming logic.2.The complex structure of object-oriented open source code brings difficult problems to programming logic analysis modeling.In view of the above-mentioned facts,this thesis propose the method of object-oriented language code analysis,on the bases of taking object as the standard for classification and the sequence of method calls as the basic structure,integrating alias merging and program structure analysis strategies and transforming the source code into a unified paradigm data format with preserving the original grammatical structural information as much as possible to lay a data foundation for the study of programming logic.3.The existing code model for large-scale data does not make full use of the syntax and structure of the code,and there are limitations to the description of the developer's programming logic.In view of the above-mentioned facts,this thesis propose the concept of the constraint relationship of the method.Based upon the above concept,it proposes an object-oriented language programming logic model CPMMC,using the sequence of methods involved in the object in building a model structure and abstracting the developer's thinking mode by describing the constraint relationships between different methods in the same sequence.To verify that the model embodies developer programming logic,we apply the model to the prediction of Java language methods.Experiments indicate that the model possesses the ability of the existing model to predict Java code,raising the accuracy by 8 percent.4.Since there exists problems such as sparse data and poor cross-program detection in the existing text-based code defect detection methods,it proposes an object-oriented code defect detection method based on CPMMC.The employment of programming logic instead of simple text analysis reduces the impact of different programming styles on detection performance.Calculate the contribution value of different features to code defects through hypothesis testing ideas and accomplish dimensionality reduction of feature vectors.A classifier is constructed based on support vector machines,resulting in the detection of code defects.This thesis evaluates the proposed method with the employment of some Android applications with tags,and the results indicate that the proposed method has significantly improved the detection performance of crossprogram code defects compared with the existing methods.