5 Model Generalization
101
Fig. 5.12. Progressive rendering in the Web client in small steps (method 1)
c. No progressive transfer and rendering (the common situation with current WFS
client software): the client waits until the complete data stream is received and
then visualizes the data at the appropriate level all at once. The data is still varioscale and in order of importance, so the basis for smooth zooming during user
interaction is also there.
Not using the topology (at client side) in the above scenarios is a limiting factor. When the client receives server-side constructed polygons and there is no line
simplification, the Painters algorithm works fine: it takes care of hiding the coarser
objects when the more detailed objects are received. However, in the case of line
generalization the Painters algorithm does not suffice because with line simplification the shape of the derived polygons will change, so the more detailed lines will
not always ‘hide’ the coarser lines. This means that in the case of line simplification
the approach will only work when the topology-to-polygon reconstruction is carried
out in the Web client and not in the database. The GAP-edges, BLG-trees, and nodes
with their geometry will now be streamed to the client; the polylines can already be
visualized during retrieval, so there is progressive transfer, but instead of the Painters
algorithm other methods are needed to hide the previously received, coarser lines.
Zooming and Panning
In the case of a WFS service (providing vector data) zooming and panning by the
user can often be handled in the client itself, without having to send new GetFeature
requests to the server. But when only a part of the objects for that spatial extent is
already in the client, new requests might be necessary: for example when zooming in,
more detailed data could be needed. With panning beyond the original spatial extent
the situation is a bit more complicated because it depends on the characteristics of
the new spatial extent (the density of objects there) what needs to happen.
When an additional GetFeature request is not necessary, zooming in will mean
hiding coarser objects and making visible more detailed ones. In the case of zooming
out this process is the opposite: now the more detailed objects have to be ‘switched’
off, and the coarser objects will be made visible. Important is that all of this is handled
in the client. Depending on the type of client this hiding/displaying could be based
101
Fig. 5.12. Progressive rendering in the Web client in small steps (method 1)
c. No progressive transfer and rendering (the common situation with current WFS
client software): the client waits until the complete data stream is received and
then visualizes the data at the appropriate level all at once. The data is still varioscale and in order of importance, so the basis for smooth zooming during user
interaction is also there.
Not using the topology (at client side) in the above scenarios is a limiting factor. When the client receives server-side constructed polygons and there is no line
simplification, the Painters algorithm works fine: it takes care of hiding the coarser
objects when the more detailed objects are received. However, in the case of line
generalization the Painters algorithm does not suffice because with line simplification the shape of the derived polygons will change, so the more detailed lines will
not always ‘hide’ the coarser lines. This means that in the case of line simplification
the approach will only work when the topology-to-polygon reconstruction is carried
out in the Web client and not in the database. The GAP-edges, BLG-trees, and nodes
with their geometry will now be streamed to the client; the polylines can already be
visualized during retrieval, so there is progressive transfer, but instead of the Painters
algorithm other methods are needed to hide the previously received, coarser lines.
Zooming and Panning
In the case of a WFS service (providing vector data) zooming and panning by the
user can often be handled in the client itself, without having to send new GetFeature
requests to the server. But when only a part of the objects for that spatial extent is
already in the client, new requests might be necessary: for example when zooming in,
more detailed data could be needed. With panning beyond the original spatial extent
the situation is a bit more complicated because it depends on the characteristics of
the new spatial extent (the density of objects there) what needs to happen.
When an additional GetFeature request is not necessary, zooming in will mean
hiding coarser objects and making visible more detailed ones. In the case of zooming
out this process is the opposite: now the more detailed objects have to be ‘switched’
off, and the coarser objects will be made visible. Important is that all of this is handled
in the client. Depending on the type of client this hiding/displaying could be based