the major difficulty, and probably the reason why this is unsupported currently, is the difficulty to distinguish between the string null (null as a word) and null. I wouldn’t claim that it’s impossible but you would need a new feature if you want to put Nulls into a custom input table.
@Viola you can use an Extended Mathematical Operations Processor with CAST(null AS DOUBLE) (or any desired target data type) to add a null valued column to your data.
Unfortunately the null value raises a red flag in many processors since they cannot cope with such data. Query processors should not have any problems with it but other processors enforce a non-null schema which is violated by such rows (either causing a schema exception or a NullPointerException in executed Spark functions - don’t worry, the exceptions will be caught but will break the execution of the affected sub-graph of the Workflow).
If you want to mix null and non-null values, a union processor can be used (or query for more sophisticated operations). It’s a bit more effort than a CIT but it yields real nulls.