If an option could be added to the kite-dataset CLI, so all columns get 'null', this would greatly help.
When using json-schema to create a AVRO schema, based on different JSON files from odata webservices (with paged data). The prediction of what types should be, works 'ok', not good.
That being said, being able to have all types [nullable] by default with a command-line option, would greatly help. So the importer wont fail as often.
If the schema is created based on a small json data file (example), importing errors frequently occurs.