TODO.rst

For a future release

• Converter: Possibility to specify other names than 'labels', 'features' for the input files

• Test convertion from h5features old versions

• read/write bigger than RAM -> catch MemoryError when np.concatenate on writing.

• labels can be of arbitrary type (optionally sorted)

• Have a h5features.File class inspired by h5py.File

  • Make Data a dict with the following syntax:

    with h5f.Reader(file, group) as reader:
        reader['item'][from_time:to_time]
        reader['item'].features
        reader['item'].labels
        reader.keys()
    

  • Make an Item class wrapping Labels and Features

• Implement sparse functionalities

• Handle h5py MPI driver for concurent reading