Annals of Library Science and Documentation 1992,39(3), 74-80

OBJECT-ORIENTED DATABASE MANAGEMENT SYSTEMS-A NEW TOOL FOR INFORMATION CENTRES

JM JOSET VISWANATHANINSDOC14 Satsang Vihar MargNew Delhi 110067

To improve service efficiency and stay competiuve,information centres have to adopt continuously newtechniques in system development, software and hard-ware which move the state of the art forward ininformation management. Object-oriented databasemanagement systems (OODBMS) are a recent devel-opment in the field of database systems. Theiradoption and use will increase the capabilities andefficiency of an information centre. OODBMSs pre-serve the main features of classical database manage-ment systems and at the same time eliminate theirdisadvantages. Inadequacies of the present datamodels for new applications are highlighted, the fea-tures of OODBMSs are presented and the usefulnessof an OODBMS is illustrated by modelling a typicalapplication in an information centre.

INTRODUCTION

At present, the society may be said to be in anelectronic era as there is a widespread usage ofelectronic products such as computers, TVs, vid-eos, etc. even by common man. Informationplays a vital role in day-to-day life. Theinformation needs of the society have undergonechanges in both qualitative and quantitativeterms. As a result, serving the present generationof patrons calls for information centres to adoptnew techniques in system development, soft-ware and hardware. Developments in hardwareand software are taking place at a rapid pace. Inorder to. derive maximum benefits from such de-velopments, information systems must have aflexible design. The developments in hardwarehave led to the availability of powerful low costsystems. Software development has beenwitnessing the emergence of different softwareparadigms such as procedural, functional, logicand object-oriented. Different paradigms aresuitable for different types of applications.Exploiting this, a number of efficient third partysoftware packages have become available in themarket supporting a variety of applications.

Database technology is vital to the functions ofinformation centres. Present database applica-

74

tions in an information centre vary much from thatof the past. For example, there is an increasingrequirement to develop application systems thatcan manipulate multimedia data types such asimages, voice and text, manage online databasesystems, handling high volumes of data and pro-vide user friendly interfaces to casual end-users.Such applications require facilities for handlingcomplex Objects, incorporating semantics intodatabases, and supporting powerful and efficienttrigger constraint and event mechanisms. Simul-taneous with the rapid changes in hardware andsoftware technologies, expert manpower to de-velop new applications is becoming scarce.Hence, the software environment must permit easydevelopment of applications by novice profes-sionals. Conventional database managementsystems are not able to meet these newrequirements.

WHY CONVENTIONAL SYSTEMS ARE NOTADEQUATE?

Relational database management systems, whichare most popularly used these days, have beendeveloped to overcome the deficiencies of' theearlier network and hierarchical systems. Eventhe relational systems fail to meet the require-ments of the advanced applications of an infor-mation centre. The relational model of datapresents a tabular view of the world. It assumesthat every entity has a fixed number of descriptiveattributes and the values of these attributes maybe presented in a tabular form. Unfortunatelymost of the real-life problems do not fit into thetabular form very well. For example, a documentdatabase containing actual documents can notbe characterised by only a set of attributes givingthe description and location of the documents. Adocument may be composed of pictures, graphs,text, etc. To handle such a database of docu-ments, the system should support complex ex-tensible objects. There are other applicationswhich demand the support of dynamic objects.Here objects may change with time and theirattributes may take different values at different

Ann Lib Sci Doc

OBJECf-ORIENTED DATABASE MANAGEMENT SYSTEMS

TitleISSNIssueVol.Year

I

ARTICLE

I

I-

IAUTHOR

r--

r----

FIG. 1(a) Journal

Journal-Id Tille ISSN Vol Issue I Year

(Key-word: Journal-Id)

Journal-Id Article-TitleArticle-No

Vol 39 No 3 Sept. 1992

(Key-word: Journal-Id, Article-No)

FIG. : l(b) Relational representation

Author

75

JOSE AND VISWNATIIAN

points of time. Users may inquire about states ofobjects corresponding to different times. This mayrequire sO/TIesort of version control. For example,in the case of serials in a library, titles of some ofthem may change with time. In conventionalsystems, it is difficult to keep track of suchhistory of objects.

In relational systems, non-redundant representa-tion of objects and dependencies call for normal i-sation which leads to multiple tables containingattributes of the objects represented. The relation-ships among these multiple tables are not visiblein the database structure but are stored in appli-cation programs. There is no way for a user to findout the relationships from the database and he/she needs the help of the database designer oradministrator for this purpose. Due to thisdrawback of relational systems, the semanticsor meaning of real world situations cannot beincorporated into the database. Much of the knowl-edge of the database designer is not captured andpreserved as part of the database. As a result,those who access the database later are rquiredto possess additional knowledge to execute forexample, meaningful adhoc queries. Moreover,partitioning a real-world object into differenttables leads to inconsistency problems duringthe process of updation and deletion. Whenchanges are incorporated in a table, the corre-sponding Changes need to be reflected in therelated tables as well. This may not be achievedcorrectly always in the absence of description ofthe relationships in the database. If the datamodel is able to support complex objectsdirectly, these problems can be avoided to a greatextent.

In the practical world of database applicationsthe difference between the end-users view ofobjects and that of database structure of relationsis a source of great frustration and difficulty. Endusers find relational and other conventional repre-sentation of objects quite cumbersome, overlycomplicated, and difficult to understand. Forexample, to represent a journal containing differ-ent articles (Fig.1 (a)) in a relational system, onemay have to decompose the attributes into differ-ent tables as shown in Fig. 1(b). End users wantto deal with objects in a fashion that seemsnatural and convenient to them.

Relational systems do not support abstractiontechniques such as generalisation and specialisa-tion. For example, an entity of the type 'publica-tion' which is a generalisation of items like books,journals, reports etc. may generally have an authorand a title. But these two attributes alone do not

76

fully characterise the entities. The general entity'publication' needs to be specialised by differentparameters for identifying the type of publicationas shown in Fig. 2. For example, books do nothave volume and issue numbers whereas journalsdo have. One has ISBN number whereas the otherhas ISSN number.

Object-oriented database management systems(OODBMS) is a post relational database manage-ment system which is capable of preserving themain advantages of classical database systemsand eliminating their disadvantages.

OBJECT·ORIENTED DATABASE MANAGEMENTSYSTEMS (OODBMSs)

An object-oriented database management sys-tem can be defined as a database managementsystem with an object-oriented data model. Ob-ject-oriented data model can be perceived, in ageneral sense, as a framework in which the realworld semantics can be expressed with ease.Instead of set oriented records as in relationalsystems, OODBMS supports objects. The majoradvantage of an OODBMS is that any real-worldentity can be represented as an object. Forexample, in the case of a serials system in alibrary, the entities concerned are journals,invoices, bank drafts, reports etc. All the entitiesare treated as objects. Objects may be as simpleas numbers or strings or as complex as thecomplete details about a journal. In addition tothe attributes of an entity the behavior character-istics are also clubbed to the object throughprocedures or methods (Fig. 3). Hence, in thecase of an object-oriented system, Objects con-tains both data and procedures. Object-orienta-tion provides a natural way of representation ofreal world problems. Objects having the samecharacteristics and behavior are grouped into aclass. For example, a particular book can beconsidered as an object belonging to a class sayBOOK. In other words, an object is an instance ofa class.

Object-oriented database management systemsare the result of integration of object-orientedlanguage features with database concepts (Fig.4).Languagefeatures include Object-identity, inheri-tance, and encapsulation. Integrating thesefeatures with database concepts such aspersistence, data sharing, integrity, query etc.results in OODBMS. Object identity is a featureby which one can distinguish Objects from oneanother, regardless of their content, locationor addressibility. In relational systems, recordsare identified using attribute values. But in

Ann Lib Sci Doc

OBJECT-ORIENTED DATABASE MANAGEMENT SYSlEMS

PUBLICATION

B

Author:Title:

OOK JOURNAL REPORT

ISBN Vol. Report No.IssueISSN

FIG:2

I Attributes I -f- Method I ==> 1,-__ O_b_je_C_t..-J

FIG:3

Object OrientedLanguage Features

DatabaseFeatures

Object IdentityInheritanceEncapsulation

PersistenceSharingIntegrityQuery

--='/ OODBMS

FIG.: 4

•

Vol 39 No 3 Sept. 1992 77

JOSE AND VISWANATHAN

object-oriented systems each object is identifiedby a unique name, which can be either systemdefined or user defined. This is analogous tothat of distinguishing objects naturally. Encapsu-lation is the formal term that describes thebundling of data and procedures or methodstogether. As an example, in the case of a desktop publishing system, documents can be consid-ered as the objects. They have attributes likename, type and number. The procedures associ-ated with the documents may be 'format', 'changefont', etc. When we consider a document as anObject, these attributes and corresponding pro-cedures are clubbed together in it (Fig.5). Thisis called the encapsulation of data and methods.These methods act as visible interfaces to theObject. So, an object is like an engineering blackbox which produces some outputs when someinputs are given. The internal processing is hid-den from the outside world. In other words, theinformation concerning an object are hiddenfrom the external world, but it is identified by thefunctions associated with that object. It assuresthat an object is independent of its implementa-tion, and this ensures data abstraction andmodularity.

As in the real world, objects in a system areindependent of one another and they communi-cate among themselves by message passing.The paradigm of communication ensures that theObjects are not directly manipulated by the users.Instead, a user sends messages to an objectand the object itself determines the method bywhich it will react.

Inheritance is the mechanism for automaticallysharing data among classes and subclasses.This mechanism is very powerful and is not foundin conventional systems. Inheritance allows usersto create new classes by programming only thedifferences from the parent class. Considerinqthe desk top publishing system, a document canbe textual or graphical. The characteristics of atextual and graphical documents are different.By inheritance, we can create new classes suchas, text and graphics, from a parent class docu-ment (Fig. 6). The new classes contain all thecharacteristics and methods of the parent classand some additional characteristics andmethods. Inheritance provides a means for reduc-ing the effort required to develop and maintain adatabase. But its ability to allow a database toage gracefully is far more important. The key ideain inheritance is to provide a simple and powerfulmechanism for defining new classes that inheritthe structure and operations from an existingclass. New methods and old structure may be

78

overridden. Polymorphism is another feature bywhich different objects can respond differently tothe same message.

These features facilitate a sophisticated way ofdealing with data and hence provide advancedmethods of managing information.

MODELLING A CIRCULATION SYSTEM

Consider a real world example of a book circula-tion system in a library. Two classes of objectsexist: the BOOK and the BORROWER. Borrowergets a book issued from a library and returns itafter perusal. Borrower is a person who has aregistration in the library. So BORROWER canalso be considered as a class which belongs to amore general class PERSON. Hence, we canconsider BOOK and PERSON as two classes ofrepresentation. A particular book is an objectwhich belongs to the class BOOK and a particularborrower is an object which belongs to the classPERSON. The appropriate data attributes for theseclasses may be

PERSON = (first name, middle name, last name,address, gender, ..)

BOOK = (title, author, publisher, year, ...)Also some of the functions associated with theseclasses may be get name, change name, getaddress, change address, etc. The attributesand associated functions together describe theclass PERSON. So the class PERSON can be rep-resented as

PERSON:DATA:

first name,middle name,last name,address,gender, etc.

METHODS:get name,change name,get address,change address, etc.

Thus the class PERSON has its own attributesand functions. A particular person or borrower isan instance of the class PERSON and is treatedas an object having all the attributes andfunctions described in the class PERSON. Theparticular attributes of these objects will be acces-sible only through the functions specified with thatObject. For example, to Change the name of a

Ann Lib Sri Doc

OBJECT-ORIENTED DATABASE MANAGEMENT SYSTEMS

DOCUMENTS

DATA

Name :FileNumber:

METHODS

Format

FIG:5

DOCUMENTNameTypeNumber

Print

Line Size IDraw

Font

JustifyChange font

FIG. :6

STUDENT.BORROWER

FIG. 7

JOSE AND VISWANATHAN

particular person one should use the function'change name' associated with the class PER-SON. These functions act as visible interface tothese objects and the attributes and are hiddenfrom the outside world. Similarly, we can considerthe class BOOK with its attributes and functions.

For a university library system, we can considertwo categories of borrowers, one is faculty bor-rower and other is a student borrower. Therefore,we can consider two subclasses of the classPERSON, namely, FACULTY_BORROWER andSTU DENT_BORROWER (Fig. 7). By this hierarchi-cal representation we have captured the idea thata FACULTY BORROWER and STUDENT BOR-- -ROWER have all the attributes such as name,address, etc. and functions, such as get name, getaddress, etc. specified in the class PERSON.These attributes and functions are inherited fromthe class PERSON. The subclasses FACULTY-BORROWER and STUDENT_BORROWER mightcontain some more attributes and functions, whichare relevant to each of these subclasses. Forexample STUDENT_BORROWER might includeattributes such as student jd, programme of study,batch no, etc. By this process, we have two spe-cialized classes of class PERSON, viz.FACULTY_,?ORROWERand STUDENT_BORRO-WER. As far as class PERSON is concerned itis a generalisation of the classesSTUDENT BORROWER and FACULTY BORRO-- -WER. There can also be subclasses ofSTUDENT_BORROWER such as,GRADUATE STUDENT and RESEARCH STU-- -DENT, etc. The advantage of object-orientedapproach is that, one can plug in any number ofclasses by inheriting characteristics of the exist-ing classes at any point of the system develop-ment. This way of system design is tantamount tothat of natural way of analysing real-life prob-lems. By capturing naturally occuring objects ofapplication domain into the classes along with theassociated behavioural pattern, the object-oriented approach simplifies developmental proc-ess considerably.

CURRENT SCENARIO

A number of groups are working towards the 4.development of OODBMS all overthe world. There

80

are about 30 OODBMS under development withcommercial vendors, industrial researchlaboratories and universities. Most of these sys-tems support the core Object-oriented conceptsand additional features with varying degrees offunctionality. Some of the commecialy availablesystems are GemStone from Servio-Logic Devel-opment Corporation, Ontos from Ontologic, Inc.and Statice from Symbolics, Inc. Some of theindustrial research prototypes are IRIS fromHewlett-Packard, O2 from Altair, France, ZEITGEISTproject at Texas Instruments, Jasmin project atFujitsu, Japan, ORION from MCC Austin, Texasetc. Some of the university research prototypesare ENCORE/Observer project at Brown Univer-sity, POSTGRESS at University of California andOZ+ at University ofToronto, AVANCE at Univer-sity of Stockholm, Sweden, etc.

CONCLUSION

Information centres require very advanced tools tomeet the present day requirements. They alsoneed a system which is easy to use even bynovice users. Object-oriented database manage-ment systems are very promising for databaseapplications of an information centre. SinceOODBMS permits incorporation of more seman-tics into its database, it is easy to developapplications. OODBMS are likely to replace con-ventional systems in the coming decade.

REFERENCES

1. Batfour A, Britton C : InfoExec : Object-Oriented/Semantic Database implemen-tation. Information and Software Technology. 1990,32(4).

2. Dittrich K R : Object-Oriented databasesystems, The next miles of marathon,Information Systems 1990, 15(1).

3. Kim W: Introduction to object-oriented da-tabase management systems. MIT Press1990.

Winnbald A L (eds.): Object-oriented software construction. Addison-Wesley (1990)

Ann Lib Sci Doc