The Researches Of Two-Phase Test Suite Reduction Based On Structural Table

As a process to authenticate the software's safety,integrity and accuracy,the software testing is deemed as an important approach of guaranteeing the software quality by analyzing the actual output and the expected output,adjusting the testing strategy or changing the testing code.For the testing requirements and the test cases,as the two components of the software testing,the test cases are designed to carry out test sufficiently as requested.However,with iteration of the software version and change of the testing requirements,the test suite will be unceasingly expanding,which will thus result in redundancy and low degree of coverage,etc.Considering these problems,this paper makes some improvement in reduction of the traditional test suite.Its contributions as follows:1.For the testing requirement code of program slicing construction,this paper proposed the structural table-based method to construct the testing requirement code.The traditional method used graphs to construct the program slicing,which demanded computing of the nodes unrelated to the point of interest,Thus it was difficult in computing.In this paper,the structural table was adopted to construct the program slicing,which has reduced the difficulty in constructing the program slicing to a certain extent.2.For the reduction of testing requirements,this paper proposed a reduced testing requirement set based on the relationship between the coupling and the testing requirement.First of all,for the testing mode,this paper considered the testing target as the point of interest of program slicing,extracted and tested the target-related testing requirement code by means of program analysis of the structural table,then computed whether there was any requirement redundancy among the computing program modes through coupling formula,and finally reduced the testing requirement set with inclusion and intersection in consideration of the inclusion,independence and intersection among the testing requirements so as to instruct the reduction of test suite.3.For the reduction of test suite,this paper proposes the k-medoids based algorithm.First of all,this paper used k-medoids algorithm to categorize the test suite.In selecting the categorization standards,this paper considered the coverage and cyclomatic complexity to categorize the test cases,then selected the test cases according to their distance to the central points,from the near to the distant in the same category.For the test cases of equivalent distance,this paper selected the test cases covering more testing requirements to guarantee selection of test cases with high coverage while reducing the test suite.4.This paper,in selecting the clustering algorithm,considered k-medoids algorithm as the clustering algorithm,by this way,the division of the entire test suite would not be affected because the algorithm was non-sensitive to the outlier test cases.In the comparison experiment,this paper adopted the traditional G algorithm,GRE algorithm and k-means algorithm respectively for comparison from such aspects of the quantity of test suite,the error detection quantity before and after reduction,and the coverage before and after reduction.Then based on the analysis of the actual cases,this paper examined the effect of each algorithm under the same data suite from the said aspects,and thus verified the validity and feasibility of this method through experiment.