What is the best way to implement transitive closure/chaining using match and add-atom?

[CICS-Oleg]

Suppose we have the following knowledge base:

(Parent Tom Bob)
(Parent Pam Bob)
(Parent Tom Liz)
(Parent Bob Ann)
(Parent Bob Pat)
(Parent Pat Jim)
(Parent Jim Lil)

How to implement transitive closure through the loop of the match-based rules below in the most efficient way?

!(match &self (Parent $x $y) (add-atom &self (Predecessor $x $y)))
!(match &self (, (Parent $x $y) (Predecessor $y $z)) (add-atom &self (Predecessor $x $z)))

The main parts of the question are:

how to effectively check immutability of the space to stop the loop?

and how to effectively populate space with intermediate inference results without duplicates of what's been done on the previous steps?

For now I'm doing the last with the "function" like that (from this issue)

(= (not_exist $pattern)
    (unify &self $pattern Fasle True))

but it seems kinda awkward.

[CICS-Oleg]

And let's enhance our example above by a gender information:

(Person (Female Pam))
(Person (Female Liz))
(Person (Female Ann))
(Person (Female Lil))
(Person (Male Tom))
(Person (Male Bob))
(Person (Male Pat))
(Person (Male Jim))

Is the following the only way to inject a gender filter for predecessors to the rules above:

!(match &self (, (Predecessor $y $z) (Person (Female $y))) (add-atom &self (FemPredecessor $y $z)))

?

[vsbogd]

and how to effectively populate space with intermediate inference results without duplicates of what's been done on the previous steps?

Duplicates are possible because space implementation allows it. Allowing duplication of atoms is a default strategy for the GroundingSpace but not the only. On the other hand there is no public MeTTa API which allows creating spaces with "No duplication" strategy. If it solves the issue then it can be introduced quickly.

[vsbogd]

how to effectively check immutability of the space to stop the loop?

Not sure I understand the question. All spaces are mutable. What do you mean by immutability here, could you please elaborate?

[vsbogd]

Is the following the only way to inject a gender filter for predecessors to the rules above:

Again I am not sure I understand. What is drawback of the solution that you provided?

[CICS-Oleg]

Again I am not sure I understand. What is drawback of the solution that you provided?

The drawback is that we need new intermediate predicate FemPredecessor to not miss any results for already infered gender agnostic predicates Predecessor

[CICS-Oleg]

Not sure I understand the question. All spaces are mutable. What do you mean by immutability here, could you please elaborate?

You're right. Immutability is a wrong term. I meant how can we detect that space stopped to change from iteration to iteration.

[CICS-Oleg]

Duplicates are possible because space implementation allows it. Allowing duplication of atoms is a default strategy for the GroundingSpace but not the only.

Yeah, I understand this. It's ok.

What I'm trying to do now is

(= (not_exist $pattern)
    (unify &self $pattern Fasle True))

!(match &self (Parent $x $y) (if (not_exist (Predecessor $x $z)) (add-atom &self (Predecessor $x $y)) Empty))
!(match &self (, (Parent $x $y) (Predecessor $y $z)) (if (not_exist (Predecessor $x $z)) (add-atom &self (Predecessor $x $z)) Empty))
!(match &self (Parent $x $y) (if (not_exist (Predecessor $x $z)) (add-atom &self (Predecessor $x $y)) Empty))
!(match &self (, (Parent $x $y) (Predecessor $y $z)) (if (not_exist (Predecessor $x $z)) (add-atom &self (Predecessor $x $z)) Empty))

!(match &self (, (Predecessor $y $z) (Person (Female $y))) (if (not_exist (FemPredecessor $y $z)) (add-atom &self (FemPredecessor $y $z)) Empty))

!(match &self (Predecessor $y Jim) $y)

And I'm getting this as an output:

[(), (), (), (), (), (), ()]
[(if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Pam Ann)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Pam Pat)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Bob Jim)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Pat Lil)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Tom Ann)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Tom Pat)) Empty)]

The query

!(match &self (Predecessor $y Jim) $y)

works well. The output is

[Tom, Pam, Pat, Bob]

But those fragments are strange:

[(if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Pam Ann)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Pam Pat)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Bob Jim)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Pat Lil)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Tom Ann)) Empty), (if Fasle (add-atom ModuleSpace(GroundingSpace-top) (Predecessor Tom Pat)) Empty)]

[vsbogd]

One probable reason is that there is a typo Fasle instead of False.

[vsbogd]

Again I am not sure I understand. What is drawback of the solution that you provided?

The drawback is that we need new intermediate predicate FemPredecessor to not miss any results for already infered gender agnostic predicates Predecessor

Ok, I see, I don't understand the final goal (or task), thus it is difficult to me to bring a good solution.

[vsbogd]

Not sure I understand the question. All spaces are mutable. What do you mean by immutability here, could you please elaborate?

You're right. Immutability is a wrong term. I meant how can we detect that space stopped to change from iteration from iteration.

Ok, for this I also may suggest introducing returning flag from add-atom which says whether the value was really injected or not.

[CICS-Oleg]

One probable reason is that there is a typo Fasle instead of False.

Thanks!