Link Details

I kind of brushed over Alex's speed improvements in the recent Groovy 1.6-beta1 release, but based on some recent benchmarks it looks like it needs some re-iterating. Alex implemented something called call site caching. Here's what goes on. When we encounter a call site we look at some notes we left the last time we ran through it, (assuming there are any). We compare notes, and if nothing changed, we don't recalculate the MOP dispatch and just do what we did last time. If our notes are different, we do the full MOP dispatch over again. It's the 'memento' design pattern....