Hi, Could someone help with reviewing the task: https://issues.apache.org/jira/browse/KAFKA-19782?
You can find two branches — we were doing development in parallel. The new implementation is up to 5 times faster. There are two approaches: 1. Using a custom PatriciaSet (extends PatriciaTrie) 2. Using PatriciaTrie directly We also added benchmarks for both cases and fixed some issues related to them. I also want to remove some of the benchmarks and would like to discuss it with someone. With best regards Kuvardin Evgeny
