Hi! I'm working on Hbase 3 support, starting with Omid. While the API changes are not major, HBase 3 does remove HTableDescriptor and some methods in Filter.
We COULD solve this with extra modules and profiles, but since the difference is rather small, I'd like to avoid that and use reflection to handle the missing methods. The Filter code IS performance sensitive, but my micro-benchmarks show that reflection doesn't cause a performance hit in this case. I'm looking for your feedback for this solution. Did I make a mistake when benchmarking ? Is there some other reason not to use this solution ? Ticket: https://issues.apache.org/jira/browse/OMID-271 Draft PR: https://github.com/apache/phoenix-omid/pull/152 Class using reflection: https://github.com/apache/phoenix-omid/pull/152/files#diff-7cfca9b7c88aeeb17c7a662a0215e904acf0854d8eeaeadacbdb45b4ebeec2e3 Reflection performance microbenchmark: https://github.com/apache/phoenix-omid/pull/152/files#diff-e9f0ca82e9c730fbe803d28531ddef05b6486e903ef46de3c09c340cb861082b best regards Istvan