Repository API

See also Repository design and Can NIPY get something interesting from BrainVISA databases?

FMRI datasets often have the structure:

• Group (sometimes) e.g. Patients, Controls
  • Subject e.g. Subject1, Subject2
    • Session e.g. Sess1, Sess1

How about an interface like:

repo = GSSRespository(
    root_dir = '/home/me/data/experiment1',
    groups = {'patients':
              {'subjects':
               {'patient1':
                {'sess1':
                 'filter': 'raw*nii'},
                {'sess2':
                 'filter': 'raw*nii'}
                },
               {'patient2':
                {'sess1':
                 'filter': 'raw*nii'}
                {'sess2':
                 'filter': 'raw*nii'}
                }
               },
              'controls':
              {'subjects':
               {'control1':
                {'sess1':
                 'filter': 'raw*nii'},
                {'sess2':
                 'filter': 'raw*nii'}
                },
               {'control2':
                {'sess1':
                 'filter': 'raw*nii'}
                {'sess2':
                 'filter': 'raw*nii'}
                }
               }
              })

for group in repo.groups:
    for subject in group.subjects:
        for session in subject.sessions:
            img = session.image
            # do something with image

We would need to think about adding metadata such as behavioral data from the scanning session, and so on. I suppose this will help us move transparently to using something like HDF5 for data storage.