Research And Implementation Of Large Scale Code Test Module Partition Method

With the rapid development of the Internet,the scale of software is becoming larger and larger,the iterative speed is getting faster and faster,and the quality of software is also facing challenges.At present,the amount of large-scale open source software code even reaches millions of lines,tens of millions of lines.In the traditional stand-alone environment,static defect detection of large-scale software will lead to long testing time,or even failure due to lack of resources.In order to solve the problem of static defect testing efficiency of large-scale code,this paper designs a large-scale code module partition method,and implements the distributed detection architecture of this method.In this paper,the dependency relationship of the project under test is modeled,and the dependency between the code files is expressed in the form of directed graph.By simplifying and analyzing the directed graph,it is divided into several minimum inseparable modules.Then the task quantity of the minimum inseparable module is evaluated,and the available resource nodes and the minimum inseparable module set are analyzed by undirected graph.According to the number of available resources and modules,the undirected graph is classified,and the maximum matching of the undirected graph is tried to find out,so as to carry out static module partition and reorganization.In order to ensure the accuracy of the static partition results,it is necessary to monitor the defect detection of the static partition results,evaluate the timeout of the exception module,and establish an undirected graph between the exception module and the available resources,so as to reorganize and partition again.On this basis,the dynamic module partition is carried out.In order to achieve the above module division method,this paper designs and implements a distributed detection system for defect detection based on yarn resource management framework,which supports resource management,task scheduling,task monitoring and other functions.In the end,the module partition method and distributed detection system are verified by experiments.Through the partition and distributed detection of six large-scale code projects,on the basis of the same hardware,the test efficiency is improved by more than 30%compared with the stand-alone environment.The more lines of code,the higher the detection efficiency of the distributed system.