Research On Code Search Technology Based On Features Of Code And Comment

In the field of software engineering,the quality,efficiency,and cost of software development are the three core issues of concern in the software development process.Since the beginning of the 21 st century,with the popularity of the Internet,information technology has shown an explosive growth,the scale and complexity of software systems are also increasing,and the efficiency of software development has also attracted more and more attention.In order to improve development efficiency,many software development technologies have been proposed one after another.Developers hope to achieve efficient code reuse through technical means such as code search.Therefore,the research of code search technology is of great significance.However,the existing code search technology is not comprehensive when representing code,and most of them only focus on the information of the code itself and ignore the comment information of the code,resulting in incomplete code representation and low accuracy of code search in actual search.In order to improve the accuracy of code search,this thesis proposes a code search technology based on code features and comment features.Different from the existing code search technology,this technology extracts the code features of the source code itself and the comment features contained in the code comments.The two types of features are combined to generate multi-dimensional code annotation,which can represent the code more completely.This thesis designs and builds a deep neural network for multi-dimensional code annotation and vectorization of query statements.Through multi-dimensional code annotation and vector results of query statements,the similarity between vectors is calculated based on similarity.In addition,this article designs and implements a code search tool-Code Hunter,which can search for qualified codes based on natural language input.Through the experimental test of 7 million Java method code segments in 100 open source projects,the results show that the code search technology based on features of code and comment and the tool Code Hunter in this paper have higher accuracy and average reciprocal in terms of metrics than other code search tools,and the search technology in this thesis has better.At the same time,this thesis proposes code search technology based on code features and comment features that can search out the corresponding code based on natural language input,which helps developers to search code.