Tweak UDT and tuple handling.
This merge request contains a few suggestions to the handling of UDT and tuple columns:
-
Remove the "keyspace" from the
UdtColumnconstructor. It makes it impossible to use a configurable keyspace in an application together with UDTs since the keyspace cannot be provided in the implementation of thectypetypeclass method in such a scenario. Furthermore, UDTs are constrained to the keyspace they are defined in [1], so that comparing keyspace names during the "typechecking" of a response to a request that is itself confined to a single keyspace does not seem to add much value. -
Adjust the typechecking of UDT columns in responses to include the type name as well as the field names and their types.
-
Adjust the typechecking of tuple columns in responses to use a proper recursive type equality check over the components (e.g. the current typecheck for tuples produces spurious errors on
varcharvstext).
Thereby I removed the encodeColumnType function since column types are
never encoded by CQL clients according to the protocol.