Re: [Geotools-devel] Big grid machine

[Martin D. <mar...@te...>]

Hello

Ian Schneider a =E9crit:
> Preliminary large grid / soft memory demo attached.
>=20
> There are three classes:
> Grid - encapsulate gridded data access
> Grid.Tile - manage tiled subset of grid
> Grid.TileReference - PhantomReference for pre-emptive tile caching
> GridImageProducer - ImageProducer implementation
> GridDisplay - cycles through modifying the entire grid, then displaying
> (main class)


Just took a look. Sound interresting. Some refactoring may help to=20
improve interoperability with J2SE imaging stuff, so that we can use it=20
with TileCache. Maybe a first step would be to make it work with=20
java.awt.image.Raster instead of Grid:

* Call Raster.getDataBuffer().

* Check the DataBuffer type we got. There is many DataBuffer types,
   in a simular way than there is many NIO buffers. For example, if the
   databuffer is an instance of java.awt.image.DataBufferDouble, then
   we could call its getData() method (which returns double[]) and wrap
   the double[] array in java.nio.DoubleBuffer.

Sound like that a bridge is possible that way. Actually (maybe I'm=20
missing the point), the bridge between various java.awt.image.DataBuffer=20
classes and corresponding java.nio.Buffer classes seems so obvious to me=20
that I would be a little bit surprised if there is not somebody (at Sun=20
or elswhere) already working on it...

	Martin.