Incorrect subdatasets handling in python bindings

[antonio]

I'm using GDAL 1.5.2 in Debian Sid x64 I have an HDF5 file with 2 datasets "/MBI" and "/QLK":

$ gdalinfo -nogcp -nomd -stats test-subdatasets.h5 
Driver: HDF5/Hierarchical Data Format Release 5
Files: test-subdatasets.h5
Size is 512, 512
Coordinate System is `'
Subdatasets:
  SUBDATASET_0_NAME=HDF5:"test-subdatasets.h5"://MBI
  SUBDATASET_0_DESC=[200x200] //MBI (8-bit unsigned integer)
  SUBDATASET_1_NAME=HDF5:"test-subdatasets.h5"://QLK
  SUBDATASET_1_DESC=[10x10] //QLK (8-bit unsigned character)
Corner Coordinates:
Upper Left  (    0.0,    0.0)
Lower Left  (    0.0,  512.0)
Upper Right (  512.0,    0.0)
Lower Right (  512.0,  512.0)
Center      (  256.0,  256.0)

When I open the dataset from python I only get on subdataset:

Python 2.5.2 (r252:60911, Jun 25 2008, 18:03:42) 
[GCC 4.3.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from osgeo import gdal
>>> d = gdal.Open('test-subdatasets.h5')
>>> d.GetSubDatasets()
[('HDF5:"test-subdatasets.h5"://QLK', '[10x10] //QLK (8-bit unsigned character)')]

The "/MBI" subdataset is missing.

[antonio]

Test HDF5 file.

[warmerdam]

I have been unable to reproduce this problem on my AMD64 ubuntu system using GDAL trunk. It works fine for me.

Python 2.5.1 (r251:54863, Oct  5 2007, 13:50:07) 
[GCC 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import gdal
/home/warmerda/gdal/swig/python/build/lib.linux-x86_64-2.5/osgeo/gdal.py:81: DeprecationWarning: gdal.py was placed in a namespace, it is now available as osgeo.gdal
  DeprecationWarning)
>>> ds = gdal.Open('test-subdatasets.h5')
GDAL: GDALOpen(test-subdatasets.h5, this=0x84f010) succeeds as HDF5.
>>> print ds.GetMetadata('SUBDATASETS')
{'SUBDATASET_0_DESC': '[200x200] //MBI (8-bit unsigned integer)', 'SUBDATASET_1_NAME': 'HDF5:"test-subdatasets.h5"://QLK', 'SUBDATASET_0_NAME': 'HDF5:"test-subdatasets.h5"://MBI', 'SUBDATASET_1_DESC': '[10x10] //QLK (8-bit unsigned character)'}
>>> 

I *suspect* you are using a different version of the GDAL library from Python than from the commandline but it is hard to say why it isn't working for you. Please reopen if you have specific suggestions for how I might reproduce the problem.

[antonio]

Replying to warmerdam: Indeed it works if you use "GetMetadata" but still not works using "GetSubDatasets":

In [1]: from osgeo import gdal

In [2]: d = gdal.Open('test-subdatasets.h5')

In [3]: d.GetSubDatasets()
Out[3]: 
[('HDF5:"test-subdatasets.h5"://QLK',
  '[10x10] //QLK (8-bit unsigned character)')]

In [4]: d.GetMetadata('SUBDATASETS')
Out[4]: 
{'SUBDATASET_0_DESC': '[200x200] //MBI (8-bit unsigned integer)',
 'SUBDATASET_0_NAME': 'HDF5:"test-subdatasets.h5"://MBI',
 'SUBDATASET_1_DESC': '[10x10] //QLK (8-bit unsigned character)',
 'SUBDATASET_1_NAME': 'HDF5:"test-subdatasets.h5"://QLK'}

[warmerdam]

Doh! I had forgotten such a thing existed, and didn't read carefully.

On inspection I see that the HDF5 driver improperly returns subdatasets numbered from zero instead of one, and that the Python implementation of GetSubDatsets() assumes a start from 1 as per the specification in the data model document.

I have corrected the HDF5 driver to return subdatsets numbered started from one in trunk (r16062), 1.6 branch (r16063) and 1.5 branch (r16064).

Thanks for the report and the followup.