A Relational Model of Data for Large Shared Data Banks.docx
- 文档编号:2744554
- 上传时间:2023-05-04
- 格式:DOCX
- 页数:16
- 大小:27.14KB
A Relational Model of Data for Large Shared Data Banks.docx
《A Relational Model of Data for Large Shared Data Banks.docx》由会员分享,可在线阅读,更多相关《A Relational Model of Data for Large Shared Data Banks.docx(16页珍藏版)》请在冰点文库上搜索。
ARelationalModelofDataforLargeSharedDataBanks
ARelationalModelofDataforLargeSharedDataBanks
---E.F.Codd
ReprintedfromCommunicationsoftheACM,Vol.13,No.6,June1970,pp.377-387.Copyright©1970,AssociationforComputingMachinery,Inc.
Abstract
Futureusersoflargedatabanksmustbeprotectedfromhavingtoknowhowthedataisorganizedinthemachine(theinternalrepresentation).Apromptingservicewhichsuppliessuchinformationisnotasatisfactorysolution.Activitiesofusersatterminalsandmostapplicationprogramsshouldremainunaffectedwhentheinternalrepresentationofdataischangedandevenwhensomeaspectsoftheexternalrepresentationarechanged.Changesindatarepresentationwilloftenbeneededas'aresultofchangesinquery,update,andreporttrafficandnaturalgrowthinthetypesofstoredinformation.
Existingnoninferential,formatteddatasystemsprovideuserswithtree-structuredfilesorslightlymoregeneralnetworkmodelsofthedata.InSection1,inadequaciesofthesemodelsarediscussed.Amodelbasedonn-aryrelations,anormalformfordatabaserelations,andtheconceptofauniversaldatasublanguageareintroduced.InSection2,certainoperationsonrelations(otherthanlogicalinference)arediscussedandappliedtotheproblemsofredundancyandconsistencyintheuser'smodel.
KeyWordsandPhrases
databank,database,datastructure,dataorganization;,hierarchiesofdata,networkofdata,relations,derivability,redundancy,consistency,composition,join,retrievallanguage,predicatecalculus,security,dataintegrity
CRCategories
3.70,3.73,3.75,420,4.22,4.29
∙RelationalModelandNormalForm
∙Introduction
∙DataHYPERLINK"http:
//www.acm.org/classics/nov95/s1p2.html"DependenciesinPresentSystems
∙OrderingDependence
∙IndexingDependence
∙ccessPathDependence
∙AHYPERLINK"http:
//www.acm.org/classics/nov95/s1p3.html"RelationalHYPERLINK"http:
//www.acm.org/classics/nov95/s1p3.html"ViewofHYPERLINK"http:
//www.acm.org/classics/nov95/s1p3.html"Data
∙NormalForm
∙SomeLinguisticAspects
∙Expressible,Names,andStoredRelations
∙RedundancyandConsistency(comingsoon)
∙References
∙HYPERLINK"http:
//www.acm.org/classics/nov95/toc.html"RelationalHYPERLINK"http:
//www.acm.org/classics/nov95/toc.html"HYPERLINK"http:
//www.acm.org/classics/nov95/toc.html"ModelHYPERLINK"http:
//www.acm.org/classics/nov95/toc.html"andNormalForm
1.1Introduction
Thispaperisconcernedwiththeapplicationofelementaryrelationtheorytosystemswhichprovidesharedaccesstolargebanksofformatteddata.ExceptforapaperbyChilds[1],theprincipalapplicationofrelationstodatasystemshasbeentodeductivequestion-answeringsystems.LeveinandMaron[2]providenumerousreferencestoworkinthisarea.
Incontrast,theproblemstreatedherearethoseofdataindependence-theindependenceofapplicationprogramsandterminalactivitiesfromgrowthindatatypesandchangesindatarepresentationÑandcertainkindsofdatainconsistencywhichareexpectedtobecometroublesomeeveninnondeductivesystems.
Therelationalview(ormodel)ofdatadescribedinSection1appearstobesuperiorinseveralrespectstothegraphornetworkmodel[3,4]presentlyinvoguefornon-inferentialsystems.Itprovidesameansofdescribingdatawithitsnaturalstructureonly--thatis,withoutsuperimposinganyadditionalstructureformachinerepresentationposes.Accordingly,itprovidesabasisforahighleveldatalanguagewhichwillyieldmaximalindependencebetweenprogramsontheonehandandmachinerepresentationandorganizationofdataontheother.
Afurtheradvantageoftherelationalviewisthatitformsasoundbasisfortreatingderivability,redundancy,andconsistencyofrelations-thesearediscussedinSection2.Thenetworkmodel,ontheotherhand,hasspawnedanumberofconfusions,nottheleastofwhichismistakingthederivationofconnectionsforthederivationofrelations(seeremarksinSection2onthe"connectiontrap").
Finally,therelationalviewpermitsaclearerevaluationofthescopeandlogicallimitationsofpresentformatteddatasystems,andalsotherelativemerits(fromalogicalstandpoint)ofcompetingrepresentationsofdatawithinasinglesystem.Examplesofthisclearerperspectivearecitedinvariouspartsofthispaper.Implementationsofsystemstosupporttherelationalmodelarenotdiscussed.
1.2DataDependenceinPresentSystems
Theprovisionofdatadescriptiontablesinrecentlydevelopedinformationsystemsrepresentsamajoradvancetowardthegoalofdataindependence[5,6,7].Suchtablesfacilitatechangingcertaincharacteristicsofthedatarepresentationstoredinadatabank.However,thevarietyofdatarepresentationcharacteristicswhichcanbechangedwithoutlogicallyimpairingsomeapplicationprogramsisstillquitelimited.Further,themodelofdatawithwhichusersinteractisstillclutteredwithrepresentationalproperties,particularlyinregardtotherepresentationofcollectionsofdata(asopposedtoindividualitems).Threeoftheprincipalkindsofdatadependencieswhichstillneedtoberemovedare:
orderingdependence,indexingdependence,andaccesspathdependence.Insomesystemsthesedependenciesarenotclearlyseparablefromoneanother.
1.2.1.OrderingDependence
Elementsofdatainadatabankmaybestoredinavarietyofways,someinvolvingnoconcernforordering,somepermittingeachelementtoparticipateinoneorderingonly,otherspermittingeachelementtoparticipateinseveralorderings.Letusconsiderthoseexistingsystemswhicheitherrequireorpermitdataelementstobestoredinatleastonetotalorderingwhichiscloselyassociatedwiththehardwired-determinedorderingofaddresses.Forexample,therecordsofafileconcerningpartsmightbestoredinascendingorderbypartserialnumber.Suchsystemsnormallypermitapplicationpro-gramstoassumethattheorderofpresentationofrecordsfromsuchafileisidenticalto(orisasuborderingof)thestoredordering.Thoseapplicationprogramswhichtakeadvantageofthestoredorderingofafilearelikelytofailtooperatecorrectlyifforsomereasonitbecomesnecessarytoreplacethatorderingbyadifferentone.Similarremarksholdforastoredorderingimplementedbymeansofpointers.
Itisunnecessarytosingleoutanysystemasanexample,becauseallthewell-knowninformationsystemsthataremarketedtodayfailtomakeacleardistinctionbetweenorderofpresentationontheonehandandstoredorderingontheother.Significantimplementationproblemsmustbesolvedtoprovidethiskindofindependence.
1.2.2.IndexingDependence
Inthecontextofformatteddata,anindexisusuallythoughtofasapurelyperformance-orientedcomponentofthedatarepresentation.Ittendstoimproveresponsetoqueriesandupdatesand,atthesametime,slowdownresponsetoinsertionsanddeletions.Fromaninformationalstandpoint,anindexisaredundantcomponentofthedatarepresentation.Ifasystemusesindicesatallandifitistoperformwellinanenvironmentwithchangingpatternsofactivityonthedatabank,anabilitytocreateanddestroyindicesfromtimetotimewillprobablybenecessary.Thequestionthenarises:
Canapplicationprogramsandterminalactivitiesremaininvariantasindicescomeandgo?
Presentformatteddatasystemstakewidelydifferentapproachestoindexing.TDMS[7]unconditionallyprovidesindexingonallattributes.ThepresentlyreleasedversionofIMS[5]providestheuserwithachoiceforeachfile:
achoicebetweennoindexingatall(thehierarchicsequentialorganization)orindexingontheprimarykeyonly(thehierarchicindexedsequentialorganization).Inneithercaseistheuser'sapplicationlogicdependentontheexistenceoftheunconditionallyprovidedindices.IDS[8],however,permitsthefiledesignerstoselectattributestobeindexedandtoincorporateindicesintothefilestructurebymeansofadditionalchains.Applicationprogramstakingadvantageoftheperformancebenefitoftheseindexingchainsmustrefertothosechainsbyname.Suchprogramsdonotoperatecorrectlyifthesechainsarelaterremoved.
1.2.3.AccessPathDependence
Manyoftheexistingformatteddatasystemsprovideuserswithtree-structuredfilesorslightlymoregeneralnetworkmodelsofthedata.Applicationprogramsdevelopedtoworkwiththesesystemstendtobelogicallyimpairedifthetreesornetworksarechangedinstructure.Asimpleexamplefollows.
Supposethedatabankcontainsinformationaboutpartsandprojects.Foreachpart,thepartnumber,partname,partdescription,quantity-on-hand,andquantity-on-orderarerecorded.Foreachproject,theprojectnumber,projectname,projectdescriptionarerecorded.Wheneveraprojectmakesuseofacertainpart,thequantityofthatpartcommittedtothegivenprojectisalsorecorded.Supposethatthesystemrequirestheuserorfiledesignertodeclareordefinethedataintermsoftreestructures.Then,anyoneofthehierarchicalstructuresmaybeadoptedfortheinformationmentionedabove(seeStructures1-5).
Now,considertheproblemofprintingoutthepartnumber,partname,andquantitycommittedforeverypartusedintheprojectwhoseprojectnameis"alpha."Thefollowingobservationsmaybemaderegardlessofwhichavailabletree-orientedinformationsystemisselectedtotacklethisproblem.IfaprogramPisdevelopedforthisproblemassumingoneofthefivestructuresaboveÑthatis,Pmakesnotesttodeterminewhichstructureisineffect-thenPwillfailonatleastthreeoftheremainingstructures.Morespecifically,ifPsucceeds
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- Relational Model of Data for Large Shared Banks
![提示](https://static.bingdoc.com/images/bang_tan.gif)
链接地址:https://www.bingdoc.com/p-2744554.html