As part of a class project, I am my teammate worked on improving the space and time performance of range filters over strings in a case where the range queries are known ahead of time. We developed a workload-dependent range filter which reduces a range emptiness query to a set of points queries, and uses a set of
bloom filters to answer the point queries. We found that this approach outperformed SuRF (the current state-of-the-art) in workloads where the average number of negative point queries is low.