[gdal-dev] Gdal and Open Data Cube

Julien Michel julien.michel at cnes.fr
Fri Jun 2 08:18:25 PDT 2017

Le 02/06/2017 à 15:20, Even Rouault a écrit :
> On vendredi 2 juin 2017 14:28:18 CEST Julien Michel wrote:
> > Dear all,
> >
> > There is a lot of activity around this Open Data Cube [1] initiative,
> > which seems promising. As far as I understand it however, what we can do
> > with data in the datacube is limited to what the python API allows you
> > to do. Is there any plan (and does it even make sense ?) to have a gdal
> > driver that allows to address data ingested in a datacube instance ?
> I've just discovered this project, and browsed quickly through the docs at
> http://datacube-core.readthedocs.io/en/latest/, so the following might 
> be not really relevant.
> According to
> http://datacube-core.readthedocs.io/en/latest/user/intro.html , it 
> seems that ingested data is put in tiled netCDF files, and referenced 
> in a PostgreSQL database (not clear if the files are blob in the DB or 
> in the filesystem)
> So if the data itself is stored in the filesystem, a driver could 
> probably read the metadata in the DB and create some sort of virtual 
> mosaic around the tiles. If their database schema is stable enough (it 
> doesn't seem to be formally documented or I didn't find where), the 
> driver could probably read it directly. Otherwise if the DB schema may 
> change, then yes you'd need to use their Python API. Could be a 
> potential use case for my experimental concept of GDAL drivers written 
> in Python (*). What would be fun here is that given they use GDAL, 
> their Python API probably uses GDAL under the hood, so you would have 
> GDAL core -> OpenDataCube driver in Python -> OpenDataCube Python API 
> -> GDAL SWIG Python bindings -> GDAL core -> netCDF driver ;-)
Thanks a lot for your response. I am still trying to figure out if this 
is of any interest or not (I mean the gdal driver for ODC). It seems to 
me that if you spent all the time and storage to build such a datacube, 
it would be better to still be able to read back data with standard 
tools. Plus the datacube could offer a nice abstraction of image 
collections. Maybe the simplest way is to ask them.
> (*) https://lists.osgeo.org/pipermail/gdal-dev/2017-April/046526.html
CNES - DSO/SI/2A - BPI 1219
18, avenue Edouard Belin
31401 Toulouse Cedex 09 - France
Tel: +33 561 282 894 - Fax: +33 561 283 109

More information about the gdal-dev mailing list