I've worked on several fairly small scale legacy HDF5 databases and each one utilizes grouping to perform lookups. For example as a contrived example lets say I have one 2 dimensional dataset where each cell maps back to a group which may store another dataset which holds the data which corresponds with the ID contained in the original dataset. This is fine but one project basically utilizes hdf5 in a relational database type system. (dataset contains value for group to open -> Group contains attributes which tell me the name of dataset to open -> finally get to data)
Since hdf5 has builtin compression would it make more since to store redundant data (say in compound datatypes).
Granted it depends on the requirements/complexity of the data, but just in general is it bad practice to store redundant data in HDF5?
