dealing with duplicates
We should find a consistent way to deal with duplicates. Especially if the index
contains dupes. Moreover but not so common are column
dupes. IMO this should not handled by the tests, instead the core should ensure dupe-free data. But this is up to discussion.