improvement to subset2, handling unclass, and related changes to data frame code, and to documentation