You've documented in the past that at least in the .NET version optimizations happen on the first run of identify on a collection that enable significantly faster run times on subsequent identifies. However, I'm testing the Java version and it seems like there isn't a significant reduction in run time between the first and subsequent runs. Is this something that's simply not implemented in the Java version at this time or am I mistaken?
Many optimizations are missing in the java version. I would believe your benchmarks. Java port is really experimental at the moment.
Ok, no problem. I'm going to dive in and try to figure out what the major issues are. I'll make sure to take a look at the C# codebase, but can you give any advice on common sense optimizations that you've made on the branch that might yield the greatest improvement in the port?
Unfortunately, no, it's a long time and I no longer remember details of the implementation. Good luck anyway and let me know when you have any patches to submit.