[gdal-dev] HDF5 product handling

[Frank Warmerdam]

On Mon, Jul 4, 2011 at 8:52 AM, Alex Mantaut
<alexmantaut at suremptec.com.ar> wrote:
>         I would like to ask what would be the best way to implement a driver
> that handles HDF5 subproducts, in a way that it doesn't interfere with the
> behavior for hdf5 for files that ain't from any specific product type, and
> allows to add new product types easly... (It's quite likely I will have to
> add a new hdf5 product type soon)
>
>        Because hdf5 products are just hdf5 files with a special order in the
> metadata, I think that the code to interpret them should go in the
> hdf5dataset (or hdf5imagedataset) to reuse the code that opens the files...
> Is this okay?
>
>        If the code for subproducts should go on the hdf5dataset (or
> hdf5imagedataset)  I have a doubt about design...

Alex,

We have a similar issue in the HDF4 driver.  The approach there has
roughly been to have one or some data items in the HDF4Dataset
class (or perhaps HDF4ImageDataset) which enumerates what the
product is, with a generic option for anything which isn't special
product.

The Open() method has special logic to identify products, and
harvest extra metadata depending on the product type.  Rather
than create subdriver's I'd suggest such using this approach,
and perhaps having one 'metadata harvest" method on the
driver for each subproduct.  In many cases that is sufficient.

I look forward to seeing your patches!

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent