[postgis-users] Re: Re: Cascaded Union Aggregate function

[Jose Carlos Martinez Llario]

Thanks guys for your comments,

Actually I implemented the algorithm about joining adjacent polygons but
i did it in Java starting from an auxiliary table with some adjacent
information, you can check the source code of the sotfware PGAT (I
tested the adjacent dissolution using SPI too but I got some problems
with the server i think originated because i had no experience about
programming using SPI in postgres..(so i made many mistakes for
sure)...anyways it worked too when postgres didnt kick me out..:)

I didnt have enough time to continue my research but i was kind
interested in using triggers for maintaining some topology tables like
disjointed polygons..and other topology tables that can help to improve
some spatial operations.

I know it is a lot work to implement a topology model in PostGIS and i
dont know if the performance would be too slow (polygons->arcs->nodes
using cross joins). Maybe a half-way position would be using topology
based on rules (like in geodatabases) using triggers for implementing
some behavior that i think we need.

Some open source GIS  projects like Kosmo or gvSIG  are trying to use
topology rules but they do in a local way (JTS). We know in PostGIS in
easy to implement topology rules (just an SQL sentence: polygons must
not overlaps..not have gaps..not pseudonodes..) and maybe with triggers
(to maintain a basic topology tables to improve the performance)

I really want to know the opinions of users and the people who is
designing PostGIS because i really love it but i still thinking that if
we want to perform spatial operations with a big volume of data we need
some work more..

Whats is your point of view?

1-To finish the topology model based on an arc-node structure and to
implement the spatial operations  (buffer, union, etc.)
2-To implement some routines using topology rules but based on a simple
feature model
3-1 and 2 and the same time
4-Delegate this work to the postgis users who really need these
characteristics.
5-Any other

I think the best is to use persistent topology and topology rules at the
same time..but i have doubts about that because other software like
arcgis didnt do that even though arc-info used persistent topology (they
changed!)
 
and what about using tolerances? is it possible to use the GEOS fixed
precision model with PostGIS?..i mean like the resolution parameter in
geodatabases or fuzzy in arc-info. I modified the source code to change
to GEOS precision model but i got unexpected results in spatial
operations (the system crashed sometimes)

Thanks everybody..
Jose