Advances in Databases: 11th British National Conference on Databases, BNCOD 11 Keele, UK, July 7–9, 1993 Proceedings

286 Pages·1993·5.6 MB·English

by  Michael Freeston (auth.), Michael Frederick Worboys, Anna Frances Grundy (eds.)

Lecture Notes in Computer Science 696
Edited by G. Goos and J. Hartmanis
Advisory Board: W. Brauer D. Gries J. Stoer
M.Worboys A.E Grundy (Eds.)
Advances in Databases
ht11 British National Conference ,sesabataDWo
BNCOD 11
Keele, UK, July 7-9, 3991
Proceedings
galreV-regnirpS
Berlin Heidelberg kroYweN
London Paris Tokyo
gnoH Kong Barcelona
tsepaduB
Series Editors
Gerhard Goos Juris Hartmanis
Universit~it Karlsruhe Cornell University
Postfach 69 80 Department of Computer Science
Vincenz-Priessnitz- e3lartS 1 4130 Upson Hall
D-76131 Karlsruhe, FRG Ithaca, NY 14853, USA
Volume Editors
Michael Frederick Worboys
Anna Frances Grundy
Department of Computer Science, Keele University
Keele, North Staffordshire, ST5 5BG, UK
CR Subject Classification (1991): H.2
ISBN 3-540-56921-9 Springer-Verlag Berlin Heidelberg New York
ISBN 0-387-56921-9 Springer-Verlag New York Berlin Heidelberg
This work is subject to copyright. All rights are reserved, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, re-use
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other
way, and storage in data banks. Duplication of this publication or parts thereof is
permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from
Springer-Vedag. Violations are liable for prosecution under the German Copyright
Law.
(cid:14)9 Springer-Verlag Berlin Heidelberg 1993
Printed in Germany
Typesetting: Camera ready by author
Printing and binding: Drnckhaus Beltz, Hemsbach/Bergstr.
45/3140-543210 - Printed on acid-free paper
Foreword
The papers following constitute the written proceedings of the eleventh British
National Conference on Databases, held at Keele University in the Midlands of
England. As with many other current database conferences, a dominant theme is
the provision of the means to enhance the capabilities of databases to handle in-
formation that has a rich semantic structure. Such provision is clearly based upon
user requirements, since many applications demand it and system configurations
call for it. A major research question is how to achieve such a semantic scale-up
without sacrificing performance. There are currently two main paradigms within
which it is possible to propose answers to this question, deduction-oriented and
object-oriented. These two approaches are not disjoint; indeed, some would argue
that the object-oriented approach is subsumed within the deductive logic-based
approach. Both paradigms are well represented in this collection, with perhaps
the balance swinging in the direction of the deductive approach, which is followed
by both the invited papers.
It has been an unfortunate consequence of the admitted failure of the Japan-
ese Fifth Generation project to develop an effective logic-based paradigm that
all approaches based on logic are seen in some quarters as doomed to failure.
Our first invited speaker takes a much more optimistic line. Michael Freeston,
from the European Computer-Industry Research Centre (ECRC) at Munich has
argued that ECRC has achieved more than both the Japanese and American
initiatives in this area and that logic-based information systems are far from
dead. His paper describes work to provide efficient indexes to support both
deductive and object-oriented paradigms.
Both deductive and active databases are based upon rules, yet unification of
these two types has not yet been achieved. Our second invited speaker, Carlo
Zaniolo from the University of California at Los Angeles, discusses the technical
problems at the heart of such a unification. He outlines a semantic framework
that results in a rule-based language, presently being developed at UCLA, with
the potential to integrate deductive and active databases.
Three of the papers following address the theme of semantic enhancement
in the context of specific applications. Goble and Glowinski discuss the handling
of integrity constraints in a medical application, Fernandes et .la test a generic
unification of deductive and object-oriented approaches with a geographic appli-
cation, and Eaglestone et .la consider how recent research on database support
for design activities can extend beyond support for standard engineering and
software design to the artistic design process.
Two papers discuss distributed configurations. Castellanos considers ways to
enrich existing models in constituent databases of a distributed system so as to
provide integrated access to the federation. Cohen and Ringwood use a logic-
based language that supports concurrency to integrate heterogeneous nodes.
Four papers deal directly with the object-oriented approach to databases.
Barclay and Kennedy incorporate the established database concept of the 'view'
1v
into their object-oriented world. Hohenstein and Odberg use an extended version
of the entity- relationship model to enrich the semantics of databases and provide
the basis for a uniform C++ interface to a variety of DBMS. Carnduff and Gray
consider how function materialisation in object-oriented databases can enhance
their use for the engineering design process. Tagg and Liew provide an excellent
summary of the current state of object-oriented DBMS.
Returning to the deductive theme, Sunderraman constructs an extension to
definite deductive databases that allows for conditional facts. Model-theoretic
and fixpoint semantics are provided for this extension. Singleton and Brereton
consider some of the practicalities in implementing logic features using a rela-
tional database.
There are two other papers in this collection which do not fit so easily into our
general theme, yet are very much in the mainstream of current database research.
McBrien discusses some of the algorithms required for querying and updating an
historical relational database. The discussion provides information about what
is required of a relational system to support an historical information system.
Reddi's paper makes a contribution to the theory of functional databases by
showing how work on optimisation strategies for integrity constraint enforcement
in deductive databases can be transferred to the functional database approach.
Acknowledgements
I am very grateful to the programme committee, whose members are listed on
a later page. The programme committee was chaired quietly and effectively by
John Hughes at a meeting in February. We selected 31 papers for presentation at
the eleventh BNCOD conference from a total of 63 full papers submitted. Other
submissions were recommended for presentation in poster form. Each full paper
was refereed by at least three members of the programme committee.
I would like to acknowledge the support of the BNCOD steering commit-
tee, especially its chair, Alex Gray. Last year's conference organiser, Peter Gray,
offered much useful practical advice. At Keele, I am grateful for the help of
the conference organising committee, without whose assistance this conference
could not have taken place. Frances Grundy was able to contribute the benefit
of her experience of organising BNCOD at Keele in 1985. My postgraduate stu-
dents have done a great deal of work behind the scenes. In particular, Priyantha
Jayawardena shared many hours with me in routine tasks. Kendal Allen, the
conference administrator, provided a friendly and efficient interface with dele-
gates.
Keele, May 1993 Michael F. Worboys
Conference Committees
Programme Committee
J Hughes (University of Ulster) - Chair
J Bocea (University of Birmingham)
TJ Bourne (SIAM Limited)
R Cooper (University of Glasgow)
MS Deen (Keele University)
B Eaglestone (University of Bradford)
AW Gray (University of Wales, Cardiff)
AF Grundy (Keele University)
KG Jeffery (SERC)
G Kemp ( University of Aberdeen)
JB Kennedy (Napier, Edinburgh)
]L1 Lucas (Keylink Computers Limited)
Z Kemp ( University of Kent)
A Poulovassilis (King's College, London University)
NW Paton (tteriot Watt University)
R Tagg (Independent Consultant)
GCIt Sharman (IBM, Hursley)
C Small (Birkbeck College, London University)
MF Worboys (Keele University)
Steering Committee
AW Gray (University of Wales, Cardiff) - Chair
PMD Gray (University of Aberdeen)
SM Jackson (Wolverhampton University)
MII Williams (I-Ieriot Watt University)
MF Worboys (Keele University)
Organising Committee
K Allen (Keele ~lniversity)
AF Grundy (Keele University)
P Jayawardena (Keele University)
MF Worboys (Keele University)
Contents
Invited Papers
Begriffsverzeichnis: A Concept Index ........................................ 1
Michael Freeston (European Computer-Industry Research Centre, Munich,
Germany)
On the Unification of Active Databases and Deductive Databases .......... 23
olraC Zaniolo (University of California ta soL Angeles, USA)
Technical Papers
Semantic Constraints in a Medical Information System .................... 40
eloraC A. Goble, Andrzej Glowinski (University of Manchester, UK),
Keith .G Jeffery (Rutherford Appleton ,yrotarobaL Oxon, )KU
A Methodology for Semantically Enriching Interoperable Databases ........ 58
i~laM Castellanos (Universitat Polit~cuica ed Catalunya, Spain)
Distributed Databases Tied with String ................................... 76
Daniel Cohen, Graem Ringwood (Univeristy of London, )KU
Viewing Objects .......................................................... 93
Peter J. Barclay, Jessie B. Kennedy (Napier University, )KU
Function Materialization Through Object Versioning in Object-Oriented
Databases ................. . ............................................. 111
T. .W Caruduff, .W A. Gray (University of Wales egelloC of Cardiff, )KU
A C++ Database Interface Based on the Entity-Relationship Approach ... 129
ewU Hohenstein (Siemens AG, Munich, Germany),
Erik grebdO (Norwegian Inslitule of ,ygolonhceT Trondheim, Norway)
Object-Oriented Database Methodology - State of the Art ................ 147
Roger ,ggaT Benny Liew (Massey University, New )dnalaeZ
Deductive Databases with Conditional Facts .............................. 162
Rajshekhar Sunderraman (Wichita State University, Kansas, USA)
A Deductive Object-Oriented Database for Data Intensive Application
Development ............................................................. 176
Alvaro A. A. Fernandes, Maria L. Barja, Norman .14I Paton,
.M Howard Williams (Heriot-Watt University, )KU
x
Storage and Retrieval of First-Order Terms Using a Relational Database .. 199
Paul Singleton, .O Pearl Brereton (Keele University, )KU
Principles of Implementing Historical Databases in RDBMS .............. 220
Peter .cM Brien (University of London, )KU
Integrity Constraint Enforcement in the Functional Database
Language PFL ........................................................... 238
Swarup Reddi (University of London, )KU
Implementation of a Version Model for Artists Using Extended Relational
Technology .............................................................. 258
Barry Eaglestone, Geoff Davies, Mick Ridley, Nigel Hulley (Bradford
University, )KU
Begriffsverzeichnis: a Concept Index
Michael Freeston
ECRC, Arabellastrasse ,71 DW-8000 Munich ,18 Germany
Abstract. This paper describes a generalised technique for the indexing of
data structures of different complexities. It represents the culmination of a
tong research effort to devise efficient and well-behaved indexing for the low-
level support of persistent programming languages: more specificaUyl per-
sistent logic programming and object-oriented programming environments,
which the author sees as the future of database systems.
1 Introduction
This paper describes a generalised technique for the indexing of data structures
of different complexities. It represents the culmination of a long research effort to
devise efficient and well-behaved indexing for the low-level support of persistent
programming languages: more specifically, persistent logic programming and object-
oriented programming environments, which the author sees as the future of database
systems.
Both programming paradigms are particularly amenable to persistent implemen-
tation. In the author's conception of such systems, the executed code locates the
target of the next execution step - whether the target be method, object, rule or
ground fact - by indirect reference via an index (i.e. search by structure and value),
rather than by direct reference to a compiled-in address location. In persistent logic
programming, location via an index acts as a clause pre-unification step. In persis-
tent object-oriented programming, it provides the location of complex objects, and
the late binding of method to object.
The relative inefficiency introduced by indirect addressing through an index has
to be balanced against the advantage of being able to apply all the conventional
DBMS operations to objects or clauses, thereby extending the power of the pro-
gramming language to the database system, while eliminating the 'impedance mis-
match' which exists between every programming language and embedded database
language. In deductive systems, the loose coupling between a logic programming
language and a (relational) DBMS is transformed into the basis of a true, fully
integrated, deductive database system. In the object-oriented paradigm, it is the
author's view that, whatever arguments there may be about the precise properties
which an object-oriented database system should possess, it is nothing if it is not a
persistent object-oriented programming language.
Efficient, generalised indexing of complex structures is thus an essential prerequi-
site, if not the key, to viable persistent programming. But it has a more fundamental
significance: it is 021 years since Frege devised a notation (Begriffsschrift) for the
representation of any arbitrary concept. Since we reason by matching the structure
of one concept against another, efficient reasoning on large bodies of knowledge in
2
any automated system must require an efficient generalised indexing mechanism,
whatever particular representation the knowledge may take.
2 Fundamentals
Classically, data structures in database systems have been restricted to fixed-structure
records or tuples. The structure of a tuple is a set of fields, or attributes. The types of
the attributes have been restricted to a few simple types, such as real, integer, and
string.
In order to build an index to a set of records, it is assumed that all the members
of the set have the same structure. A key value must be associated with each. This
value may or may not be unique to the record. In principle, it may be the direct
value of a single attribute, or of several attributes, or it may be generated by some
conversion operation on one or more attributes. The unit of memory allocation for
the data and the index, in main memory or secondary storage, is a page, and this
page is almost invariably of fixed size.
The best known and most widely used dynamic method of indexing a set of
records in a database is the B-tree. In general, a tree structure is composed of a root
node, branch nodes and leaf nodes. By convention, the tree is represented in inverted
form i.e. with the root at the top. A traversal path through the tree is defined by the
sequence of nodes encountered along the path. The height of the tree is the length of
the longest direct path traversed from root to leaf. The fan-out ratio is the number of
branches leading from a node in the direction of the leaves. This ratio usually has a
range of allowed values, depending on the details of the design and implementation.
The limits of this range are the same for all the index nodes.
The B-tree takes the value of a single attribute in a record, or the lexical con-
catenation of several attributes, as the index key. Each index node corresponds to a
page of memory, and contains an ordered set of index keys. The index is constructed
as a hierarchy of index keys: at any particular level of the tree, each node contains
an ordered set of key values and, associated with each key, a pointer to a node at the
index level below. Each key represents an upper (or lower) bound to the key values
stored in the node to which it points. At the lowest index level, the keys point to
data pages containing records within the ranges defined by the lowest level index
keys.
When the insertion of an additional record causes a data page to overflow:
.1 the page is split into two pages about the median value of the index key at-
tribute(s).
2. the median key attribute value, together with an additional pointer for the new
page, is inserted in the index leaf node which holds the pointer to the original
page. An index node thus consists of key, pointer pairs, stored in key value order.
Plus one additional pointer, for the extreme upper or lower range partition.
If an insertion in an index node causes it to overflow, then the index node is
similarly split about its median key value, which is promoted upwards, together
with a pointer to the newly created index page.