I did not had time to analyze what jess is doing, but note that what is important is the final answer. In your example, with Move(1,2) and Move(2,3), the final answer must be Win(2), right? And that is what Drools will answer, does not matter the order in which the data is entered into the engine.
BUT, *very important*: the following construct in backward chaining:Is better represented in forward chaining using *logicalInsert* instead of a regular *insert*:
win(X):- move(X,Y), not(win(Y)).logicalInsert(new Win(m.getFirst()));
rule "direct" % Drools
whenthen
m : Move(x : first, y : second)
not Win(first == y)
end
Since in your backward chaining rule, only one win() predicate instantiation will remain true.
So, even with differences in the reasoning algorithm, the answer is correct.
Please explain further if I am missing anything.
Edson
2009/4/17 Paul Fodor <paul.i.fodor@gmail.com>
Hi Edson, Greg,
I don't think the rule is written wrong. This is how the win-nowin program is written in logic programming: X wins if there is a move from X to some Y and Y doesn't win:win(X):- move(X,Y), not(win(Y)).rule "direct" % Drools
then
when
m : Move(x : first, y : second)
not Win(first == y)
insert(new Win(m.getFirst()));
endI think that it's interesting that, in Jess (another production rule system), the stratified model is always computed right, no matter what was the order of the facts in the database. If you want to take a look, please see the equivalent program in Jess for win-nowin that I attached. Just run it with:
jess test.clpwin_upper1_jess.jess
(move (cur 1) (next 2))
(move (cur 1) (next 3))
(move (cur 2) (next 4))
(move (cur 2) (next 5))
...win_upper2_jess.jess
(move (cur 2) (next 4))
(move (cur 2) (next 5))
(move (cur 1) (next 2))
(move (cur 1) (next 3))
...test.clp:
(deftemplate move (slot cur) (slot next))
(deftemplate win (slot val))(defrule find_win
(move (cur ?cur) (next ?next))
(not (win (val ?next)))
=>
(assert (win (val ?cur)))
)(defquery query-win
(win (val ?val))
)
(open "win_result.txt" output a)
(printout output ./win_upper1_jess.jess crlf)
(reset)
(load-facts "./win_upper1_jess.jess")
(bind ?tmx (call java.lang.management.ManagementFactory getThreadMXBean))
(deffunction cputime () (return (* (?tmx getCurrentThreadCpuTime) 1E-9)))
(bind ?starttime_wall (time))
(bind ?starttime_cpu (cputime))
(run)
(bind ?query_result (run-query* query-win))
(bind ?count 0)
(while (?query_result next)
(++ ?count)
)
(printout output "solutions: " ?count crlf)
(bind ?endtime_cpu (cputime))
(bind ?endtime_wall (time))
(bind ?walltime (- ?endtime_wall ?starttime_wall))
(bind ?cputime (- ?endtime_cpu ?starttime_cpu))
(printout output "computing cputime: " ?cputime crlf)
(printout output "computing walltime: " ?walltime crlf)
(close output)
Regards,
Paul Fodor.2009/4/16 Edson Tirelli <tirelli@post.com>
Ha, thanks a lot Greg. I need new glasses... he is actually comparing with the parameter "second", but when creating the win fact, using the parameter "first".
not Win(first == m.second)insert(new Win(m.first));Yes, in this case the engine is working exactly as it should.
Anyway, I added the (fixed) test case to the codebase, just in case. :)
Thanks,
Edson
2009/4/16 Greg Barton <greg_barton@yahoo.com>
You don't have to worry. The engine is acting as it should.
The rule Paul had was this, a bit simplified for clarity:
rule "direct"
when
m : Move()
not Win(first == m.second)
then
insert(new Win(m.first));
end
If the insertion order is [Move(1,2), Move(2,3)] then the rule matches first on Move(2,3) and Win(2) is inserted. No other rule fires because now Move(1,2) and Win(2) match up, removing the instantiation with Move(1,2) from the agenda.
If the insertion order is [Move(2,3), Move(1,2)] then the order is this:
matched Move(1,2) insert Win(1)
matched Move(2,3) insert Win(2)
The insertion of Win(1) in the first firing does NOT prevent the instantiation with Move(2,3) from then firing.
So it's all good. :) Sample code and output attached.
--- On Thu, 4/16/09, Greg Barton <greg_barton@yahoo.com> wrote:
> From: Greg Barton <greg_barton@yahoo.com>
> Subject: Re: [rules-users] Negation semantics in Drools
> To: "Rules Users List" <rules-users@lists.jboss.org>> Date: Thursday, April 16, 2009, 8:50 PM
> It is on the latest snapshot release,
> 5.0.0.20090417.005612-483
>
> --- On Thu, 4/16/09, Edson Tirelli <tirelli@post.com>
> wrote:
>
> > We need to investigate if that is still happening
> in
> > latest trunk.
>
>
>
> _______________________________________________
> rules-users mailing list
> rules-users@lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
rules-users@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users