Definitely a lot to read and improvements to make. I will probably do a more complete benchmark with more datasets on a later post.

Thanks for the suggestions.