Scaling on-demand feature views

[VladOnMyOwn]

I want to use feature view only to load raw data vectors (Arrays), and on-demand feature view to unpack these arrays and calculate features. That is, for each entity I have a set of records in the DB, which I "collapse" into one record containing lists of values in each of the fields - this is the feature view. Then in the on-demand feature view I expand these lists of values back into a set of records for each object and calculate features on this set of records. Do I understand correctly that in this case in order to be able to calculate an arbitrary subset of all features via get_historical_features, I need to define a separate on-demand feature view for each feature, so that I don't have to calculate a lot of unnecessary features?