[rules-dev] new algorithm

[Mark Proctor]

We've managed to emulate/map 100% of our existing features and functionality. Which wasn't easy, and why it took much longer. The algorithm itself is lazy, yet some features such as dynamic salience or auto-focus, must be eager. Also several features, such as no-loop, could be done different and much more robustly in other ways.

There will be more work to do, this is just the beginning. Initially we'll just be tracking to make sure that we don't have worse than Rete performance for any use cases, before we focus on making it faster. Those who use root "context" objects to select rules to evaluate and fire, will not likely see much difference. Those that use a large number of rules, especially with rule flow or agenda-groups, should hopefully see some gains. Over all the lazy nature of algorithm should make it more forgiving of bad rule authoring. Also we've added in new tracing capabilities, so you can turn logging on and see how your rules are being evaluated, and data propagated.

Mark
On 6 Jun 2013, at 01:13, Mark Proctor <mproctor at codehaus.org> wrote:

> For those interested, we just got the new rule algorithm to pass 100% of our existing tests. So we'll be merging those changes in tomorrow :)
> http://blog.athico.com/2013/01/life-beyond-rete-rip-rete-2013.html
> 
> Mark