compose a more complex object. Some underlying 
data are propagated from the components to the 
composite object in terms of the identifiers or 
the pointers. The data models including the 
primitive elements, the simple and the complex 
features are shown in the following diagram. 
| Complex feature | 
Simple feature rds. 
    
2 
  
  
  
UT ei Rm Ths 
finie) tere) (ensem 
008 
  
  
x : 
CML Sl. Mi, M2, - x ( (M1, Pointer } j 
e reet et © S S ET 
Fig.6 Data model based on unified data structure 
The arrows in the diagram express the identifier 
linking or the pointers in the data files, the 
ellipses contain the most primitive data in GIS. 
The diagram shows that the vector structures and 
the raster structures do not have any obvious 
distinctions because they use the same primitive 
data format--Morton key (see Li & Gong, 1992), 
which can guarantee that the unified data 
structure serves the two kinds of data and 
support object-oriented data model. 
3.3 Encapsulation Approach 
An outstanding characteristic is that the data 
and operation of an object are encapsulated in an 
entity, which is helpful for large software 
design and database protection. It is also 
required in GIS. Our system is designed with an 
object-oriented program language C++, which 
support encapsulation of data and functions, 
besides the operations for creating, deleting, 
modifying, displaying, it may contain various 
spatial queries and analysis operations. If all 
operations are encapsulated with data, it is not 
efficient. A new strategy presented here is that 
the operations acting on geometric data are 
divided into basic ones such as 
creating,deleting, modifying, displaying objects, 
which are encapsulated in a class, and appendant 
ones including query and analysis, which are set 
outside the class. Program construction is 
written by Turbo C*t in figure 7 
Class: name Appendant operations 
{ main() 
values of type 
public: common values 
basic operations call basic operations 
) call appendant operation 
) 
(a) Definition of 
a class construction 
(b) main program 
Fig.7 program construction and 
definition of a class 
4  OBJECT-ORIENTED MODELS 
FOR ATTRIBUTE DATA MANAGEMENT 
716 
4.1 Relational Approach and 
Object-oriented Data Model 
Relational database management system are often 
used to manage attribute data in GIS. There are 
many similarities between the relational approach 
and the object-oriented model. The relational 
model has to a certain extent semantic mechanisms 
of classification, generalization, and 
aggregation. In a loose sense, a table such as 
Relation 1 represents a class of parcels. The 
tuples in this table such as that for parcel I, 
are objects of this class. Generalization can be 
used when there is a need to store properties 
common to several classes. For example, the 
classes of parcels and highways can be 
generalized by the superclass of spatial features 
in Relation 6. An aggregation can model composite 
objects from their components, Relation 2 implies 
an aggregation relationship between a parcel and 
its boundaries. A map composed of different kinds 
of spatial features is also an aggregation. 
However, the fact that they share common behavior 
can not be deduced from the relational data 
model, which is lack of inheritance implement. A 
strategy proposed here is to integrate a object- 
oriented language such as HyperTalk with a RDBMS 
like Oracle, which manages attribute data in GIS. 
4.2 Object-oriented Programming in HyperTalk 
Object-Oriented Programming (OOP) is different 
from the traditional approach like Pascal, 
Fortran. In OOP, data and procedures that operate 
on the data are together, packaged in something 
called an object. There are five central ideas in 
OOP: object, message, method, class, and 
inheritance. 
HyperTalk is a programming language for Apple 
Computers. In HyperTalk, some key ideas are very 
close to OOP. HyperTalk uses some of the same 
terminology, adopts some of the same methods and 
adapts others, and in many ways looks and "feels" 
like OOP. But it is not a real object-oriented 
programming language, so it is called Object- 
Like Programming language. 
4.2.1 Object As mentioned above an object is an 
entity that includes data and procedures that 
operate on that data. Viewed from a programming 
standpoint, objects are the elements of an OOP 
system that send and receive messages. In 
HyperTalk, there are five types of objects: 
stacks, backgrounds, cards, buttons and fields, 
and they can be considered as the operations 
"icon" or data "container". Like OOP objects, 
each of these can send and receive messages. Each 
type of object can also be associated with a 
script that contains handlers, which correspond 
to methods (as will be seen in a moment). 
4.2.2 Message A message in an OOP world 
corresponds to a procedure or function called in 
a procedural language like Pascal. Everything in 
OOP is accomplished in only one way: one object 
sends a message to another object and the 
receiving object reacts. In HyperTalk, the 
messages can be held by five types of objects. 
Each type of message can be addressed to one or 
more of the types of objects encompassed by 
HyperTalk. The objects In HyperTalk can also 
receive the messages that are sent by using other 
language like SQL. 
Ge 
st 
phe 
da 
cl: 
agt 
ob, 
cl: 
ob; 
prc 
Ope 
inl 
anc