Hi, I am using the "forall" keyword on the LHS and it seems the condition is satisfied when there is nothing to match. e.g. "forall (MyClass(value == "test"))" fires the rule when there is no MyClass() object in the working memory. My workaround is add "exists (MyClass())". Also, same for "not (exists (MyClass(value != "test")))", but I can understand this, as there is no MyClass() object, it doesn't exists and hence "not" gives true. But for "forall", it doesnt sound right to me. I wonder if "forall" is actually implemented as "not exists" in the engine. Please clarify if this is by design or a bug. Build is 5.5.0.Final Thank you -- View this message in context: http://drools.46999.n3.nabble.com/forall-is-satisfied-when-there-is-nothi... Sent from the Drools: User forum mailing list archive at Nabble.com. _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

First order logic does permit empty domains, and the universal quantifier evaluates to true in that case. Davide On 01/07/2014 11:11 AM, Wolfgang Laun wrote: > It is true that first-order logic usually assumes that the domain of a > formula > be a nonempty set. As so often, things aren't quite so simple when > formulae > are evaluated on a computer. What should be done in this case, forall > with an empty domain? Throw an exception? Not very convenient, since > there's no reasonable way of handling exceptions thrown on the LHS. > Return false? That doesn't make sense, because you can't inspect what > isn't there. Return true? If it isn't false - what else? > > If first-order logic does permit empty domains, it must be treated as > a special case. > > -W > > On 07/01/2014, Sonata <plz.write.to(a)gmail.com&gt; wrote: >> Hi, I am using the "forall" keyword on the LHS and it seems the >> condition >> is >> satisfied when there is nothing to match. e.g. "forall (MyClass(value == >> "test"))" fires the rule when there is no MyClass() object in the >> working >> memory. My workaround is add "exists (MyClass())". >> >> Also, same for "not (exists (MyClass(value != "test")))", but I can >> understand this, as there is no MyClass() object, it doesn't exists and >> hence "not" gives true. >> >> But for "forall", it doesnt sound right to me. I wonder if "forall" is >> actually implemented as "not exists" in the engine. >> >> Please clarify if this is by design or a bug. Build is 5.5.0.Final >> >> Thank you >> >> >> >> -- >> View this message in context: >> http://drools.46999.n3.nabble.com/forall-is-satisfied-when-there-is-nothi... >> Sent from the Drools: User forum mailing list archive at Nabble.com. >> _______________________________________________ >> rules-users mailing list >> rules-users(a)lists.jboss.org >> https://lists.jboss.org/mailman/listinfo/rules-users >> > _______________________________________________ > rules-users mailing list > rules-users(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/rules-users > _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

Indeed it is true by convention, see also http://en.wikipedia.org/wiki/Universal_quantification The only other alternative is to deprecate quantifiers altogether ;)

Davide On 01/07/2014 11:53 AM, Wolfgang Laun wrote: > Please comment on > http://en.wikipedia.org/wiki/First-order_logic#Empty_domains > -W > > On 07/01/2014, Davide Sottara <dsotty(a)gmail.com&gt; wrote: >> First order logic does permit empty domains, and the universal >> quantifier evaluates to true in that case. >> Davide >> >> On 01/07/2014 11:11 AM, Wolfgang Laun wrote: >>> It is true that first-order logic usually assumes that the domain of a >>> formula >>> be a nonempty set. As so often, things aren't quite so simple when >>> formulae >>> are evaluated on a computer. What should be done in this case, forall >>> with an empty domain? Throw an exception? Not very convenient, since >>> there's no reasonable way of handling exceptions thrown on the LHS. >>> Return false? That doesn't make sense, because you can't inspect what >>> isn't there. Return true? If it isn't false - what else? >>> >>> If first-order logic does permit empty domains, it must be treated as >>> a special case. >>> >>> -W >>> >>> On 07/01/2014, Sonata <plz.write.to(a)gmail.com&gt; wrote: >>>> Hi, I am using the "forall" keyword on the LHS and it seems the >>>> condition >>>> is >>>> satisfied when there is nothing to match. e.g. "forall (MyClass(value >>>> == >>>> "test"))" fires the rule when there is no MyClass() object in the >>>> working >>>> memory. My workaround is add "exists (MyClass())". >>>> >>>> Also, same for "not (exists (MyClass(value != "test")))", but I can >>>> understand this, as there is no MyClass() object, it doesn't exists >>>> and >>>> hence "not" gives true. >>>> >>>> But for "forall", it doesnt sound right to me. I wonder if "forall" is >>>> actually implemented as "not exists" in the engine. >>>> >>>> Please clarify if this is by design or a bug. Build is 5.5.0.Final >>>> >>>> Thank you >>>> >>>> >>>> >>>> -- >>>> View this message in context: >>>> http://drools.46999.n3.nabble.com/forall-is-satisfied-when-there-is-nothi... >>>> Sent from the Drools: User forum mailing list archive at Nabble.com. >>>> _______________________________________________ >>>> rules-users mailing list >>>> rules-users(a)lists.jboss.org >>>> https://lists.jboss.org/mailman/listinfo/rules-users >>>> >>> _______________________________________________ >>> rules-users mailing list >>> rules-users(a)lists.jboss.org >>> https://lists.jboss.org/mailman/listinfo/rules-users >>> >> _______________________________________________ >> rules-users mailing list >> rules-users(a)lists.jboss.org >> https://lists.jboss.org/mailman/listinfo/rules-users >> > _______________________________________________ > rules-users mailing list > rules-users(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/rules-users > _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

On 08/01/2014, Davide Sottara <dsotty(a)gmail.com&gt; wrote: > Indeed it is true by convention, see also > http://en.wikipedia.org/wiki/Universal_quantification > > The only other alternative is to deprecate quantifiers altogether ;) > Ah, *by convention*, yes. One should exercise some care with "vacuous truths", however. Uttering statements such as "All the little green men in my room are from Mars" may not brand you as a liar, but you could be called "batty". ;-) -W > Davide > > On 01/07/2014 11:53 AM, Wolfgang Laun wrote: >> Please comment on >> http://en.wikipedia.org/wiki/First-order_logic#Empty_domains >> -W >> >> On 07/01/2014, Davide Sottara <dsotty(a)gmail.com&gt; wrote: >>> First order logic does permit empty domains, and the universal >>> quantifier evaluates to true in that case. >>> Davide >>> >>> On 01/07/2014 11:11 AM, Wolfgang Laun wrote: >>>> It is true that first-order logic usually assumes that the domain of a >>>> formula >>>> be a nonempty set. As so often, things aren't quite so simple when >>>> formulae >>>> are evaluated on a computer. What should be done in this case, forall >>>> with an empty domain? Throw an exception? Not very convenient, since >>>> there's no reasonable way of handling exceptions thrown on the LHS. >>>> Return false? That doesn't make sense, because you can't inspect what >>>> isn't there. Return true? If it isn't false - what else? >>>> >>>> If first-order logic does permit empty domains, it must be treated as >>>> a special case. >>>> >>>> -W >>>> >>>> On 07/01/2014, Sonata <plz.write.to(a)gmail.com&gt; wrote: >>>>> Hi, I am using the "forall" keyword on the LHS and it seems the >>>>> condition >>>>> is >>>>> satisfied when there is nothing to match. e.g. "forall (MyClass(value >>>>> == >>>>> "test"))" fires the rule when there is no MyClass() object in the >>>>> working >>>>> memory. My workaround is add "exists (MyClass())". >>>>> >>>>> Also, same for "not (exists (MyClass(value != "test")))", but I can >>>>> understand this, as there is no MyClass() object, it doesn't exists >>>>> and >>>>> hence "not" gives true. >>>>> >>>>> But for "forall", it doesnt sound right to me. I wonder if "forall" is >>>>> actually implemented as "not exists" in the engine. >>>>> >>>>> Please clarify if this is by design or a bug. Build is 5.5.0.Final >>>>> >>>>> Thank you >>>>> >>>>> >>>>> >>>>> -- >>>>> View this message in context: >>>>> http://drools.46999.n3.nabble.com/forall-is-satisfied-when-there-is-nothi... >>>>> Sent from the Drools: User forum mailing list archive at Nabble.com. >>>>> _______________________________________________ >>>>> rules-users mailing list >>>>> rules-users(a)lists.jboss.org >>>>> https://lists.jboss.org/mailman/listinfo/rules-users >>>>> >>>> _______________________________________________ >>>> rules-users mailing list >>>> rules-users(a)lists.jboss.org >>>> https://lists.jboss.org/mailman/listinfo/rules-users >>>> >>> _______________________________________________ >>> rules-users mailing list >>> rules-users(a)lists.jboss.org >>> https://lists.jboss.org/mailman/listinfo/rules-users >>> >> _______________________________________________ >> rules-users mailing list >> rules-users(a)lists.jboss.org >> https://lists.jboss.org/mailman/listinfo/rules-users >> > _______________________________________________ > rules-users mailing list > rules-users(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/rules-users > _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

This is actually the way it is implemented internally, and no, I don't think that it would be appropriate to change it.

The "all and some" combination is actually quite popular and might be implemented at some point. By no means it is the only possibility to enhance the expressivity of the language: http://plato.stanford.edu/entries/generalized-quantifiers/ On 01/08/2014 01:50 PM, Sonata wrote: > Davide Sottara wrote >> This is actually the way it is implemented internally, and no, I don't >> think that >> it would be appropriate to change it. > Yes I agree, so may be we users actually do not need forall, but *every*, > which just means forall and exists > Now, look at it again: > when every Cloth( dried ) then collect() > See how pretty it is, simple beauty, fully expressive, just like a > sentence > :) > as oppose to > when forall Cloth ( dried ) AND exists Cloth() then collect() X( > > Nah, I guess people can live with that :P > > > > -- > View this message in context: > http://drools.46999.n3.nabble.com/forall-is-satisfied-when-there-is-nothi... > Sent from the Drools: User forum mailing list archive at Nabble.com. > _______________________________________________ > rules-users mailing list > rules-users(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/rules-users > _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

forall is implemented as two notes, see the Clips documentation here. In reality it is just a bit of sugar for those two notes. You might be able to achieve what you want, with your own combination of nots. http://www.csie.ntu.edu.tw/~sylee/courses/clips/bpg/node5.4.7.html Mark On 8 Jan 2014, at 11:32, Sonata <plz.write.to(a)gmail.com&gt; wrote: > laune wrote >> On 08/01/2014, Davide Sottara &lt; > >> dsotty@ > >> &gt; wrote: >>> Indeed it is true by convention, see also >>> http://en.wikipedia.org/wiki/Universal_quantification >>> >>> The only other alternative is to deprecate quantifiers altogether ;) >>> >> >> Ah, *by convention*, yes. One should exercise some care with "vacuous >> truths", however. Uttering statements such as "All the little green >> men in my room are from Mars" may not brand you as a liar, but you >> could be called "batty". ;-) >> >> -W >> >>> Davide >>> > > Guys, I am not sure whats the arguing point here. But from a programmer's > and a Drools engine user's perspective, I would find forall(void) to return > false more convenient. > > Say for example: > forall(Cloth(dried)) then collect() > Programmingwise, I dont want to execute collect() when the Cloth() > collection is empty and do extra null checkings. > Performancewise, I also dont want to call collect() when there is actually > nothing to perform. > In terms of semantics, I purposely want to check for isDried is true, it is > easy to realize from the LHS that I only want to collect clothes when all of > them are dried, rather than I want to collect also empty air. > > On the contrary, if I really want to do the RHS even though the LHS has > nothing, I would have written: > not( exists( Customer() ) ) then takeanap() > and I will not write > forall( Customer(status == "gone") ) then takeanap() > because the latter needs to do extra property visits and string comparisons > and it is so indirect to express what I really want. > > I really dont care what the mathematical definition is unless you tell me > Drools is a math. engine rather than a rule engine to express business > requirements. That, I should update myself and take a step back to look at > Drools again. Just my own opinion :P > > > > > -- > View this message in context: http://drools.46999.n3.nabble.com/forall-is-satisfied-when-there-is-nothi... > Sent from the Drools: User forum mailing list archive at Nabble.com. > _______________________________________________ > rules-users mailing list > rules-users(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/rules-users