Comparison of Post-Relational and Object-Relational modelling for real-world database applications
| Published date | 04 November 2014 |
| DOI | https://doi.org/10.1108/JSIT-05-2014-0034 |
| Pages | 313-340 |
| Date | 04 November 2014 |
| Author | Nikitas N. Karanikolas,Michael Vassilakopoulos |
| Subject Matter | Information & knowledge management,Information systems |
Comparison of Post-Relational and
Object-Relational modelling for
real-world database applications
Nikitas N. Karanikolas
Department of Informatics, Technological Educational Institute (TEI) of
Athens, Athens, Greece, and
Michael Vassilakopoulos
Department of Electrical and Computer Engineering,
University of Thessaly, Volos, Greece
Abstract
Purpose – The purpose of this paper is to compare the use of two Object-Relational models against the
use of a post-Relational model for a realistic application. Although real-world applications, in most
cases, can be adequately modeled by the Entity-Relationship (ER) model, the transformation to the
popular Relational model alters the representation of structures common in reality, like multi-valued
and composite elds. Alternative database models have been developed to overcome these
shortcomings.
Design/methodology/approach – Based on the ER model of a medical application, this paper
compares the information representation, manipulation and enforcement of integrity constraints
through PostgreSQL and Oracle, against the use of a post-Relational model composed of the Conceptual
Universal Database Language (CUDL) and the Conceptual Universal Database Language Abstraction
Level (CAL).
Findings – The CAL/CUDL pair, although more periphrastic for data denition, is simpler for data
insertions, does not require the use of procedural code for data updates, produces clearer output for
retrieval of attributes, can accomplish retrieval of rows based on conditions that address composite data
with declarative statements and supports data validation for relationships between composite data
without the need for procedural code.
Research limitations/implications – To verify, in practice, the conclusions of the paper, complete
implementation of a CAL/CUDL system is needed.
Practical implications – The use of the CAL/CUDL pair would advance the productivity of database
application development.
Originality/value – This paper highlights the properties of realistic database-applications modelling
and management that are desirable by developers and shows that these properties are better satised
by the CAL/CUDL pair.
Keywords Database Modelling, Object-Relational Databases, Post-Relational Databases,
Relational Databases
Paper type Research paper
1. Introduction
The Relational model has been proposed during the late 1960s by Codd (1969,1970),
which was based on set theory and rst-order predicate logic to overcome problems
encountered when using other database models, such as data redundancy, dependence
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1328-7265.htm
Post-Relational
and
Object-Relational
modelling
313
Journal of Systems and Information
Technology
Vol. 16 No. 4, 2014
pp. 313-340
© Emerald Group Publishing Limited
1328-7265
DOI 10.1108/JSIT-05-2014-0034
on physical implementation and weak data integrity. Thanks to its mathematics
foundation, properties and theorems and techniques for improving design, the
Relational model can be applied to many different applications. Moreover, its simplicity,
the separation of logical from physical level and the ease of expressing operations
declaratively by the Structured Query Language (SQL) led to the wide adoption of the
Relational model by Database Managements Systems (DBMSs) developers and
database-application developers.
However, conceptual modelling of a real-world application is better accomplished by
the use of the Entity-Relationship (ER) model (Chen, 1976;Elmasri and Navathe, 2010,
chapter 7-8) and ER diagrams. ER diagrams directly represent the key components of a
business process, that is entities (persons, objects, places or events for which data are
stored) described by attributes and the relationships that express the dependencies and
requirements between these entities. The ER model is transformed to a Relational model
(Elmasri and Navathe, 2010, chapter 9) (entity and relationship sets are mapped to
relations) through a systematic process, which is afterwards normalized (Elmasri and
Navathe, 2010, chapter 15) to limit data redundancy and avoid update anomalies. This is
accomplished by the specication of a set of functional dependencies among attributes
and the decomposition of relations into smaller ones. This transformation of information
representation alters the representation of structures common in reality, like
multi-valued and composite elds, disconnecting the database model from the real
world. Nevertheless, during data manipulation, the combination of attributes through
the Relational join operator is needed. These drawbacks led to criticism of the Relational
model: “the Relational model is limited with respect to semantic content (i.e. expressive
power) and there are many design problems which are not naturally expressible in terms
of relations” (Worboys et al., 1990); “The Relational model is weak when showing
many-to-one relationships” (Pavkovi´
cet al., 2001).
However, the real-world that we are called to model with an Information System
(often with the use of a database) seldom incorporates repetitions of data (data
redundancy). As an example, consider a non-computerized managed library, where we
do not incorporate multiple series of books (copies) and a corresponding number of
bookshelves, to place the rst set of book copies sorted according to the title, the second
set of book copies sorted according to the rst writer, the third set of book copies sorted
according to the theme category, etc. On the contrary, we do not use any ordering (more
precisely, we use the ordering of books according to the date of import into the library)
or use some of the orderings that interest us (usually thematic), and, in addition, we
create indexes (with cards) for each one of the ordering that interest us. Each card
contains the key of the classication and a reference to the natural ordering (the location
of book in the bookshelves). Thus, we claim that the Information System designers
should be able to portray directly the real world in a model that provides more powerful
structures (through composite data types), as those of the real world.
To overcome the shortcomings of the Relational model, several models have been
proposed (popular or representative ones are reviewed in the next section of this paper).
Among them, the Object-Relational model (Stonebraker and Moore, 1996) is the one
most widely adopted by DBMS developers, as it is a compromise between the querying
capability of the Relational model and the ability of the Object-Oriented (OO) model to
model composite data. SQL3 is the SQL standard that supports concepts of the
JSIT
16,4
314
To continue reading
Request your trial