As outlined in other tickets, support for null tokens can be problematic in some situations.
It can trigger:
analysis issues (when querying on a null token, you probably don't want your query to be analyzed)
mapping issues (see or HSEARCH-2415)
As suggested in HSEARCH-2389, a better solution might be to implement "exists" predicates. They already exist in Elasticsearch, and maybe we could implement them easily with recent Lucene versions.
As a follow-up, we'll try to implement these predicates for object fields in HSEARCH-2389.