Hi,
Layers that contain Features represented by different Geometry types is
a great improvement in data modelization as it separates abstract models
from implementation constraints. Breaking a building layer into two
building layers (points and polygons) or breaking the river layer into
two layers (linestring and polygons) is generally not an acceptable
modelization from a user point of view, it is the result of soft
constraints : in these cases both layers have same attributes and when a
user want to query the data, he generally doesn't mind how it is
represented : linestring or polygons.
In this sens modelization with layers containing mixed geometries is
closer to the reality, those with one layer by geometry type is an
adaptation to make softs simpler.
Mixed layers are made possible thanks to JTS which allows powerful
spatial operation between any type of geometry (which is surely not a
simple matter !)
I think there are some drawbacks : mixed layers does not free the user
from implementation constraints completely. He still has several kind of
features to deal with, and it may be a pain. It may not be adapted for
some kind of applications (I think it is still more easy to work with a
pure linestring layer if you have to make network calculations), and can
give strange results if results depends on the spatial representation of
features (ex. you'll get into problem if you want to compute a ratio
"attribute/feature area" and your features are sometimes polygons and
sometime points...)
my 2 cents
Michael
Layers with several types of geometries are not easy to manage because a
few softwares around there can accept this (ex. thoses based on
shapefiles), but there are a demonstration of the powerness of JTS which
allows most of spatial operation between heterogeneous geometries).
Such layers, make it possible to separate data
Sunburned Surveyor a écrit :
Thank you for these scenarios Paul. I will include them in my article.
Do you mind if I mention you in the credits?
The Sunburned Surveyor
On 8/29/07, Paul Austin <[EMAIL PROTECTED]> wrote:
One reason would be precision and scale. Consider a building small
buildings could be 1m sq so these we would encode as points and larger
buildings we would use polygons.
On a related note why would I want to have two geometries on the same
feature, consider a two sided river I may want to have the polygon of
water boundary and a MultLineString to represent the banks of the river.
This would allow be to render the river in light blue and the banks in
darker blue but not have the darker blue lines where another river joins
in with this one.
Paul
Sunburned Surveyor wrote:
I'm writing a short article about working with layers that contain
Features represented by different Geometry types. One of the questions
that I am trying to answer in the article is:
"When is it appropriate to store features represented by different
geometry types on the same layer?"
I must say that I'm having a difficult time answering this question,
since this is not something I do in OpenJUMP. Instead of trying to
guess at a reason, I thought I would ask other users.
Why do you have layers that contain features represented by different
geometry types?
Thanks,
The Sunburned Surveyor
_______________________________________________
jump-users mailing list
[email protected]
http://lists.refractions.net/mailman/listinfo/jump-users
_______________________________________________
jump-users mailing list
[email protected]
http://lists.refractions.net/mailman/listinfo/jump-users
_______________________________________________
jump-users mailing list
[email protected]
http://lists.refractions.net/mailman/listinfo/jump-users
_______________________________________________
jump-users mailing list
[email protected]
http://lists.refractions.net/mailman/listinfo/jump-users
