TODO

OCTAVE
======
- i/o works for sparse and dense real valued things and cell arrays of
strings

MATLAB
======
- i/o works for sparse and dense real valued things and cell arrays of
strings

CSV
===
- should work for all types except sparse data but won't work with strings
containing "," - no escaping is done here

ARFF
====
- should work for all types except sparse data but won't work with strings
containing "," - no escaping is done here

uci-20070111 spambase (3842) failed slurping


LIBSVM
======
works for all even multi-label data sets, however slurping failed for

mediamill (exp1) data set (empty labels are not handled)


CLIENT API
==========

- use ml2h5.converter.basehandler is the class to load h5 files
  Only works for Data?

- Get list of all Tasks from mldata.org

- Use as a testing/validation tool

- Does it make sense to use abstract base classes (abc)?
  http://www.doughellmann.com/PyMOTW/abc/