h is 
C++ 
tion 
ect- 
for 
age, 
the 
this 
and 
ase. 
s of 
hics; 
the 
ated 
the 
uter 
the 
rder 
| the 
the 
from 
gned 
blem 
nted 
sign 
OORA 
ions 
ode l 
s of 
ses» 
ment 
and 
3. CONSIDERATIONS ON DATA MODEL AND DATA STRUCTURE 
The data model for database has been walking 
towards the object-oriented model from hierarchical 
model, network model and relational model. In 
hierarchical model, data are organized as tree 
structure which has direction and order. In network 
model, data are organized as directional graph 
structure. In relational model, data are organized as 
2-D relational table which is based on relational 
algebra. These conventional data models are not fit 
for new applications, especially for engineering 
applications (such as CAD, GIS, PIS etc.) because 
they have lack of capabilities of managing and 
manipulating highly-structured complex objects which 
exist everywhere in real world. Object-oriented model 
has extended and improved the conventional models. It 
is capable of handling and simulating complex objects. 
The hierarchical structure, network structure and 
relational structure can exist in a complex object 
simultaniously. Object-oriented model has expansibili- 
ty so that new contents can be added to the existing 
data model, and different types of data objects can 
be held and manipulated with a unified mechanism of 
management. 
  
  
  
The data types in PIS include vector data, raster 
data and attribute data. There are two kinds of PIS 
systems: raster systems and vector-based systems. 
Both have an important place in PIS and will continue 
to prevail for a long time.The latter works well when 
real world spatial conditions can accurately be 
defined as lines or edges. The vector approach can 
obtain important topological information which is 
difficult to achieve with the raster model, but it is 
rather ineffective for performing Boolean and overlay 
operations on different data layers. The raster 
( including compressed raster encoding such as 
quadtree ) model is more appropriate when the problem 
can be described as discrete samples of continuous 
fields. It supports image algebra operations and has 
Powerful backing of image processing techniques. Due 
to no clear answer for which one is better, the 
combination of vector and raster structures would be 
more appropriate in PIS (M. Molenaar et al., 1990). 
In object-oriented model, the complex objects can 
be composed of different types of objects which can 
have different data structures and can be distributed 
in different databases. In this way, the vctors, 
rasters and attributes are combined into a unified 
data structure —— object-oriented structure. For 
example, a geographic entity may be defined as a 
complex object as shown in Fig. 1. 
High-Level 
  
  
  
  
  
  
  
  
[ d 
| A Geographic Entity | Complex Object 
L J 
id 
U li 3 
| | | 
r À sr i "11 i 1  Lower-Level 
| Vector Data | | Attribute data | | Raster Data | Complex Objects 
1 FL ) L ] 
1 T 
| | 
pl-——————— = pn 
| | | | 
L 1 L 1 
I up i i ui i d 
| Complex Graphics] | Attributes | | Imagery | | DEMs | 
L I LL. J L 1 L J 
T Y ; 
r p " raser 4 pi——73 
| | | | | | | 
i i 41 i ^1 T i I i edd L ; F L E i i 
| Nodes | | Arcs | | Polygons | | Attributes | | Encoding | | Attributes | | Encoding | 
L. VoL |i 1 l 1 | 1 ]) 1 A 4 J 
  
  
  
Topological I dE 4 
Structure | | | 
pl r 1 7) 4 i 1 
| Nodes | | Arcs | | Attributes| Simple Objects 
L A tL AA -J 
  
Fig.1 A geographic entity as a complex object. 
617