Hey @amueller! I want to work on this issue, can you explain what exactly i need to do.
Thanks!

@amueller basically mentions two issues:

1. split.py has absolutely no documentation - no docstrings whatsoever: https://openml.github.io/openml-python/stable/generated/openml.OpenMLSplit.html#openml.OpenMLSplit
   I'm pretty sure that Andy would like to have a general docstring here to have the class and its methods explained.
2. The class being instantiated in a private method of a different class. This is actually not correct: https://github.com/openml/openml-python/blob/master/openml/tasks/task.py#L83
   Also, I'm not sure why this would be an issue.

Therefore, it would be great if you add some documentation to the class, maybe @amueller can comment on issue number two in the meantime. Unfortunately, the documentation on OpenML.org is a bit sparse ATM, but you can find some improvements made by Andy, including a description of what a split and a task is, here: openml/openml.org#210

If you have any questions please do not hesitate to ask.

Sorry if I wasn't clear on the second issue.
My problem was that you can not construct an OpenMLSplit object class, or at least it's not clear to me how. The line you're listing is exactly what I meant. It's a private class method of OpenMLSplit. The only way to create an OpenMLSplit that I can see is calling OpenMLSplit._from_arff_file.

I guess in general that would be fine as a pattern, but I just found it pretty confusing. I guess we can't currently create tasks from the Python interface, so there is no reason why anyone would create a split. If we were to create tasks in Python, I would imagine that somehow would involve creating a split?

Given the this is necessary to create new tasks I can see your confusion and why we need documentation on how the split class can be instantiated.