The user outlines the workflow interactively by combining
these operators to form processing plans. Selecting the WiT^
menu item Run starts the processing of the flowchart. WiT®
calls for each operator a C program with a corresponding name.
Input and output filenames, as well as operator parameters if
necessary, are transferred. The C program processes them in
such a way that one or more GRASS commands can be issued.
GRASS carries them out and saves the results on the hard disk.
The names for these temporary files produced by GRASS are
automatically generated. The next VGIS operator uses the tem-
porary file as input. In case that not all necessary parameters
have been provided, the C program calls a C++ module that
prompts for the required data in a Motif” window. Reclass is
an example for such an operator requiring further information
from the user.
By employing several display- and output operators, interme-
diate results can be displayed and compared. Figure 7 depicts
the previous example with the display of intermediate results.
Output #1
distance: 100
Figure 7. Display of provisional results in VGIS
5 FIRST RESULTS
User resonance to the first VGIS prototype is quite positive.
The interface is easy to handle and the learning curve excep-
tionally low.
A few problems, mainly resulting from the use of WiT®,
should nevertheless be mentioned. Direct querying of parame-
ters in self-defined windows is not supported and can even
with tricks only be accomplished while processing a flow chart.
During the creation of flow charts only some very simple
WiT®-provided windows can be called. This is an enormous
obstacle.
In addition, it is not acceptable for a GIS application that the
number of inputs of an operator needs to be fixed and cannot
be parameterized. There is an unnecessary large set of overlay
operators only because the number layers needs to be defined
by the developer rather than the user. C is the only language
86
supported by the WiT® APL In general, WiT ® offers too few
means for customer-driven extensions.
The probably biggest handicap of the current VGIS prototype is
the lack of a real interface between its modules. Data keeps
being written to disk and read from there, causing the perform-
ance to drop dramatically. The original concept of WiT® allows
for images to be transferred from one operator to the next one
directly. GRASS (and other GIS) however, do not accepts other
input and output media than the hard disk.
The development of this prototype proved the correctness of
the underlying concept. The current development of an
Arc/Info? interpreter for VGIS will show that true data struc-
ture-independence is feasible. The authors came nevertheless
to the conclusion that due - to the restrictions caused by em-
ploying standard software packages - some major changes to
the current path of development are necessary.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996