基于遗传算法的软件测试技术文档格式.docx

基于遗传算法的软件测试技术文档格式.docx

《基于遗传算法的软件测试技术文档格式.docx》由会员分享，可在线阅读，更多相关《基于遗传算法的软件测试技术文档格式.docx（22页珍藏版）》请在冰豆网上搜索。

AzadHindFauzMarg,Dwarka,Sector-3,NewDelhi-110078,India

Abstract

Theoverallaimofthesoftwareindustryistoensuredeliveryofhighqualitysoftwaretotheenduser.Toensurehighqualitysoftware,itisrequiredtotestsoftware.Testingensuresthatsoftwaremeetsuserspecificationsandrequirements.However,thefieldofsoftwaretestinghasanumberofunderlyingissueslikeeffectivegenerationoftestcases,prioritisationoftestcasesetcwhichneedtobetackled.Theseissuesdemandoneffort,timeandcostofthetesting.Differenttechniquesandmethodologieshavebeenproposedfortakingcareoftheseissues.Useofevolutionaryalgorithmsforautomatictestgenerationhasbeenanareaofinterestformanyresearchers.GeneticAlgorithm（GA）isonesuchformofevolutionaryalgorithms.Inthisresearchpaper,wepresentasurveyofGAapproachforaddressingthevariousissuesencounteredduringsoftwaretesting.

Keywords:

Softwaretesting,GeneticAlgorithm

1.Introduction

Testingisprimarilydoneonsoftwareaswellasinwebfortestingclientandserverarchitecture.Softwaretestingisoneofthemajorandprimarytechniquesforachievinghighqualitysoftware.Softwaretestingisdonetodetectpresenceoffaults,whichcausesoftwarefailure.However,softwaretestingisatimeconsumingandexpensivetask[29],[20],[28].Itconsumesalmost50%ofthesoftwaresystemdevelopmentresources[3],[20].Softwaretestingcanalsobedefinedasprocessofverifyingandvalidatingsoftwaretoensurethatsoftwaremeetsthetechnicalaswellasbusinessrequirementsasexpected[16].

Verificationisdonetoensurethatthesoftwaremeetsspecificationandisclosetostructuraltestingwhereasvalidationisclosetothefunctionaltestingandisdone

byexecutingsoftwareundertest（SUT）[18].Broadly,testingtechniquesincludefunctional（blackbox）andstructural（whitebox）testing.Functionaltestingisbased

onfunctionalrequirementswhereasstructuraltestingisdoneoncodeitself[13][10][24].Grayboxtestingishybridofwhiteboxtestingandblackboxtesting[8].

Testingcanbedoneeithermanuallyorautomaticallybyusingtestingtools.Itisfoundthatautomatedsoftwaretestingisbetterthanmanualtesting.However,veryfewtestdatagenerationtoolsarecommerciallyavailabletoday[14].Varioustechniqueshavebeenproposedforgeneratingtestdataortestcasesautomatically.Recently,lotofworkisbeingdonefortestcasesgenerationusingsoftcomputingtechniqueslikefuzzylogic,neuralnetworks,GA,geneticprogrammingandevolutionarycomputationprovidingkeystotheproblemareasofsoftwaretesting.

Evolutionarytestingisanemergingmethodologyforautomaticallyproducinghighqualitytestdata[10].GAiswellknownformoftheevolutionaryalgorithmsconceivedbyJohnHollandinUnitedStatesduringlatesixties[6][25].In[21],evolutionaryblackboxtestingisalsoappliedonembeddedsystemstotestitsfunctionalandnon-functionalproperties.GAhasbeenappliedinmanyoptimizationproblemsforgeneratingtestplansforfunctionalitytesting,feasibletestcasesandinmanyotherareas[5][15].GAhasalsobeenusedinmodelbasedtestcasegeneration[3][23][26][27].Inobjectorientedunittestingaswellasintheblackboxtesting,GAisusedforautomaticgenerationoftestcases[23],[10],[15].Concerningtestingofwebapplications,manytools,newtechniquesandmethodshavebeendevelopedtoaddressissueslikemaintainability,testability,security,performance,correctnessandreliabilityofwebapplication[8].Webapplicationsarecomposedofwebpagesandcomponentsandinteractionbetweenthemexecuteswebservers,HTTP,browser（theclientside）andnetworks.Awebpageisinformationviewedontheclientsideinasinglebrowserwindow[16].In[30],usersessiondataofwebapplicationisusedtogeneratetestcasesbyapplyingGA.

Inthisresearchpaper,asurveyofdifferentsoftwaretestingtechniqueswhereGAisefficientlyusedispresented.Thispaperisdividedinto4sections.Section2describesbrieflytheworkingofaGA.Insection3,applicationsofGAindifferenttypesofsoftwaretestingisdescribed.Section4concludesthepaperandgivesanoverviewofourfuturework.

2.GeneticAlgorithm:

ABriefIntroduction

Inthepast,evolutionaryalgorithmshavebeenappliedinmanyreallifeproblems.GAisonesuchevolutionary

algorithm.GAhasemergedasapractical,robustoptimizationtechniqueandsearchmethod.AGAisasearchalgorithmthatisinspiredbythewaynatureevolvesspeciesusingnaturalselectionofthefittestindividuals.

Thepossiblesolutionstoproblembeingsolvedarerepresentedbyapopulationofchromosomes.Achromosomeisastringofbinarydigitsandeachdigitthatmakesupachromosomeiscalledagene.Thisinitialpopulationcanbetotallyrandomorcanbecreatedmanuallyusingprocessessuchasgreedyalgorithm.ThepseudocodeofabasicalgorithmforGAisasfollows[6]:

-

Initialize（population）

Evaluate（population）

While（stoppingconditionnotsatisfied）

{

Selection（population）

Crossover（population）

Mutate（population）

}

GAusesthreeoperatorsonitspopulationwhicharedescribedbelow:

ΣSelection:

Aselectionschemeisappliedtodeterminehowindividualsarechosenformatingbasedontheirfitness.Fitnesscanbedefinedasacapabilityofanindividualtosurviveandreproduceinanenvironment.Selectiongeneratesthenewpopulationfromtheoldone,thusstartinganewgeneration.Eachchromosomeisevaluatedinpresentgenerationtodetermineitsfitnessvalue.Thisfitnessvalueisusedtoselectthebetterchromosomesfromthepopulationforthenextgeneration.

ΣCrossoverorRecombination:

Afterselection,thecrossoveroperationisappliedtotheselectedchromosomes.Itinvolvesswappingofgenesorsequenceofbitsinthestringbetweentwoindividuals.Thisprocessisrepeatedwithdifferentparentindividualsuntilthenextgenerationhasenoughindividuals.Aftercrossover,themutationoperatorisappliedtoarandomlyselectedsubsetofthepopulation.

ΣMutation:

Mutationalterschromosomesinsmallwaystointroducenewgoodtraits.Itisappliedtobringdiversityinthepopulation.

3.UsingGeneticAlgorithminSoftware

Testing

InthissectionwewilldiscussindetailtheapplicationsofGAindifferentareasoftestingliketestplanning[5],minimizationoftestcasesinregressiontesting[11],modelbasedtesting[3][23][26][27]andwebtesting[30].

3.1ApplicationsofGAinWhiteBoxTesting

Structuraltestingcanbedoneintheformofdataflowtestingorpathtesting.Pathtestinginvolvesgeneratingasetofpathsthatwillcovereverybranchintheprogramandfindingthesetoftestcasesthatwillexecuteeverypathinthissetofprogrampath[16][18].Indataflowtesting,thefocusisonthepointsatwhichvariablesreceivevaluesandthepointsatwhichthesevaluesareused[2].Next,wewilldiscussbrieflysomeoftheresearchworkregardingtheapplicationsofGAinwhiteboxtesting.

3.1.1DataFlowTesting

M.R.Girgis

Girgis[7]hasproposedastructuralorientedautomatictestdatagenerationtechniquethatusesaGAguidedbythedataflowdependenciesintheprogramtofulfiltheall-usescriterion.TheprogramtobetestedisconvertedintoaControlFlowGraph（CFG）whereeachnoderepresentsablockinaprogramandtheedgesoftheflowgraphdepictthecontrolflowofthestatements.

Variablesinaprogramundertestaredividedinto'

cuses'

and'

p-uses’variables.c-usesvariablesarethosewhichareusedincomputationsorasapredicatesinaprogramwhereasp-usesvariablesareassociatedwithedgesoftheflowgraph.Inordertofulfiltheall-usescriteria,thedef-clearpath（apathcontainingnonewdefinitionofacurrentvariable）fromeachdefinitionofavariabletoeachuseofthatvariableneedtobedetermined.Tofindoutthesetofpathssatisfyingallusescriteria,itisnecessarytodeterminedefc-use（dcu）anddefp-use（dpu）ofavariablei.e.thedef-clearpathstotheirc-useatnodeianddef-clearpathstotheirp-useatedge（i,j）.

Usingthelocationofavariabledefsandusesinaprogramundertest,combinedwiththe‘Basicstatereachalgorithm’,thesetsdcu（i）anddpu（i,j）aredetermined.Fromthe‘Basicstatereachalgorithm’twosetsreach（i）andavail（i）aredeterminedwherereach（i）isthesetofallvariabledefsthat“reach”nodeiandsetavail（i）isthesetofall“available”variabledefsatnodeii.e.Theunionofsetofglobaldefsatnodeiandthesetofalldefsthatreachthisnode.

dcu（i）:

reach（i）«

c-use（i）

（1）

dpu（i,j）:

avail（i）«

p-use（i,j）

Listofalldcuanddpusetsintheprocedurecallssatisfyingtheall-usescriterionaredeterminedalongwithkillingnodes（nodescontainingotherdefinitionofavariableinacurrentpath）thatmustnotbeincludedinthecurrentpath.Inthisapproach,GAacceptsinstrumentedversionoftheprogramundertest,thelistofdef-usesetstobecovered,thenumberofinputvariables,andthedomainandtheprecisionofeachinputvariablesasaninput.Abinaryvectorisusedtorepresentachromosome.Thelengthoftheinputisdeterminedbythedomainandtheprecision.ThedomainisrepresentedbyDi=[ai,bi]whereeachvariableinaprogramtakesvaluesfromtherange[ai,bi].EachdomainDishouldbecutintoequalsizerangesifdecimalplacesdiisdesiredforavariab