[rules-users] 答复: Performance consideration in rule writing

Mon Dec 26 09:48:22 EST 2011

Hi, Wolfgang,

I think I¡¯ve figured out some answers and see my shares below. Thanks and
let me know your thoughts about #3.

Best

Abe

1.       Will I gain better performance if I put the rule differentiator
condition ¡°compareToPostThreshold($txn)==2¡± at the beginning of both rule
1 and 2? 

One kind pf Rete optimization is based on evaluating common constraints
once, therefore: no.

//Abe: it definitely makes a difference if you put differentiator conditions
at the beginning ¨C this way RETE won¡¯t waste efforts constructing networks
which will not fulfill. See below example.

rule "Evaluation of assignment in r-value position"
no-loop true
when
 $statement:CStatement($value:value)
 eval(Matcher.isRVExpression($value))

 eval(0==Matcher.getClauseNum())
then
 System.out.println($value);
 modify($statement){
  setSemantics(Matcher.getRVSemantics(memory,$value));
 }
end

rule "Evaluation of assignment clause in r-value position without updating
memory"
no-loop true
when
 $statement:CStatement($value:value)
 eval(Matcher.isRVExpression($value))

 eval(0<Matcher.getClauseNum())
then
 System.out.println("=="+$value);
 Matcher.decreaseClauseNum();
 modify($statement){
  setSemantics(Matcher.getRVSemanticsWithoutUpdate(memory,$value));
 }
end

These two rules try to validate if Matcher.getClauseNum() ==0 or <0. If you
put differentiator ¡°Matcher.getClauseNum()¡± at the end of LHS, Drools will
start two threads to deduct RETE network so even if the fact hits rule 1,
rule 2 will still be stepped in. But if we make the two rules as below, RETE
will identify the shortcut correctly and no thread will be triggered to step
into rule 2:

rule "Evaluation of assignment in r-value position"
no-loop true
when
 eval(0==Matcher.getClauseNum())
 $statement:CStatement($value:value)
 eval(Matcher.isRVExpression($value))
then
 System.out.println($value);
 modify($statement){
  setSemantics(Matcher.getRVSemantics(memory,$value));
 }
end

rule "Evaluation of assignment clause in r-value position without updating
memory"
no-loop true
when
 eval(0<Matcher.getClauseNum())
 $statement:CStatement($value:value)
 eval(Matcher.isRVExpression($value))
then
 System.out.println("=="+$value);
 Matcher.decreaseClauseNum();
 modify($statement){
  setSemantics(Matcher.getRVSemanticsWithoutUpdate(memory,$value));
 }
end

2.       I saw salaboy¡¯s video claiming that to avoid using eval() in the
rule. Do we have any alternative way to do that from a performance
consideration

Constraints based on fields using == are best. Other things may result in
eval-like evaluations anyway. Most of the time, it isn't eval that causes
performance setbacks.

or I¡¯d better collect/ prepare all the data before I send them into the
session?

Not clear what you mean by this, but if you can provide attributes that lend
themselves to straightforward constraints it might be worthwhile considering
some up-front processing of facts.
 //Abe: I saw below statement from Drools document 5.2.0. As Eval is not
indexed, overuse of evale reduces the rules¡¯ clarity and will result in a
bad performance.

5.8.3.7. Conditional Element eval

eval

Figure 5.29. eval

The conditional element eval is essentially a catch-all which allows any
semantic code (that returns a primitive boolean) to be executed. This code
can refer to variables that were bound in the LHS of the rule, and functions
in the rule package. Overuse of eval reduces the declarativeness of your
rules and can result in a poorly performing engine. While eval can be used
anywhere in the patterns, the best practice is to add it as the last
conditional element in the LHS of a rule.

Evals cannot be indexed and thus are not as efficient as Field Constraints.
However this makes them ideal for being used when functions return values
that change over time, which is not allowed within Field Constraints.

3.       What¡¯s you guys¡¯ naming convention for rule¡¯s salience?

Not clear what you mean by that.

//Abe: I mean how do you weight your salience values across different rules.
I¡¯ve seen various styles in my project ¨C somebody uses 100, 200, 300 but
somebody uses 90, 100, 110, 120, etc. This is not a big problem as they are
working on different rules and won¡¯t pollute each other. However I would
still try to make it consistent so maintain each other¡¯s rule files will be
easier¡

·¢¼þÈË: rules-users-bounces at lists.jboss.org
[mailto:rules-users-bounces at lists.jboss.org] ´ú±í Wolfgang Laun
·¢ËÍÊ±¼ä: 2011Äê12ÔÂ26ÈÕ 22:20
ÊÕ¼þÈË: Rules Users List
Ö÷Ìâ: Re: [rules-users] Performance consideration in rule writing

See below.

2011/12/26 Zhuo Li <milanello1998 at gmail.com>

Hi, team,

I have some quick questions here regarding performance best practices of
rule writing. See below two pieces of rules:

Rule ¡°1¡±

         Salience 100

         No-loop true

         When $txn : data(sourceid == 5&&txnjustify==¡±995¡±
&&eval(creditOption($txn)==1)&&eval(isGCSwitch($txn))&&isCurrencyEquals($txn
)==0&&compareToPostThreshold($txn)==2);

         Then

                   ¡

         End

Rule ¡°2¡±

         Salience 100

         No-loop true

         When $txn : data(sourceid == 5&&txnjustify==¡±995¡±
&&eval(creditOption($txn)==1)&&eval(isGCSwitch($txn))&&isCurrencyEquals($txn
)==0&&compareToPostThreshold($txn)==1);

         Then

                  ¡

         End

Questions: