[Gdal-dev] Re: Minor changes to gdal_java.i

[Andrea Aime]

Simone Gannecchini ha scritto:
> Hi all,
> I am attaching a very minor patch for the gdal_java.i swig  file that me and
> Daniele would really like to have accepted.
> If you look at it here below, you'll see that it simply allows to write and
> read without going through each single raster band.
> 
> It should provide a nice optimization in case you work tile by tile since it
> reduces the number JNI calls to read/write data by the number of #bands.
> 
> Index: C:/work/VC++/gdal/swig/include/java/gdal_java.i
> ===================================================================
> --- C:/work/VC++/gdal/swig/include/java/gdal_java.i	(revision 11187)
> +++ C:/work/VC++/gdal/swig/include/java/gdal_java.i	(working copy)
> +%extend GDALDatasetShadow {
> +  CPLErr ReadRaster_Direct( int xoff, int yoff, int xsize, int ysize,
> +                            int buf_xsize, int buf_ysize,
> +                            GDALDataType buf_type, int bands, void *buf)
> +{
> +
> +
> +   return GDALDatasetRasterIO( self, GF_Read, xoff, yoff, xsize, ysize,
> +                                 buf, buf_xsize, buf_ysize,
> +                                 buf_type, bands, 0,0,0,0);
> +
> +}
> +
> +
> + CPLErr WriteRaster_Direct( int xoff, int yoff, int xsize, int ysize,
> +                            int buf_xsize, int buf_ysize,
> +                            GDALDataType buf_type, int bands, void *buf)
> +{
> +
> +  return GDALDatasetRasterIO( self, GF_Write, xoff, yoff, xsize, ysize,
> +                                 buf, buf_xsize, buf_ysize,
> +                                 buf_type, bands, 0,0,0,0);
> +
> +}
> +
> +} /* extend */
> +
> +
>  %include typemaps_java.i

Change looks good, I see the advantage, I'm just wondering if this is
general enough to push it one level up, where every language can use it.
I mean, I don't see anything java specific there?

> We have other optimizations that we are working on for accessing large
> raster in java on Tile base but for the moment they are not stable enough
> (wonder if we'll have have time to make them stable though :-) )

Well, it may be nice to discuss their design, aim, and implementation. 
There are lots of skilled people here, someone may have ideas on how to 
stabilize them, whatever that means :-)

Cheers
Andrea