数据库应用程序测试中英文对照外文翻译文献.docx

数据库应用程序测试中英文对照外文翻译文献.docx

《数据库应用程序测试中英文对照外文翻译文献.docx》由会员分享，可在线阅读，更多相关《数据库应用程序测试中英文对照外文翻译文献.docx（7页珍藏版）》请在冰豆网上搜索。

数据库应用程序测试中英文对照外文翻译文献

中英文翻译

DatabaseApplicationTesting

1.Introduction

Databasesplayapivotalroleinalmosteveryorganizationintoday’sinformation-basedsociety.CommercialDatabasemanagementsystems（DBMSs）provideorganizationswithefficientaccesstohugeamountsofdatawithoutaffectingtheintegrityofdataandrelievingtheuseroftheanyneedtounderstandthelow-levelimplementationdetails.OvertheyearstremendouseffortshavebeendevotedtoensuringuseofefficientandintegrityprotectingdatastructuresandalgorithmsbyDBMSs.However,littlehasbeendonetodevelopsystematictechniquesforensuringcorrectnessofapplicationsusingtheseDBMSs.Manytestingtechniqueshavebeendevelopedtohelpensurethatbehaviourofaprogramisinaccordancewiththespecifications.However,thesetechniquesmostlytargetprogramswrittenintraditionalimperativelanguagesandcan’tbeofmuchhelpwhenitcomestodatabaseapplications.Likeanyotherprogram,databaseapplicationprogramcanbeviewedasanattempttoimplementafunction.Consideredthisway,boththeinputandoutputspacesofthisfunctionwillincludedatabasestateapartfromtheexplicitinputandoutputparametersoftheapplication.Thisaffectssubstantiallythewayatestcaseisdefined,generatedandexecutedtocheckcorrectnessofapplication.Hencethereisaneedfornewapproachesspecificallyorientedtowardstestingdatabaseapplications.

Testingdatabaseapplicationprogramsinvolvesthefollowingphases:

•Extractionofinformationfromdatabaseschema

•GenerationoftestdataandPopulatingtestdatabase

•Generationoftestcasesasinputtotheapplicationprogram

•Validationofdatabasestateandoutputafterexecution

Usinglivedatahasseverallimitations.Itmaynotreflectsufficientlywidevarietyofpossiblesituationsandevenifitdoes,itmightbedifficulttofindtheminalargedatabase.Secondly,privacyorsecurityconstraintsmightpreventtheuserfromseeingsensitivedata.Hence,variousmethodsforgeneratingsynthetictestdatahavebeenproposed.Whengeneratingdataandpopulatingthetestdatabase,itsimportanttogeneratevalidandinterestingdatae.g.itwouldbeadvisabletoselectdatasoastoincludesituationswhichthetesterbelievesarelikelytooccurorwillexposefaultsinapplication.Thetechniqueusedfortestdatagenerationwilldeterminetheextentofcoverageoftestdatabase.Selectingagoodinitialdatabasestatesoastoincludeawidevarietyofscenariosresemblingrealdatafortheparticularapplicationisverybeneficial.Sincedatabasestateplaysanimportantroleindeterminingtheoutput,ithastobecheckedaftereachexecutionthatonlythespecifiedmodificationsandnoneothershaveoccurred.

2.AGENDA-toolsetfortestingDBapplications

AGENDAisatoolsethasbeendesigned.AGENDAtakesasinputtheapplicationdatabaseschema,applicationsourcecodeandfilescontainingsamplevalueswhichcontainsuggestedvaluesfortheattributesprovidedbytheuser.Theuserinteractivelyselectstestheuristicsandprovidesinformationaboutexpectedbehaviouroftestcases.UsingthisinformationAGENDA,populatesthedatabase,generatesinputstotheapplication,executestheapplicationonthoseinputsandcheckssomeaspectsofcorrectnessoftheresultingdatabasestateandapplicationoutput.1

3.InputGenerator:

ItgeneratestheinputdatatobesuppliedtotheapplicationbyusinginformationderivedfromAgendaparserandStategeneratorinadditiontotheinformationgainedbyparsingtheSQLstatementsintheapplicationprogramandinformationusefulforcheckingtestresults.Informationderivedfromparsingthesourcecodemaybeusefulinsuggestinginputsthattestershouldsupplytotheapplication.Theinputgeneratorthusgeneratestestinputsbyinstantiatingtheinputparameterswithactualparameters.

4.StateValidator:

ThevalidatormonitorsthechangeinapplicationDBstateduringexecutionofatest.Itautomaticallylogsthechangesintheapplicationtablesandsemi-automaticallychecksthestatechange.

5.OutputValidator:

Itcapturestheapplication’soutputsandchecksthemagainstthequerypreconditionsandpostconditionsthathavebeengeneratedbythetoolorsuppliedbythetester.

6.DesignandImplementation

6.1Parsingtool

TheAgendaParsingtoolisbasedonPostgresSQLparser.PostgresSQLparsercreatesanAbstractSyntaxTreecontainingrelevantinformationabouttables,attributesandconstraintsfromagivenschema.However,thisinformationisspreadoutatdifferentlocationsinthetree.Inaddition,itispossibletohavedifferenttreestructureshavingthesameunderlyinginformationaboutthetables,becauseofuseofdifferentSQLDDLsyntacticconstructsexpressingthesameinformation.Consequently,theexactlocationofrelevantinformationdependsontheexactsyntaxofschemadefinition.SomeoftheinformationfromDBMS’sinternalcatalogtablesisneededbyothercomponentsofAGENDA.AllowingthemtodirectlyquerythesetableswouldhaveintroducedinterdependencybetweenAGENDAcomponentswhichisnotdesirable.HencealltheinformationthatneedstobeprocessedisstoredinAgendaDBwhichismadeavailabletoothercomponents.ThisdecouplingofPostgresSQLfromrestofthecomponentsallowsAGENDAtobeportedtodifferentDBMSjustbychangingtheParser.TheParserextractsinformationaboutintegrityconstraintssuchasuniquenessconstraints,referentialconstraintsandnotNULLconstraintsfromschema.Italsoextractslimitedinformationfromsemanticconstraints,particularlyboundaryvalues.Thisisveryusefulinautomaticdata-partitioningandinputgeneration.Next,theAgendaParserparsesthesample-valuefilescontaininguser-supplieddataandstoresthesamplevalues,theirdatagroupsandassociatedattributesintheAgendaDB.Attributesinvolvedincompositeconstraintsaremarkedsothattheycanbecorrectlyhandledbyinputgenerator.

7ExtensionstoAGENDA

7.1TestingWebDBapplications

WiththetremendousgrowthofWorldWideWeb,manynewweb-basedservicesthataredrivenbydatastoredindatabasesaregainingimportance.ExamplesincludeE-commerceapplicationssuchasonlinestores,andbusiness-to-businesssupportproducts.Someoftheseareofcriticalimportanceandhenceitisessentialtoensuretheircorrectfunctioning.MostwebDBapplicationsconsistofthreeslayers-atthebaseisDBMSandadatabase,atthetopisclientwebbrowserandinbetweenliestheapplicationlogic-usuallydevelopedwithaserver-sidescriptinglanguageorJavaextendedwithlibrarythatcaninterfacewithDBMSs,andcandecodeandproduceHTMLpagesdisplayedintheclientbrowser.Forawebapplicationatestcaseisconsideredasasequenceofpagestobevisitedalongwiththeinputvaluestobeprovidedtothepagescontainingforms.Thewhiteboxapproachinvolvesfollowingsteps:

1.Informationextractionfromapplicationsource:

UsefulinformationsuchasURLlinks（whichincludesallotherURLsthatcanbereachedfromthecurrentpage）andparameterinformation（name-valuepairsthatarepassedtotheServlet）foreachURLisextractedfromapplicationsource.URLsarepartitionedintotwocategoriesdependingontheircontent-staticanddata-based（dynamic）page.

2.Webapplicationgraphgenerationandpathselection:

Basedontheinformationextractedearlier,anapplicationgraph,whereeachnoderepresentsaURLandedgesrepresentURLlinks,isgeneratedandthensimplifiedaccordingtoURLlinktypes.ThereisanedgefromURLAtoURLBifURLAproducesalinktoURLBintheHTMLpageitgenerates.PathsthroughthegraphrepresentnaturalsequencesofexecutionofURLsasausernavigatesthroughthewebapplication.Hence,someofthesepathsareselectedastestcasestorepresentpossiblescenariosofuseoftheapplication.

3.InputGeneration:

Foreachpathselected,AGENDAisusedtogenerateinputsforeachURL.Thepathalongwithinputsconstituteatestcase.AnXMLfileisgeneratedcorrespondingtoeachsuchtestcase.

4.TestExecution:

TheXMLfileisparsedusingXMLparsertoextractURLinformationandthetestcaseisexecutedautomaticallyusingopensourceJakartaHttpClientintegratedwithAGENDA.Afterexecutionofeachupdateorinsertion,AGENDAchecksthenewdatabasestates.Outputpagesarecheckedbymanualinspectionorothertools.ThetoolinitscurrentformistargetedtotheJavaServletmodel,usingJDBCfordatabaseaccess,andmakessomeassumptionsaboutprogrammingstyle.However,thebasictechniquecanbeappliedtomoregeneralservletstylesandotherwebapplicationlanguages.

8.RegressionTestsForDatabaseApplications

Anyapplicationisconstantlygoingthroughtheprocessofevolutionsuchasitscomponentsgettingreplacedwithmorepowerfulcomponents,variousoptimizationsbeingincorporatedandsoon.Wheneversuchmodificationisintroducedinanapplication,itisimportanttocheckfortheintegrityoftheapplicationandthatisthepurposeofregressiontests.Therearevarioustoolsbuiltforautomatingtheregressiontestingprocedure,mostpopularbeingJUnitframeworkdevelopedforcarryingoutregressiontestsforJavaapplications.Databaseapplicationswhicharecomposedofmanylayersandstackedinvariouslayersare,inparticular,subjecttoconstantchangeforinstancere-engineeringofbusinessprocesses,authorizationrulesbeingchangedetc.Changingdatabaseapplicationsisverycostlyandinvolvesgreatdealofmanualworksincetherearen’tanytoolsavailablethatcanautomaticallycarryingoutregressiontestsonthem.

9.ConclusionsandFutureWork

TheAGENDAtoolsetwasdesignedandimplementedinrespo