When working with a partitioned dataset, it would be nice if null partitions were allowed. I often find myself writing custom SQL to cast null values of a partition key column to a special value so that I can use Dataiku partitioning. This makes sense for a query type dataset where partition values are inserted directly into a template, but for a table type dataset, it's frustrating to have to convert it to a query to add null support, especially when the partition list in a dataset changes frequently and a previously-working partition column breaks due to a record with a null value being added to the dataset. Supporting null partitions would speed up workflows and make pipelines more reliable.
