軟件工程教學(xué)課件：Chapter-12-Review Techniques

TheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.1Chapter12ReviewTechniquesTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.2Reviews...thereisnoparticularreasonwhyyourfriendandcolleaguecannotalsobeyoursternestcritic.JerryWeinbergTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.3WhatAreReviews?ameetingconductedbytechnicalpeoplefortechnicalpeopleatechnicalassessmentofaworkproductcreatedduringthesoftwareengineeringprocessasoftwarequalityassurancemechanismatraininggroundTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.4WhatReviewsAreNotAprojectsummaryorprogressassessmentAmeetingintendedsolelytoimpartinformationAmechanismforpoliticalorpersonalreprisal!TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.5WhatDoWeLookFor?ErrorsanddefectsError—aqualityproblemfoundbeforethesoftwareisreleasedtoendusersDefect—aqualityproblemfoundonlyafterthesoftwarehasbeenreleasedtoend-usersWemakethisdistinctionbecauseerrorsanddefectshaveverydifferenteconomic,business,psychological,andhumanimpactHowever,thetemporaldistinctionmadebetweenerrorsanddefectsinthisbookisnotmainstreamthinkingTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.6DefectAmplificationAdefectamplificationmodelcanbeusedtoillustratethegenerationanddetectionoferrorsduringthedesignandcodegenerationactionsofasoftwareprocess.ErrorspassedthroughAmplifiederrors1:xNewlygeneratederrorsDevelopmentstepErrorsfromPreviousstepErrorspassedTonextstepDefectsDetectionPercentEfficiencyTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.7DefectAmplificationIntheexampleprovidedinSEPA,asoftwareprocessthatdoesNOTincludereviews,yields94errorsatthebeginningoftestingandReleases12latentdefectstothefieldasoftwareprocessthatdoesincludereviews,yields24errorsatthebeginningoftestingandreleases3latentdefectstothefieldAcostanalysisindicatesthattheprocesswithNOreviewscostsapproximately3timesmorethantheprocesswithreviews,takingthecostofcorrectingthelatentdefectsintoaccountTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.8MetricsThetotalrevieweffortandthetotalnumberoferrorsdiscoveredaredefinedas:Ereview=Ep

+Ea+Er

Errtot=Errminor+ErrmajorDefectdensityrepresentstheerrorsfoundperunitofworkproductreviewed.Defectdensity=Errtot/WPSwhere…TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.9MetricsPreparationeffort,Ep—theeffort(inperson-hours)requiredtoreviewaworkproductpriortotheactualreviewmeetingAssessmenteffort,Ea—theeffort(inperson-hours)thatisexpendingduringtheactualreviewReworkeffort,Er—theeffort(inperson-hours)thatisdedicatedtothecorrectionofthoseerrorsuncoveredduringthereviewWorkproductsize,WPS—ameasureofthesizeoftheworkproductthathasbeenreviewed(e.g.,thenumberofUMLmodels,orthenumberofdocumentpages,orthenumberoflinesofcode)Minorerrorsfound,Errminor—thenumberoferrorsfoundthatcanbecategorizedasminor(requiringlessthansomepre-specifiedefforttocorrect)Majorerrorsfound,Errmajor—thenumberoferrorsfoundthatcanbecategorizedasmajor(requiringmorethansomepre-specifiedefforttocorrect)TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.10AnExample—IIfpasthistoryindicatesthattheaveragedefectdensityforarequirementsmodelis0.6errorsperpage,andanewrequirementmodelis32pageslong,aroughestimatesuggeststhatyoursoftwareteamwillfindabout19or20errorsduringthereviewofthedocument.Ifyoufindonly6errors,you’vedoneanextremelygoodjobindevelopingtherequirementsmodeloryourreviewapproachwasnotthoroughenough.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.11AnExample—IITheeffortrequiredtocorrectaminormodelerror(immediatelyafterthereview)wasfoundtorequire4person-hours.Theeffortrequiredforamajorrequirementerrorwasfoundtobe18person-hours.Examiningthereviewdatacollected,youfindthatminorerrorsoccurabout6timesmorefrequentlythanmajorerrors.Therefore,youcanestimatethattheaverageefforttofindandcorrectarequirementserrorduringreviewisabout6person-hours.Requirementsrelatederrorsuncoveredduringtestingrequireanaverageof45person-hourstofindandcorrect.Usingtheaveragesnoted,weget:Effortsavedpererror= Etesting–Ereviews

45–6=30person-hours/errorSince22errorswerefoundduringthereviewoftherequirementsmodel,asavingofabout660person-hoursoftestingeffortwouldbeachieved.Andthat’sjustforrequirements-relatederrors.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.12OverallEffortexpendedwithandwithoutreviewswithreviewsTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.13ReferenceModelTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.14InformalReviewsInformalreviewsinclude:asimpledeskcheckofasoftwareengineeringworkproductwithacolleagueacasualmeeting(involvingmorethan2people)forthepurposeofreviewingaworkproduct,orthereview-orientedaspectsofpairprogrammingpairprogrammingencouragescontinuousreviewasaworkproduct(designorcode)iscreated.Thebenefitisimmediatediscoveryoferrorsandbetterworkproductqualityasaconsequence.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.15FormalTechnicalReviewsTheobjectivesofanFTRare:touncovererrorsinfunction,logic,orimplementationforanyrepresentationofthesoftwaretoverifythatthesoftwareunderreviewmeetsitsrequirementstoensurethatthesoftwarehasbeenrepresentedaccordingtopredefinedstandardstoachievesoftwarethatisdevelopedinauniformmannertomakeprojectsmoremanageableTheFTRisactuallyaclassofreviewsthatincludeswalkthroughsandinspections.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.16TheReviewMeetingBetweenthreeandfivepeople(typically)shouldbeinvolvedinthereview.Advancepreparationshouldoccurbutshouldrequirenomorethantwohoursofworkforeachperson.Thedurationofthereviewmeetingshouldbelessthantwohours.Focusisonaworkproduct(e.g.,aportionofarequirementsmodel,adetailedcomponentdesign,sourcecodeforacomponent)TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.17ThePlayersreviewleaderproducerrecorderreviewerstandardsbearer(SQA)maintenanceoracleuserrepTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.18ThePlayersProducer—theindividualwhohasdevelopedtheworkproductinformstheprojectleaderthattheworkproductiscompleteandthatareviewisrequiredReviewleader—evaluatestheproductforreadiness,generatescopiesofproductmaterials,anddistributesthemtotwoorthreereviewersforadvancepreparation.Reviewer(s)—expectedtospendbetweenoneandtwohoursreviewingtheproduct,makingnotes,andotherwisebecomingfamiliarwiththework.Recorder—reviewerwhorecords(inwriting)allimportantissuesraisedduringthereview.

TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.19ConductingtheReviewReviewtheproduct,nottheproducer.Setanagendaandmaintainit.

Limitdebateandrebuttal.Enunciateproblemareas,butdon'tattempttosolveeveryproblemnoted.Takewrittennotes.Limitthenumberofparticipantsandinsistuponadvancepreparation.Developachecklistforeachproductthatislikelytobereviewed.

AllocateresourcesandscheduletimeforFTRs.

Conductmeaningfultrainingforallreviewers.

Reviewyourearlyreviews.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.20ReviewOptionsMatrixtrainedleaderagendaestablishedreviewersprepareinadvanceproducerpresentsproduct“reader”presentsproductrecordertakesnoteschecklistsusedtofinderrorserrorscategorizedasfoundissueslistcreatedteammustsign-offonresultIPR—informalpeerreviewWT—WalkthroughIN—InspectionRRR—roundrobinreviewIPRWTINRRRnomaybemaybemaybenomaybenonononoyesyesyesyesnoyesnonoyesyesyesyesyesnoyesyesyesyesyesyesyesyesyesnonoyesnonoyesmaybe**TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.21Sample-DrivenReviews(SDRs)SDRsattempttoquantifythoseworkproductsthatareprimarytargetsforfullFTRs.Toaccomplishthis…

Inspectafractionaiofeachsoftwareworkproduct,i.Recordthenumberoffaults,fifoundwithinai.Developagrossestimateofthenumberoffaultswithinworkproductibymultiplyingfiby1/ai.Sorttheworkproductsindescendingorderaccordingtothegrossestimateofthenumberoffaultsineach.Focusavailablereviewresourcesonthoseworkproductsthathavethehighestestimatednumberoffaults.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.22MetricsDerivedfromReviewsinspectiontimeperpageofdocumentationinspectiontimeperKLOCorFPerrorsuncoveredperreviewerhourerrorsuncoveredperpreparationhourerrorsuncoveredperSEtask(e.g.,design)numberofminorerrors(e.g.,typos)numberoferrorsfoundduringpreparationnumberofmajorerrors(e.g.,nonconformancetoreq.)inspectioneffortperKLOCorFPTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.23Chapter12ReviewTechniquesTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.24Reviews...thereisnoparticularreasonwhyyourfriendandcolleaguecannotalsobeyoursternestcritic.JerryWeinbergTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.25WhatAreReviews?ameetingconductedbytechnicalpeoplefortechnicalpeopleatechnicalassessmentofaworkproductcreatedduringthesoftwareengineeringprocessasoftwarequalityassurancemechanismatraininggroundTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.26WhatReviewsAreNotAprojectsummaryorprogressassessmentAmeetingintendedsolelytoimpartinformationAmechanismforpoliticalorpersonalreprisal!TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.27WhatDoWeLookFor?ErrorsanddefectsError—aqualityproblemfoundbeforethesoftwareisreleasedtoendusersDefect—aqualityproblemfoundonlyafterthesoftwarehasbeenreleasedtoend-usersWemakethisdistinctionbecauseerrorsanddefectshaveverydifferenteconomic,business,psychological,andhumanimpactHowever,thetemporaldistinctionmadebetweenerrorsanddefectsinthisbookisnotmainstreamthinkingTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.28DefectAmplificationAdefectamplificationmodelcanbeusedtoillustratethegenerationanddetectionoferrorsduringthedesignandcodegenerationactionsofasoftwareprocess.ErrorspassedthroughAmplifiederrors1:xNewlygeneratederrorsDevelopmentstepErrorsfromPreviousstepErrorspassedTonextstepDefectsDetectionPercentEfficiencyTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.29DefectAmplificationIntheexampleprovidedinSEPA,asoftwareprocessthatdoesNOTincludereviews,yields94errorsatthebeginningoftestingandReleases12latentdefectstothefieldasoftwareprocessthatdoesincludereviews,yields24errorsatthebeginningoftestingandreleases3latentdefectstothefieldAcostanalysisindicatesthattheprocesswithNOreviewscostsapproximately3timesmorethantheprocesswithreviews,takingthecostofcorrectingthelatentdefectsintoaccountTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.30MetricsThetotalrevieweffortandthetotalnumberoferrorsdiscoveredaredefinedas:Ereview=Ep

+Ea+Er

Errtot=Errminor+ErrmajorDefectdensityrepresentstheerrorsfoundperunitofworkproductreviewed.Defectdensity=Errtot/WPSwhere…TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.31MetricsPreparationeffort,Ep—theeffort(inperson-hours)requiredtoreviewaworkproductpriortotheactualreviewmeetingAssessmenteffort,Ea—theeffort(inperson-hours)thatisexpendingduringtheactualreviewReworkeffort,Er—theeffort(inperson-hours)thatisdedicatedtothecorrectionofthoseerrorsuncoveredduringthereviewWorkproductsize,WPS—ameasureofthesizeoftheworkproductthathasbeenreviewed(e.g.,thenumberofUMLmodels,orthenumberofdocumentpages,orthenumberoflinesofcode)Minorerrorsfound,Errminor—thenumberoferrorsfoundthatcanbecategorizedasminor(requiringlessthansomepre-specifiedefforttocorrect)Majorerrorsfound,Errmajor—thenumberoferrorsfoundthatcanbecategorizedasmajor(requiringmorethansomepre-specifiedefforttocorrect)TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.32AnExample—IIfpasthistoryindicatesthattheaveragedefectdensityforarequirementsmodelis0.6errorsperpage,andanewrequirementmodelis32pageslong,aroughestimatesuggeststhatyoursoftwareteamwillfindabout19or20errorsduringthereviewofthedocument.Ifyoufindonly6errors,you’vedoneanextremelygoodjobindevelopingtherequirementsmodeloryourreviewapproachwasnotthoroughenough.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.33AnExample—IITheeffortrequiredtocorrectaminormodelerror(immediatelyafterthereview)wasfoundtorequire4person-hours.Theeffortrequiredforamajorrequirementerrorwasfoundtobe18person-hours.Examiningthereviewdatacollected,youfindthatminorerrorsoccurabout6timesmorefrequentlythanmajorerrors.Therefore,youcanestimatethattheaverageefforttofindandcorrectarequirementserrorduringreviewisabout6person-hours.Requirementsrelatederrorsuncoveredduringtestingrequireanaverageof45person-hourstofindandcorrect.Usingtheaveragesnoted,weget:Effortsavedpererror= Etesting–Ereviews

45–6=30person-hours/errorSince22errorswerefoundduringthereviewoftherequirementsmodel,asavingofabout660person-hoursoftestingeffortwouldbeachieved.Andthat’sjustforrequirements-relatederrors.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.34OverallEffortexpendedwithandwithoutreviewswithreviewsTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.35ReferenceModelTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.36InformalReviewsInformalreviewsinclude:asimpledeskcheckofasoftwareengineeringworkproductwithacolleagueacasualmeeting(involvingmorethan2people)forthepurposeofreviewingaworkproduct,orthereview-orientedaspectsofpairprogrammingpairprogrammingencouragescontinuousreviewasaworkproduct(designorcode)iscreated.Thebenefitisimmediatediscoveryoferrorsandbetterworkproductqualityasaconsequence.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.37FormalTechnicalReviewsTheobjectivesofanFTRare:touncovererrorsinfunction,logic,orimplementationforanyrepresentationofthesoftwaretoverifythatthesoftwareunderreviewmeetsitsrequirementstoensurethatthesoftwarehasbeenrepresentedaccordingtopredefinedstandardstoachievesoftwarethatisdevelopedinauniformmannertomakeprojectsmoremanageableTheFTRisactuallyaclassofreviewsthatincludeswalkthroughsandinspections.TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.38TheReviewMeetingBetweenthreeandfivepeople(typically)shouldbeinvolvedinthereview.Advancepreparationshouldoccurbutshouldrequirenomorethantwohoursofworkforeachperson.Thedurationofthereviewmeetingshouldbelessthantwohours.Focusisonaworkproduct(e.g.,aportionofarequirementsmodel,adetailedcomponentdesign,sourcecodeforacomponent)TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.39ThePlayersreviewleaderproducerrecorderreviewerstandardsbearer(SQA)maintenanceoracleuserrepTheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.40ThePlayersProducer—theindividualwhohasdevelopedtheworkproductinformstheprojectleaderthattheworkproductiscompleteandthatareviewisrequiredReviewleader—evaluatestheproductforreadiness,generatescopiesofproductmaterials,anddistributesthemtotwoorthreereviewersforadvancepreparation.Reviewer(s)—expectedtospendbetweenoneandtwohoursreviewingtheproduct,makingnotes,andotherwisebecomingfamiliarwiththework.Recorder—reviewerwhorecords(inwriting)allimportantissuesraisedduringthereview.

TheseslidesaredesignedtoaTheseslidesaredesignedtoaccompanySoftwareEngineering:APractitioner’sApproach,7/e(McGraw-Hill2009).Slidescopyright2009byRogerPressman.41ConductingtheReviewReviewtheproduct,nottheproducer.