Hi Arjun. Yes discussions are very helpful - and this is what dev is for (we
don't have enough discussions like this here in my opinion). We base a lot
of our designs on feedback here.
As for rule flow, yes it would be a set of rules. The idea that was
implemented in jBPM was using agenda-groups to control flow, so the "set" of
rules what covered by membership in an agenda group.
Of course, the set could be arbitrary, but agenda group makes a lot of
sense. Flow between rules would not be wise, as they are not rules then but
procedures, defeating the purpose (in which case, the logic could/should be
put in the flow itself, not the rule).
As for rules/files - well in the repo logical model, essentually it is one
rule per "file" in a sense - so you could group them how you like.
Regarding performance, yes there are optimisations to be had by examining
the network and "dependencies" (I am working on a rule analytics module in
my spare time that would help) which can allow futher optimisations, perhaps
even static agendas in some cases. In most cases, it requires knowing
something about what the RHS of the rule is doing, which currently we treat
as opaque (but that will change in future).
Always lots to be done, jump in, the water is warm.
On 12/1/06, Arjun Dhar <[EMAIL PROTECTED]> wrote:
Hi Michael,
Rule Flow; Great:
-----------------
Yes, a generic rule flow would be ideal.
..At a conceptual; how would one represent mutual exclusions? For some
reason
I'm more comfortable looking at them as "Sets".
And also when we talk of flows, do we mean from a rule to another or it
could
be a set of rules to another set? And, in the process flow it is possible
to
skip an entire Set? Maybe you can see why I'm looking at it from a set
point of
view. Where a set is a logical grouping, not necessarly in the same file.
..ok, I think I'll conclude here and give yr eye balls a rest! :) I'm sure
you
guys have a lot to think about.
The idea of the questions is not to get answers, but to throw the idea in
the
open for discussion. Being focussed on some things can make one loose
perspective of other things :o)
...One last thought regarding PERFORMANCE;...its a simple idea but I dont
know
its worth. Often, rules may have inherent dependece (im not sayin we
introduce
them, some can be free to live as they are).
Correct me i Im wrong, we are able to prepare the agenda based on Object
type
node matching and the codition matching the facts...right?!
If the above is accurate, then when when we know that there is dependence
on
other rules (rule flows can really help here) why cant that be taken as a
factor also? ==> Use a pre-condition to filer out matches?? My knowledge
on
this is based on the documentation provided and hence I'm not confident.
But I
did not see any thing like this mentioned and thought it was worth
mentioning.
I'll try get into this code as soon as I have my bread n butter work
regarding
earlier concerns done. :)
thanks a lot.
-Arjun
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
