It is the natural complement to as.data.frame
it will have initially one method for data frames, but more are possible for local files, hive tables and such
can be a focal point to progressively refactor all conversion code to these two as.something functions.
the data.frame methods is almost indispensable to write quick test cases.