of the coincident 
| it, it that it does 
iting details in a 
'onvention is that 
) of the stack is 
ted by all editing 
ate in the feature 
0 make room for 
ng feature off of 
stack to provide 
the stack. These 
ure stack editing 
| in achieving a 
which does not 
editing tools. 
avior 
oincident feature 
me time was the 
uld it behave as a 
k, or as a FIFQ 
r first inclination 
| Since these were 
: the stack would 
'S or books lying 
aced on the stack 
ed off. When, 
stem we used for 
clined for FIFO 
he value of LIFO 
two reasons the 
more serviceable 
is a FIFO stack. 
age users from 
  
  
  
ick 
  
thinking that the stack order had any significance 
in itself. Features needed to be freely rotated to 
be accessible to other editing tools, because only 
the top feature would be editable at any given 
moment. Since the top feature would have to 
change frequently, there needed to be no other 
significance to being on top. Secondly, we found 
that the number of editing operations required for 
feature stack editing was less with a FIFO stack 
than with a LIFO stack. In addition to Push and 
Pop operations, a LIFO stack would require a 
Rotate operation to make any feature in the stack 
randomly accessible. A FIFO stack has no such 
requirement because feature rotation can be 
accomplished by a Push/Pop sequence. 
4.3 The Insignificance of Stack Order 
More should be said about the 
insignificance of stack order as it applies to 
coincident features. Some early attempts at 
automated mapping allowed feature geometry to 
be affected by the physical storage sequence such 
that a feature’s definition was affected by what 
came before it in the file (DMA's SLF format is 
an example). Today, this concept is generally 
discarded as being inconsistent with object- 
oriented data management. All significant object 
properties should be explicitly stated regardless of 
storage sequence. 
Similarly, in the business of making maps, 
drawing order usually has a profound affect on the 
overall appearance of the final product. This is 
especially true when maps include a number of 
polygon fill patterns or area colors. A swatch of 
green color can make an effective portrayal of 
vegetation when it lies below all cultural features, 
for example. But it can obfuscate critical detail if 
it is drawn on top of cultural features. 
That a device will draw cartographic 
features in a fixed sequence may be inevitable. 
The important thing is that feature discrimination 
for drawing order should be based on real-world 
attributes, not on any conventions in the physical 
data storage. In other words, the green vegetation 
area should be drawn first because the system 
recognizes it as an area of vegetation and it knows 
to draw vegetation before culture--not because the 
vegetation polygon occurs in the file before the 
city tint polygon. 
The same principle should be observed in 
the processing of coincident features as we have 
discussed here. Even though a system using the 
Auxiliary Table Method does store a "primary" 
151 
feature differently from optional "secondary" 
coincident features, the user interface should never 
suggest that being on the top of the feature stack 
is a kind of pseudo-property of the feature in that 
position. When it comes to coincident features, 
the question "Who's on first?" is irrelevant. The 
relative importance of coincident features (and 
consequently, their drawing behavior) should be 
determined from their real-world attributes. 
S. CONCLUSIONS 
In this paper we have reviewed various 
methods for managing coincident features in GIS 
applications. Although a number of advanced 
applications for raster data are available, we have 
focused our attention on strategies for dealing with 
vector data, especially when the system 
requirement for spatial topology forces a one-to- 
one linkage between graphic objects and attribute 
table records. We have seen how this is often at 
odds with the one-to-many relationship which can 
exist between graphic objects and real-world 
features. 
These coincident feature management 
strategies, or some combination of them, are used 
by many GIS systems today, although the user 
interface of any particular GIS may not reveal the 
true nature of its inner workings. We have 
argued that most GIS applications are likely to 
evolve to the state where all the system 
requirements in Table 1 are considered essential. 
We have also stressed that as GIS applications 
evolve, care should be taken to ensure that 
coincident features behave in accordance with 
object-oriented data management principles.