Correlation Code Smell Research Based On Software Version Changes

Code smell usually refers to the problems existing in the code,which is a bad symptom in software design.Most code smells do not exist alone,and they are related to other code smells.The combination of smells with correlation will make the code more difficult to understand and prone to errors,thus increasing the difficulty of software maintenance.In the software life cycle,the version of the software is constantly changing,accompanied by the change of code smell.At the same time,the combination of code smell with correlation will also change.The changes mainly focus on the intensity and quantity of code smell with correlation.The current research lacks attention on this aspect,which makes it difficult for developers to formulate a cost-effective refactoring practice strategy.Aiming at this research direction,this paper puts forward the intensity model of variable with correlation,and studies the intensity of code smell with correlation under different versions,At the same time,applies the existing Apriori algorithm to conduct empirical research on the quantity of code smell with correlation under different software versions.The main contents of this paper are as follows:(1)The change of the quantity and proportion of code smell with the change of software version is studied.(2)The change of the intensity of the combination of correlation code with the change of software version is studied.Based on the factor score and the contribution rate of factor explanation,the intensity model of the combination of correlation variables is proposed.The model relies on the factor analysis model to calculate the intensity of the combination of correlation variables and analyze the changes.(3)The change of the quantity of correlation code combinations with the change of software version is studied.Apriori algorithm is used to extract rules of code smell with correlation in different versions,and the change of combination is analyzed.This paper analyzes the correlation changes of 14 code smells,and carries out experiments on 361 versions of 27 systems.With the change of software version,the experimental results show that: the number of code smell shows an overall upward trend,and a few show a downward trend;the intensity of code smell with correlation depends on the combination,and there is no obvious consistent change trend.The number of code smell with correlation shows an overall upward trend,and specific to a single code smell will appear different changes.At the same time,it is found that long method affects many other code smells,so we should pay attention to this code smell when we refactor the code.