EducationalTestingService/skll

Inconsistent use of ids in FeatureSet.split_by_ids

Closed this issue · 0 comments

The method split_by_ids receives two sets of ints: ids_for_split1, ids_for_split2, according to the docstring these are the actual ids (assuming the ids are integers which is not necessarily the case).

However the method uses ids_for_split1 as indices of ids in the id list in one case (L519-521), and as the actual ids in the other (L523). ids_for_split2 is used as indices (L527-529).

def split_by_ids(fs, ids_for_split1, ids_for_split2=None):