Chris wrote:
> I added some debug, and apparently, the LineString intersection is not
> due to my algorithm. It happens approximately 12 times in the first loop
> iteration, then it does not happen anymore. This was probably due to my
> shapefile.
>
> So I chose to ignore the LineString intersections and now my algorithm
> has been running for approximately half an hour and it is the fourth
> iteration of the while loop (still unfinished). I added some debug to
> see how many LineStrings were created for each iteration:
> Added 2200 new lines
> Added 7666 new lines
> Added 26234 new lines
>
> Is my algorithm ok? Is it normal it is taking so long? Is there any way
> I can optimize this? If not, I guess I need to save the result somewhere
> in a file and work from the result from now on because it is taking too
> much time.
Intersection is quite an expensive operation. I am not surprised it is
taking this long with any non-trivial amount of data. Luckily there are
some easy ways to optimize this. The best way to be to use a spatial
index instead of looping through every line string and doing an
intersection. So the the first step of your algorithm will be to
populate the index with all of your lines. Its pretty easy to use:
SpatialIndex index = ...;
for ( LineString l : lines ) {
index.insert( l.getEnvelopInternal(), l );
}
Then you can replace the second loop with a lookup in the index.
//current line being processed
LineString l = ...;
//do a looup in the index
List close = index.query( l.getEnvelopeInternal() );
//do intersection on close lines
for ( LineString c : close ) {
l.intersect( c );
}
.. etc..
This will greatly reduce the number of intersections you have to do and
you should see a pretty drastic performance improvement. There are two
implementations of com.vividsolutions.jts.index.SpatialIndex: STRTree
and QuadTree. The STRTree will give you best performance but it is
static, which means you cant update over the stage of your algorithm
which you will probably want to do with the new lines that are produced.
So you will probably have to just use a the QuadTree implementation.
Either one should suffice in this case.
Try that and let me know how it works.
-Justin
>
> Regards,
> Chris.
>
> On Jan 11, 2008 9:28 PM, Chris <[EMAIL PROTECTED]
> <mailto:[EMAIL PROTECTED]>> wrote:
>
> I tried to use another Vector as you advised. This way, the size the
> the lines Vector is not changing while iterating on it. I'm joining
> the new code to this mail.
>
> It is taking quite some time (2-3 minutes) but this time I got an
> exception:
> java.lang.ClassCastException: com.vividsolutions.jts.geom.LineString
> cannot be cast to com.vividsolutions.jts.geom.Point
>
> at this line:
> Point p = (Point) gc.getGeometryN(0);
>
> where gc is the intersection between the two lines. Apparently, the
> intersection between two LineStrings can be a LineString. I guess
> this means that the two lines are identical? How could this happen?
> Is my algorithm still wrong? or maybe I should just ignore when this
> case happens?
>
> Regards,
> Chris.
>
>
> On Jan 11, 2008 7:28 PM, Justin Deoliveira < [EMAIL PROTECTED]
> <mailto:[EMAIL PROTECTED]>> wrote:
>
> Hi Chris,
>
> I would have to mock up some data and test it but I dont think the
> problem is the method but in the management of the data
> structure. The
> code is modifying the same list it is iterating over... which is
> generally not a great idea.
>
> I recommend restructuring the algorithm so that when two new
> lines are
> created they are added to a separate list. When the outer loop
> termines
> add the new lines to the original list and repeat the algorithm
> terminating when there are no more new lines.
>
> Hope that helps. I could be wrong as i said since i have not run the
> code. If you find you continue to have the same problem let us
> know .
>
> -Justin
>
> Chris wrote:
> > Hi,
> >
> > After loading a shapefile (ESRI), I created a graph using a
> > DirectedLineStringGraphGenerator. The creation went fine but
> when using
> > a DijkstraShortestPathFinder.getPath() on it, I noticed that
> there was a
> > problem because it returned null. Apparently, the linestring graph
> > generator does not take internal intersections of
> > lines into account when building the graph. Hence, I decided
> to code a
> > routine to split the LineStrings at their intersection as
> advised there:
> >
> http://www.nabble.com/Re%3A-LineString-Graph-Traversal-p1489791.html
> <http://www.nabble.com/Re%3A-LineString-Graph-Traversal-p1489791.html>
> >
> > The problem is that My routine is taking "forever" to do the
> splitting.
> > Either, I'm doing it wrong or it is not optimized enough.
> After waiting
> > for 10 minutes, it is still not finished (although, the
> shapefile is not
> > that big). Could someone have a look at my function and see
> what the
> > problem is?
> >
> > Thanks in advance,
> >
> > Best regards,
> > Chris.
> >
> >
> >
> >
>
> ------------------------------------------------------------------------
> >
> >
>
> -------------------------------------------------------------------------
> > Check out the new SourceForge.net Marketplace.
> > It's the best place to buy or sell services for
> > just about anything Open Source.
> >
>
> http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
> >
> > !DSPAM:4007,4787a0db264908992556831!
> >
> >
> >
>
> ------------------------------------------------------------------------
>
> >
> > _______________________________________________
> > Geotools-gt2-users mailing list
> > Geotools-gt2-users@lists.sourceforge.net
> <mailto:Geotools-gt2-users@lists.sourceforge.net>
> > https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users
> >
> >
> > !DSPAM:4007,4787a0db264908992556831!
>
>
> --
> Justin Deoliveira
> The Open Planning Project
> http://topp.openplans.org
>
>
>
> !DSPAM:4007,4787dd5553261439371379!
--
Justin Deoliveira
The Open Planning Project
http://topp.openplans.org
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
Geotools-gt2-users mailing list
Geotools-gt2-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users
