[Geotools-gt2-users] Fwd: length calculator

[Oleksandr H. <guz...@gm...>]

I got an improvement when using featureiterator instead of the visitor.
Probably because the features are not copied here.

Cheers
--
Oleksandr (Sasha) Huziy


---------- Forwarded message ----------
From: Oleksandr Huziy <guz...@gm...>
Date: 2012/12/7
Subject: length calculator
To: geotools-gt2-users <geo...@li...>


Hi,

I have a question concerning performance.

I need to select multilinestrings from a shape file (only the ones inside
given polygons), and calculate their lengths. The problem is that the
number of the polygons is big. So I decided to do it with multiple threads
(since I have many cores on the machine).

I did it as follows

       ShapefileDataStore dataStore = new ShapefileDataStore(shapeUrl);
       SimpleFeatureSource featureSource = dataStore.getFeatureSource();
       FilterFactory2 ff = CommonFactoryFinder.getFilterFactory2(null);

       ExecutorService service = Executors.newFixedThreadPool(processors);


        DrainageDensityCalculatorThread densityCalculatorThread;
        for(i = 0; i < nx; i++){
            for(j = 0; j < ny; j++){
                theCell = cells[i][j];
                //select rivers intersecting the gridcell
                densityCalculatorThread = new
DrainageDensityCalculatorThread(theCell, featureSource);
                service.execute(densityCalculatorThread);

            }
            System.out.println("i/nx = " + i +"/"+nx);
        }


        service.shutdown();
        try {
            service.awaitTermination(Long.MAX_VALUE, TimeUnit.NANOSECONDS);
        } catch (InterruptedException ex) {
            Logger.getLogger(this.getClass().getName()).log(Level.SEVERE,
null, ex);
        }


Inside the thread I apply the filter to the featureSource,  and then using
a featurevisitor I visit the selected featurecollection and calculate its
total length.
It takes long even with 20 threads, is it possible to improve the
performance? It is not a problem to wait since the operation is not
performed that often, though it would be nice to make it faster.

Thanks.


--
Oleksandr (Sasha) Huziy