The partitioners extending FieldPartitioner do not support nested partition fields. The loaded values from MongoDB are read via document.get(partitionField) which does not work with nested path.

The returned value would be null and lead to unclear erros about duplicated keys or not supported topology, like "com.mongodb.spark.sql.connector.exceptions.MongoSparkException: Partitioning failed. Invalid partitioner configuration. The partitions generated contain duplicates:..."