when does a Row object have a schema

[Koert Kuipers]

the Spark-SQL Row trait has a schema that by default is null. when the
schema is null operations that rely on fieldIndex such as
getAs[T](fieldName: String): T do not work.

i noticed that when i convert a DataFrame to Rdd[Row] that the Row objects
do have schemas. can i rely on this?

when can i be sure that the schema is not null? what is the expectation
here?

thanks! koert