Before someone ask me why I can't simply use HashMap? here is the answer.
I'm already using HashMap, and I'm looking for a way to decrease the number of temporary objects created during an object parsing.
Having such structure would help me doing that.
This is why Bill using the trove stuff. See the gnu.trove.TIntIntHashMap etc classes to see if they will work. The jar is thirdparty/trove/lib/trove.jar
I second the trove recommendation. The code in the trove project is very good.
It looks perfect.
I got a question about it though.
Our repository info points a 2.1.1 release:
While trove just released 1.1b5:
Am I missing something?
Looks like it's the wrong version number. I'll figure out which version it really is, fix it, and update the builds.
Just commited the code.
JBossSerialization now needs trove.jar.
If dependency is an issue, we could consider copying TIntObjectHashMap sourcecode into jbser sourcetree. (As it's a LGPL, if I only changed the package name, I don't think this would be an issue).
I have testcases working, but I will make some soakTests to make sure I'm not breaking anything.
Please, don't use jboss-serialization/HEAD unless you want to give a developer's input.
For now you can use CR5.
I don't think the trove dependency is a big issue. We may want to roll into the jboss common util module when its broken out.