yeah, csv type guess is just one of case. i more likely to have a metadata management solution beyond the data lake, maybe Ontotext Refine's goal is what i want, but it's approach from RDF really sucks.