We need a way to read records knowing just the schema and format. Currently we use Java Objects, but this will no longer be possible if we don't have access to what object was originally written, which will be in case in the explore context. Therefore, we need some object that will be used to represent a Record (calling it StructuredRecord for now) that essentially acts as a map of fields.
We would want this class to be re-usable by tables, streams, and filesets. We also want to keep in mind that though it may start out as an internal class, we would want to expose it through the API for programmatic use.