I haven't done performance benchmarking, nor tuned it yet. That's my next plan. The library is largely meant for firing ad-hoc queries in production, or for tweaking data-sets and database queries.

Though reflection is slower, it allows us to fire queries on List<Object> sort of collections and the built in type conversions may allow us to compare a Calendar instance to a Date, or a Long to a String (with implicit parsing and conversions).

I ran some performance tests and added a basic caching layer on reflection - that improved it by around 50% in the very basic of tests. You may look at the results in README at https://github.com/sangupta/gather repo.